diff --git a/.gitattributes b/.gitattributes index 7f1424434..865da2ca2 100644 --- a/.gitattributes +++ b/.gitattributes @@ -12,6 +12,11 @@ *.jpg binary *.gif binary *.ico binary +*.jpeg binary +*.mp3 binary +*.zip binary +*.bin binary + # Preserve original line endings for specific document files *.doc text eol=crlf diff --git a/.github/ISSUE_TEMPLATE/show_me_the_bug.md b/.github/ISSUE_TEMPLATE/show_me_the_bug.md index 504a2bd12..0c33f0319 100644 --- a/.github/ISSUE_TEMPLATE/show_me_the_bug.md +++ b/.github/ISSUE_TEMPLATE/show_me_the_bug.md @@ -19,6 +19,7 @@ - LLM type and model name: - System version: - Python version: +- MetaGPT version or branch: diff --git a/.github/workflows/build-package.yaml b/.github/workflows/build-package.yaml index 7f4fee53e..294a13f71 100644 --- a/.github/workflows/build-package.yaml +++ b/.github/workflows/build-package.yaml @@ -1,8 +1,9 @@ name: Build and upload python package on: + workflow_dispatch: release: - types: [created] + types: [created, published] jobs: deploy: diff --git a/.gitignore b/.gitignore index 6bc67fa61..aa5edd74a 100644 --- a/.gitignore +++ b/.gitignore @@ -1,7 +1,7 @@ ### Python template # Byte-compiled / optimized / DLL files -__pycache__/ +__pycache__ *.py[cod] *$py.class @@ -27,6 +27,8 @@ share/python-wheels/ .installed.cfg *.egg MANIFEST +metagpt/tools/schemas/ +examples/data/search_kb/*.json # PyInstaller # Usually these files are written by a python scripts from a template @@ -151,9 +153,14 @@ allure-results .vscode key.yaml -data +/data/ data.ms examples/nb/ +examples/default__vector_store.json +examples/docstore.json +examples/graph_store.json +examples/image__vector_store.json +examples/index_store.json .chroma *~$* workspace/* @@ -168,6 +175,7 @@ output tmp.png .dependencies.json tests/metagpt/utils/file_repo_git +tests/data/rsp_cache_new.json *.tmp *.png htmlcov @@ -178,4 +186,5 @@ cov.xml *.faiss *-structure.csv *-structure.json -metagpt/tools/schemas \ No newline at end of file +*.dot +.python-version diff --git a/MANIFEST.in b/MANIFEST.in new file mode 100644 index 000000000..292433f80 --- /dev/null +++ b/MANIFEST.in @@ -0,0 +1,3 @@ +recursive-include metagpt/ext/stanford_town/prompts *.txt +recursive-include metagpt/ext/stanford_town/static_dirs *.csv +recursive-include metagpt/ext/stanford_town/static_dirs *.json \ No newline at end of file diff --git a/README.md b/README.md index 72ce03412..edb2066a3 100644 --- a/README.md +++ b/README.md @@ -26,7 +26,7 @@ # MetaGPT: The Multi-Agent Framework

## News -🚀 March. 01, 2024: Our Data Interpreter paper is on arxiv. Find all design and benchmark details [here](https://arxiv.org/abs/2402.18679)! +🚀 Mar. 14, 2024: Our **Data Interpreter** paper is on [arxiv](https://arxiv.org/abs/2402.18679). Check the [example](https://docs.deepwisdom.ai/main/en/DataInterpreter/) and [code](https://github.com/geekan/MetaGPT/tree/main/examples/di)! 🚀 Feb. 08, 2024: [v0.7.0](https://github.com/geekan/MetaGPT/releases/tag/v0.7.0) released, supporting assigning different LLMs to different Roles. We also introduced [Data Interpreter](https://github.com/geekan/MetaGPT/blob/main/examples/di/README.md), a powerful agent capable of solving a wide range of real-world problems. @@ -55,21 +55,30 @@ ## Software Company as Multi-Agent System

Software Company Multi-Agent Schematic (Gradually Implementing)

-## Install +## Get Started -### Pip installation +### Installation > Ensure that Python 3.9+ is installed on your system. You can check this by using: `python --version`. > You can use conda like this: `conda create -n metagpt python=3.9 && conda activate metagpt` ```bash -pip install metagpt -# https://docs.deepwisdom.ai/main/en/guide/get_started/configuration.html -metagpt --init-config # it will create ~/.metagpt/config2.yaml, just modify it to your needs +pip install --upgrade metagpt +# or `pip install --upgrade git+https://github.com/geekan/MetaGPT.git` +# or `git clone https://github.com/geekan/MetaGPT && cd MetaGPT && pip install --upgrade -e .` ``` +For detailed installation guidance, please refer to [cli_install](https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html#install-stable-version) + or [docker_install](https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html#install-with-docker) + ### Configuration +You can init the config of MetaGPT by running the following command, or manually create `~/.metagpt/config2.yaml` file: +```bash +# Check https://docs.deepwisdom.ai/main/en/guide/get_started/configuration.html for more details +metagpt --init-config # it will create ~/.metagpt/config2.yaml, just modify it to your needs +``` + You can configure `~/.metagpt/config2.yaml` according to the [example](https://github.com/geekan/MetaGPT/blob/main/config/config2.example.yaml) and [doc](https://docs.deepwisdom.ai/main/en/guide/get_started/configuration.html): ```yaml @@ -82,13 +91,13 @@ ### Configuration ### Usage -After installation, you can use it as CLI +After installation, you can use MetaGPT at CLI ```bash metagpt "Create a 2048 game" # this will create a repo in ./workspace ``` -or you can use it as library +or use it as library ```python from metagpt.software_company import generate_repo, ProjectRepo @@ -96,47 +105,19 @@ ### Usage print(repo) # it will print the repo structure with files ``` -detail installation please refer to [cli_install](https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html#install-stable-version) - or [docker_install](https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html#install-with-docker) +You can also use its [Data Interpreter](https://github.com/geekan/MetaGPT/tree/main/examples/di) -### Docker installation -
⏬ Step 1: Download metagpt image and prepare config2.yaml :: click to expand :: -
+```python +import asyncio +from metagpt.roles.di.data_interpreter import DataInterpreter -```bash -docker pull metagpt/metagpt:latest -mkdir -p /opt/metagpt/{config,workspace} -docker run --rm metagpt/metagpt:latest cat /app/metagpt/config/config2.yaml > /opt/metagpt/config/config2.yaml -vim /opt/metagpt/config/config2.yaml # Change the config +async def main(): + di = DataInterpreter() + await di.run("Run data analysis on sklearn Iris dataset, include a plot") + +asyncio.run(main()) # or await main() in a jupyter notebook setting ``` -
-
- -
⏬ Step 2: Run metagpt container :: click to expand :: -
- -```bash -docker run --name metagpt -d \ - --privileged \ - -v /opt/metagpt/config/config2.yaml:/app/metagpt/config/config2.yaml \ - -v /opt/metagpt/workspace:/app/metagpt/workspace \ - metagpt/metagpt:latest -``` - -
-
- -
⏬ Step 3: Use metagpt :: click to expand :: -
- -```bash -docker exec -it metagpt /bin/bash -$ metagpt "Create a 2048 game" # this will create a repo in ./workspace -``` - -
-
### QuickStart & Demo Video - Try it on [MetaGPT Huggingface Space](https://huggingface.co/spaces/deepwisdom/MetaGPT) @@ -156,6 +137,7 @@ ## Tutorial - 🧑‍💻 Contribution - [Develop Roadmap](docs/ROADMAP.md) - 🔖 Use Cases + - [Data Interpreter](https://docs.deepwisdom.ai/main/en/guide/use_cases/agent/interpreter/intro.html) - [Debate](https://docs.deepwisdom.ai/main/en/guide/use_cases/multi_agent/debate.html) - [Researcher](https://docs.deepwisdom.ai/main/en/guide/use_cases/agent/researcher.html) - [Recepit Assistant](https://docs.deepwisdom.ai/main/en/guide/use_cases/agent/receipt_assistant.html) @@ -179,7 +161,9 @@ ### Contact Information ## Citation -For now, cite the [arXiv paper](https://arxiv.org/abs/2308.00352): +To stay updated with the latest research and development, follow [@MetaGPT_](https://twitter.com/MetaGPT_) on Twitter. + +To cite [MetaGPT](https://arxiv.org/abs/2308.00352) or [Data Interpreter](https://arxiv.org/abs/2402.18679) in publications, please use the following BibTeX entries. ```bibtex @misc{hong2023metagpt, @@ -190,4 +174,14 @@ ## Citation archivePrefix={arXiv}, primaryClass={cs.AI} } +@misc{hong2024data, + title={Data Interpreter: An LLM Agent For Data Science}, + author={Sirui Hong and Yizhang Lin and Bang Liu and Bangbang Liu and Binhao Wu and Danyang Li and Jiaqi Chen and Jiayi Zhang and Jinlin Wang and Li Zhang and Lingyao Zhang and Min Yang and Mingchen Zhuge and Taicheng Guo and Tuo Zhou and Wei Tao and Wenyi Wang and Xiangru Tang and Xiangtao Lu and Xiawu Zheng and Xinbing Liang and Yaying Fei and Yuheng Cheng and Zongze Xu and Chenglin Wu}, + year={2024}, + eprint={2402.18679}, + archivePrefix={arXiv}, + primaryClass={cs.AI} +} + ``` + diff --git a/SECURITY.md b/SECURITY.md index 623e6e556..924ce5015 100644 --- a/SECURITY.md +++ b/SECURITY.md @@ -4,9 +4,9 @@ ## Supported Versions | Version | Supported | |---------|--------------------| - | 7.x | :x: | - | 6.x | :x: | -| < 6.x | :x: | + | 0.7.x | :x: | + | 0.6.x | :x: | +| < 0.6.x | :x: | ## Reporting a Vulnerability diff --git a/config/config2.example.yaml b/config/config2.example.yaml index 3a5cc3585..c5454ec32 100644 --- a/config/config2.example.yaml +++ b/config/config2.example.yaml @@ -4,6 +4,7 @@ llm: api_key: "YOUR_API_KEY" model: "gpt-4-turbo-preview" # or gpt-3.5-turbo-1106 / gpt-4-1106-preview proxy: "YOUR_PROXY" # for LLM API requests + # timeout: 600 # Optional. If set to 0, default value is 300. pricing_plan: "" # Optional. If invalid, it will be automatically filled in with the value of the `model`. # Azure-exclusive pricing plan mappings： # - gpt-3.5-turbo 4k: "gpt-3.5-turbo-1106" diff --git a/docs/README_CN.md b/docs/README_CN.md index 7a0db4974..8aea5e4cb 100644 --- a/docs/README_CN.md +++ b/docs/README_CN.md @@ -116,7 +116,7 @@ ### 联系信息 ## 引用 -引用 [arXiv paper](https://arxiv.org/abs/2308.00352): +如果您在研究论文中使用 MetaGPT 或 Data Interpreter，请引用我们的工作： ```bibtex @misc{hong2023metagpt, @@ -127,4 +127,12 @@ ## 引用 archivePrefix={arXiv}, primaryClass={cs.AI} } +@misc{hong2024data, + title={Data Interpreter: An LLM Agent For Data Science}, + author={Sirui Hong and Yizhang Lin and Bang Liu and Bangbang Liu and Binhao Wu and Danyang Li and Jiaqi Chen and Jiayi Zhang and Jinlin Wang and Li Zhang and Lingyao Zhang and Min Yang and Mingchen Zhuge and Taicheng Guo and Tuo Zhou and Wei Tao and Wenyi Wang and Xiangru Tang and Xiangtao Lu and Xiawu Zheng and Xinbing Liang and Yaying Fei and Yuheng Cheng and Zongze Xu and Chenglin Wu}, + year={2024}, + eprint={2402.18679}, + archivePrefix={arXiv}, + primaryClass={cs.AI} +} ``` diff --git a/docs/README_JA.md b/docs/README_JA.md index c6b99461c..91155532b 100644 --- a/docs/README_JA.md +++ b/docs/README_JA.md @@ -295,7 +295,7 @@ ## クイックスタート ## 引用 -現時点では、[arXiv 論文](https://arxiv.org/abs/2308.00352)を引用してください: +研究論文でMetaGPTやData Interpreterを使用する場合は、以下のように当社の作業を引用してください： ```bibtex @misc{hong2023metagpt, @@ -306,6 +306,14 @@ ## 引用 archivePrefix={arXiv}, primaryClass={cs.AI} } +@misc{hong2024data, + title={Data Interpreter: An LLM Agent For Data Science}, + author={Sirui Hong and Yizhang Lin and Bang Liu and Bangbang Liu and Binhao Wu and Danyang Li and Jiaqi Chen and Jiayi Zhang and Jinlin Wang and Li Zhang and Lingyao Zhang and Min Yang and Mingchen Zhuge and Taicheng Guo and Tuo Zhou and Wei Tao and Wenyi Wang and Xiangru Tang and Xiangtao Lu and Xiawu Zheng and Xinbing Liang and Yaying Fei and Yuheng Cheng and Zongze Xu and Chenglin Wu}, + year={2024}, + eprint={2402.18679}, + archivePrefix={arXiv}, + primaryClass={cs.AI} +} ``` ## お問い合わせ先 diff --git a/examples/data/rag/travel.txt b/examples/data/rag/travel.txt new file mode 100644 index 000000000..f72ad5c59 --- /dev/null +++ b/examples/data/rag/travel.txt @@ -0,0 +1 @@ +Bob likes traveling. \ No newline at end of file diff --git a/examples/data/rag/writer.txt b/examples/data/rag/writer.txt new file mode 100644 index 000000000..1dc055901 --- /dev/null +++ b/examples/data/rag/writer.txt @@ -0,0 +1,109 @@ +Productivity +I think I am at least somewhat more productive than average, and people sometimes ask me for productivity tips. So I decided to just write them all down in one place. + +Compound growth gets discussed as a financial concept, but it works in careers as well, and it is magic. A small productivity gain, compounded over 50 years, is worth a lot. So it’s worth figuring out how to optimize productivity. If you get 10% more done and 1% better every day compared to someone else, the compounded difference is massive. + +What you work on + +Famous writers have some essential qualities, creativity and discipline + +It doesn’t matter how fast you move if it’s in a worthless direction. Picking the right thing to work on is the most important element of productivity and usually almost ignored. So think about it more! Independent thought is hard but it’s something you can get better at with practice. + +The most impressive people I know have strong beliefs about the world, which is rare in the general population. If you find yourself always agreeing with whomever you last spoke with, that’s bad. You will of course be wrong sometimes, but develop the confidence to stick with your convictions. It will let you be courageous when you’re right about something important that most people don’t see. + +I make sure to leave enough time in my schedule to think about what to work on. The best ways for me to do this are reading books, hanging out with interesting people, and spending time in nature. + +I’ve learned that I can’t be very productive working on things I don’t care about or don’t like. So I just try not to put myself in a position where I have to do them (by delegating, avoiding, or something else). Stuff that you don’t like is a painful drag on morale and momentum. + +By the way, here is an important lesson about delegation: remember that everyone else is also most productive when they’re doing what they like, and do what you’d want other people to do for you—try to figure out who likes (and is good at) doing what, and delegate that way. + +If you find yourself not liking what you’re doing for a long period of time, seriously consider a major job change. Short-term burnout happens, but if it isn’t resolved with some time off, maybe it’s time to do something you’re more interested in. + +I’ve been very fortunate to find work I like so much I’d do it for free, which makes it easy to be really productive. + +It’s important to learn that you can learn anything you want, and that you can get better quickly. This feels like an unlikely miracle the first few times it happens, but eventually you learn to trust that you can do it. + +Doing great work usually requires colleagues of some sort. Try to be around smart, productive, happy, and positive people that don’t belittle your ambitions. I love being around people who push me and inspire me to be better. To the degree you able to, avoid the opposite kind of people—the cost of letting them take up your mental cycles is horrific. + +You have to both pick the right problem and do the work. There aren’t many shortcuts. If you’re going to do something really important, you are very likely going to work both smart and hard. The biggest prizes are heavily competed for. This isn’t true in every field (there are great mathematicians who never spend that many hours a week working) but it is in most. + +Prioritization + +Writers have to work hard to be successful + +My system has three key pillars: “Make sure to get the important shit done”, “Don’t waste time on stupid shit”, and “make a lot of lists”. + +I highly recommend using lists. I make lists of what I want to accomplish each year, each month, and each day. Lists are very focusing, and they help me with multitasking because I don’t have to keep as much in my head. If I’m not in the mood for some particular task, I can always find something else I’m excited to do. + +I prefer lists written down on paper. It’s easy to add and remove tasks. I can access them during meetings without feeling rude. I re-transcribe lists frequently, which forces me to think about everything on the list and gives me an opportunity to add and remove items. + +I don’t bother with categorization or trying to size tasks or anything like that (the most I do is put a star next to really important items). + +I try to prioritize in a way that generates momentum. The more I get done, the better I feel, and then the more I get done. I like to start and end each day with something I can really make progress on. + +I am relentless about getting my most important projects done—I’ve found that if I really want something to happen and I push hard enough, it usually happens. + +I try to be ruthless about saying no to stuff, and doing non-critical things in the quickest way possible. I probably take this too far—for example, I am almost sure I am terse to the point of rudeness when replying to emails. + +Passion and adaptability are key qualities to writers + +I generally try to avoid meetings and conferences as I find the time cost to be huge—I get the most value out of time in my office. However, it is critical that you keep enough space in your schedule to allow for chance encounters and exposure to new people and ideas. Having an open network is valuable; though probably 90% of the random meetings I take are a waste of time, the other 10% really make up for it. + +I find most meetings are best scheduled for 15-20 minutes, or 2 hours. The default of 1 hour is usually wrong, and leads to a lot of wasted time. + +I have different times of day I try to use for different kinds of work. The first few hours of the morning are definitely my most productive time of the day, so I don’t let anyone schedule anything then. I try to do meetings in the afternoon. I take a break, or switch tasks, whenever I feel my attention starting to fade. + +I don’t think most people value their time enough—I am surprised by the number of people I know who make $100 an hour and yet will spend a couple of hours doing something they don’t want to do to save $20. + +Also, don’t fall into the trap of productivity porn—chasing productivity for its own sake isn’t helpful. Many people spend too much time thinking about how to perfectly optimize their system, and not nearly enough asking if they’re working on the right problems. It doesn’t matter what system you use or if you squeeze out every second if you’re working on the wrong thing. + +The right goal is to allocate your year optimally, not your day. + +Physical factors + +Very likely what is optimal for me won’t be optimal for you. You’ll have to experiment to find out what works best for your body. It’s definitely worth doing—it helps in all aspects of life, and you’ll feel a lot better and happier overall. + +It probably took a little bit of my time every week for a few years to arrive at what works best for me, but my sense is if I do a good job at all the below I’m at least 1.5x more productive than if not. + +Sleep seems to be the most important physical factor in productivity for me. Some sort of sleep tracker to figure out how to sleep best is helpful. I’ve found the only thing I’m consistent with are in the set-it-and-forget-it category, and I really like the Emfit QS+Active. + +I like a cold, dark, quiet room, and a great mattress (I resisted spending a bunch of money on a great mattress for years, which was stupid—it makes a huge difference to my sleep quality. I love this one). Not eating a lot in the few hours before sleep helps. Not drinking alcohol helps a lot, though I’m not willing to do that all the time. + +I use a Chili Pad to be cold while I sleep if I can’t get the room cold enough, which is great but loud (I set it up to have the cooler unit outside my room). + +When traveling, I use an eye mask and ear plugs. + +Writers usually have empathy to write good books. + +This is likely to be controversial, but I take a low dose of sleeping pills (like a third of a normal dose) or a very low dose of cannabis whenever I can’t sleep. I am a bad sleeper in general, and a particularly bad sleeper when I travel. It likely has tradeoffs, but so does not sleeping well. If you can already sleep well, I wouldn’t recommend this. + +I use a full spectrum LED light most mornings for about 10-15 minutes while I catch up on email. It’s great—if you try nothing else in here, this is the thing I’d try. It’s a ridiculous gain for me. I like this one, and it’s easy to travel with. + +Exercise is probably the second most important physical factor. I tried a number of different exercise programs for a few months each and the one that seemed best was lifting heavy weights 3x a week for an hour, and high intensity interval training occasionally. In addition to productivity gains, this is also the exercise program that makes me feel the best overall. + +The third area is nutrition. I very rarely eat breakfast, so I get about 15 hours of fasting most days (except an espresso when I wake up). I know this is contrary to most advice, and I suspect it’s not optimal for most people, but it definitely works well for me. + +Eating lots of sugar is the thing that makes me feel the worst and that I try hardest to avoid. I also try to avoid foods that aggravate my digestion or spike up inflammation (for example, very spicy foods). I don’t have much willpower when it comes to sweet things, so I mostly just try to keep junk food out of the house. + +I have one big shot of espresso immediately when I wake up and one after lunch. I assume this is about 200mg total of caffeine per day. I tried a few other configurations; this was the one that worked by far the best. I otherwise aggressively avoid stimulants, but I will have more coffee if I’m super tired and really need to get something done. + +If a writer want to be super, then should include innovative thinking. + +I’m vegetarian and have been since I was a kid, and I supplement methyl B-12, Omega-3, Iron, and Vitamin D-3. I got to this list with a year or so of quarterly blood tests; it’s worked for me ever since (I re-test maybe every year and a half or so). There are many doctors who will happily work with you on a super comprehensive blood test (and services like WellnessFX). I also go out of my way to drink a lot of protein shakes, which I hate and I wouldn’t do if I weren’t vegetarian. + +Other stuff + +Here’s what I like in a workspace: natural light, quiet, knowing that I won’t be interrupted if I don’t want to be, long blocks of time, and being comfortable and relaxed (I’ve got a beautiful desk with a couple of 4k monitors on it in my office, but I spend almost all my time on my couch with my laptop). + +I wrote custom software for the annoying things I have to do frequently, which is great. I also made an effort to learn to type really fast and the keyboard shortcuts that help with my workflow. + +Like most people, I sometimes go through periods of a week or two where I just have no motivation to do anything (I suspect it may have something to do with nutrition). This sucks and always seems to happen at inconvenient times. I have not figured out what to do about it besides wait for the fog to lift, and to trust that eventually it always does. And I generally try to avoid people and situations that put me in bad moods, which is good advice whether you care about productivity or not. + +In general, I think it’s good to overcommit a little bit. I find that I generally get done what I take on, and if I have a little bit too much to do it makes me more efficient at everything, which is a way to train to avoid distractions (a great habit to build!). However, overcommitting a lot is disastrous. + +Don’t neglect your family and friends for the sake of productivity—that’s a very stupid tradeoff (and very likely a net productivity loss, because you’ll be less happy). Don’t neglect doing things you love or that clear your head either. + +Finally, to repeat one more time: productivity in the wrong direction isn’t worth anything at all. Think more about what to work on. + +Open-Mindedness and curiosity are essential to writers + diff --git a/examples/example.json b/examples/data/search_kb/example.json similarity index 100% rename from examples/example.json rename to examples/data/search_kb/example.json diff --git a/examples/example.xlsx b/examples/data/search_kb/example.xlsx similarity index 100% rename from examples/example.xlsx rename to examples/data/search_kb/example.xlsx diff --git a/examples/di/arxiv_reader.py b/examples/di/arxiv_reader.py new file mode 100644 index 000000000..6e1939b81 --- /dev/null +++ b/examples/di/arxiv_reader.py @@ -0,0 +1,21 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +from metagpt.roles.di.data_interpreter import DataInterpreter + + +async def main(): + template = "https://arxiv.org/list/{tag}/pastweek?skip=0&show=300" + tags = ["cs.ai", "cs.cl", "cs.lg", "cs.se"] + urls = [template.format(tag=tag) for tag in tags] + prompt = f"""This is a collection of arxiv urls: '{urls}' . +Record each article, remove duplicates by title (they may have multiple tags), filter out papers related to +large language model / agent / llm, print top 100 and visualize the word count of the titles""" + di = DataInterpreter(react_mode="react", tools=["scrape_web_playwright"]) + + await di.run(prompt) + + +if __name__ == "__main__": + import asyncio + + asyncio.run(main()) diff --git a/examples/di/custom_tool.py b/examples/di/custom_tool.py new file mode 100644 index 000000000..cbe7380c7 --- /dev/null +++ b/examples/di/custom_tool.py @@ -0,0 +1,36 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +""" +@Time : 2024/3/22 10:54 +@Author : alexanderwu +@File : custom_tool.py +""" + +from metagpt.roles.di.data_interpreter import DataInterpreter +from metagpt.tools.tool_registry import register_tool + + +@register_tool() +def magic_function(arg1: str, arg2: int) -> dict: + """ + The magic function that does something. + + Args: + arg1 (str): ... + arg2 (int): ... + + Returns: + dict: ... + """ + return {"arg1": arg1 * 3, "arg2": arg2 * 5} + + +async def main(): + di = DataInterpreter(tools=["magic_function"]) + await di.run("Just call the magic function with arg1 'A' and arg2 2. Tell me the result.") + + +if __name__ == "__main__": + import asyncio + + asyncio.run(main()) diff --git a/examples/di/data_visualization.py b/examples/di/data_visualization.py index 1a21ab7cb..184e04f26 100644 --- a/examples/di/data_visualization.py +++ b/examples/di/data_visualization.py @@ -1,14 +1,17 @@ import asyncio +from metagpt.logs import logger from metagpt.roles.di.data_interpreter import DataInterpreter +from metagpt.utils.recovery_util import save_history async def main(requirement: str = ""): di = DataInterpreter() - await di.run(requirement) + rsp = await di.run(requirement) + logger.info(rsp) + save_history(role=di) if __name__ == "__main__": requirement = "Run data analysis on sklearn Iris dataset, include a plot" - asyncio.run(main(requirement)) diff --git a/examples/di/machine_learning.py b/examples/di/machine_learning.py index a58735831..c674e66e8 100644 --- a/examples/di/machine_learning.py +++ b/examples/di/machine_learning.py @@ -2,11 +2,21 @@ import fire from metagpt.roles.di.data_interpreter import DataInterpreter +WINE_REQ = "Run data analysis on sklearn Wine recognition dataset, include a plot, and train a model to predict wine class (20% as validation), and show validation accuracy." -async def main(auto_run: bool = True): - requirement = "Run data analysis on sklearn Wine recognition dataset, include a plot, and train a model to predict wine class (20% as validation), and show validation accuracy." - di = DataInterpreter(auto_run=auto_run) - await di.run(requirement) +DATA_DIR = "path/to/your/data" +# sales_forecast data from https://www.kaggle.com/datasets/aslanahmedov/walmart-sales-forecast/data +SALES_FORECAST_REQ = f"""Train a model to predict sales for each department in every store (split the last 40 weeks records as validation dataset, the others is train dataset), include plot total sales trends, print metric and plot scatter plots of +groud truth and predictions on validation data. Dataset is {DATA_DIR}/train.csv, the metric is weighted mean absolute error (WMAE) for test data. Notice: *print* key variables to get more information for next task step. +""" + +REQUIREMENTS = {"wine": WINE_REQ, "sales_forecast": SALES_FORECAST_REQ} + + +async def main(use_case: str = "wine"): + mi = DataInterpreter() + requirement = REQUIREMENTS[use_case] + await mi.run(requirement) if __name__ == "__main__": diff --git a/examples/rag_pipeline.py b/examples/rag_pipeline.py new file mode 100644 index 000000000..b5111b75c --- /dev/null +++ b/examples/rag_pipeline.py @@ -0,0 +1,248 @@ +"""RAG pipeline""" + +import asyncio + +from pydantic import BaseModel + +from metagpt.const import DATA_PATH, EXAMPLE_DATA_PATH +from metagpt.logs import logger +from metagpt.rag.engines import SimpleEngine +from metagpt.rag.schema import ( + BM25RetrieverConfig, + ChromaIndexConfig, + ChromaRetrieverConfig, + ElasticsearchIndexConfig, + ElasticsearchRetrieverConfig, + ElasticsearchStoreConfig, + FAISSRetrieverConfig, + LLMRankerConfig, +) +from metagpt.utils.exceptions import handle_exception + +DOC_PATH = EXAMPLE_DATA_PATH / "rag/writer.txt" +QUESTION = "What are key qualities to be a good writer?" + +TRAVEL_DOC_PATH = EXAMPLE_DATA_PATH / "rag/travel.txt" +TRAVEL_QUESTION = "What does Bob like?" + +LLM_TIP = "If you not sure, just answer I don't know." + + +class Player(BaseModel): + """To demonstrate rag add objs.""" + + name: str = "" + goal: str = "Win The 100-meter Sprint." + tool: str = "Red Bull Energy Drink." + + def rag_key(self) -> str: + """For search""" + return self.goal + + +class RAGExample: + """Show how to use RAG.""" + + def __init__(self, engine: SimpleEngine = None): + self._engine = engine + + @property + def engine(self): + if not self._engine: + self._engine = SimpleEngine.from_docs( + input_files=[DOC_PATH], + retriever_configs=[FAISSRetrieverConfig(), BM25RetrieverConfig()], + ranker_configs=[LLMRankerConfig()], + ) + return self._engine + + @engine.setter + def engine(self, value: SimpleEngine): + self._engine = value + + async def run_pipeline(self, question=QUESTION, print_title=True): + """This example run rag pipeline, use faiss&bm25 retriever and llm ranker, will print something like: + + Retrieve Result: + 0. Productivi..., 10.0 + 1. I wrote cu..., 7.0 + 2. I highly r..., 5.0 + + Query Result: + Passion, adaptability, open-mindedness, creativity, discipline, and empathy are key qualities to be a good writer. + """ + if print_title: + self._print_title("Run Pipeline") + + nodes = await self.engine.aretrieve(question) + self._print_retrieve_result(nodes) + + answer = await self.engine.aquery(question) + self._print_query_result(answer) + + async def add_docs(self): + """This example show how to add docs. + + Before add docs llm anwser I don't know. + After add docs llm give the correct answer, will print something like: + + [Before add docs] + Retrieve Result: + + Query Result: + Empty Response + + [After add docs] + Retrieve Result: + 0. Bob like..., 10.0 + + Query Result: + Bob likes traveling. + """ + self._print_title("Add Docs") + + travel_question = f"{TRAVEL_QUESTION}{LLM_TIP}" + travel_filepath = TRAVEL_DOC_PATH + + logger.info("[Before add docs]") + await self.run_pipeline(question=travel_question, print_title=False) + + logger.info("[After add docs]") + self.engine.add_docs([travel_filepath]) + await self.run_pipeline(question=travel_question, print_title=False) + + @handle_exception + async def add_objects(self, print_title=True): + """This example show how to add objects. + + Before add docs, engine retrieve nothing. + After add objects, engine give the correct answer, will print something like: + + [Before add objs] + Retrieve Result: + + [After add objs] + Retrieve Result: + 0. 100m Sprin..., 10.0 + + [Object Detail] + {'name': 'Mike', 'goal': 'Win The 100-meter Sprint', 'tool': 'Red Bull Energy Drink'} + """ + if print_title: + self._print_title("Add Objects") + + player = Player(name="Mike") + question = f"{player.rag_key()}" + + logger.info("[Before add objs]") + await self._retrieve_and_print(question) + + logger.info("[After add objs]") + self.engine.add_objs([player]) + + try: + nodes = await self._retrieve_and_print(question) + + logger.info("[Object Detail]") + player: Player = nodes[0].metadata["obj"] + logger.info(player.name) + except Exception as e: + logger.error(f"nodes is empty, llm don't answer correctly, exception: {e}") + + async def init_objects(self): + """This example show how to from objs, will print something like: + + Same as add_objects. + """ + self._print_title("Init Objects") + + pre_engine = self.engine + self.engine = SimpleEngine.from_objs(retriever_configs=[FAISSRetrieverConfig()]) + await self.add_objects(print_title=False) + self.engine = pre_engine + + async def init_and_query_chromadb(self): + """This example show how to use chromadb. how to save and load index. will print something like: + + Query Result: + Bob likes traveling. + """ + self._print_title("Init And Query ChromaDB") + + # 1. save index + output_dir = DATA_PATH / "rag" + SimpleEngine.from_docs( + input_files=[TRAVEL_DOC_PATH], + retriever_configs=[ChromaRetrieverConfig(persist_path=output_dir)], + ) + + # 2. load index + engine = SimpleEngine.from_index(index_config=ChromaIndexConfig(persist_path=output_dir)) + + # 3. query + answer = await engine.aquery(TRAVEL_QUESTION) + self._print_query_result(answer) + + @handle_exception + async def init_and_query_es(self): + """This example show how to use es. how to save and load index. will print something like: + + Query Result: + Bob likes traveling. + """ + self._print_title("Init And Query Elasticsearch") + + # 1. create es index and save docs + store_config = ElasticsearchStoreConfig(index_name="travel", es_url="http://127.0.0.1:9200") + engine = SimpleEngine.from_docs( + input_files=[TRAVEL_DOC_PATH], + retriever_configs=[ElasticsearchRetrieverConfig(store_config=store_config)], + ) + + # 2. load index + engine = SimpleEngine.from_index(index_config=ElasticsearchIndexConfig(store_config=store_config)) + + # 3. query + answer = await engine.aquery(TRAVEL_QUESTION) + self._print_query_result(answer) + + @staticmethod + def _print_title(title): + logger.info(f"{'#'*30} {title} {'#'*30}") + + @staticmethod + def _print_retrieve_result(result): + """Print retrieve result.""" + logger.info("Retrieve Result:") + + for i, node in enumerate(result): + logger.info(f"{i}. {node.text[:10]}..., {node.score}") + + logger.info("") + + @staticmethod + def _print_query_result(result): + """Print query result.""" + logger.info("Query Result:") + + logger.info(f"{result}\n") + + async def _retrieve_and_print(self, question): + nodes = await self.engine.aretrieve(question) + self._print_retrieve_result(nodes) + return nodes + + +async def main(): + """RAG pipeline""" + e = RAGExample() + await e.run_pipeline() + await e.add_docs() + await e.add_objects() + await e.init_objects() + await e.init_and_query_chromadb() + await e.init_and_query_es() + + +if __name__ == "__main__": + asyncio.run(main()) diff --git a/examples/rag_search.py b/examples/rag_search.py new file mode 100644 index 000000000..258c5ba60 --- /dev/null +++ b/examples/rag_search.py @@ -0,0 +1,21 @@ +"""Agent with RAG search.""" + +import asyncio + +from examples.rag_pipeline import DOC_PATH, QUESTION +from metagpt.logs import logger +from metagpt.rag.engines import SimpleEngine +from metagpt.roles import Sales + + +async def search(): + """Agent with RAG search.""" + + store = SimpleEngine.from_docs(input_files=[DOC_PATH]) + role = Sales(profile="Sales", store=store) + result = await role.run(QUESTION) + logger.info(result) + + +if __name__ == "__main__": + asyncio.run(search()) diff --git a/examples/search_kb.py b/examples/search_kb.py deleted file mode 100644 index 995720cc1..000000000 --- a/examples/search_kb.py +++ /dev/null @@ -1,33 +0,0 @@ -#!/usr/bin/env python -# -*- coding: utf-8 -*- -""" -@File : search_kb.py -@Modified By: mashenquan, 2023-12-22. Delete useless codes. -""" -import asyncio - -from langchain.embeddings import OpenAIEmbeddings - -from metagpt.config2 import config -from metagpt.const import DATA_PATH, EXAMPLE_PATH -from metagpt.document_store import FaissStore -from metagpt.logs import logger -from metagpt.roles import Sales - - -def get_store(): - llm = config.get_openai_llm() - embedding = OpenAIEmbeddings(openai_api_key=llm.api_key, openai_api_base=llm.base_url) - return FaissStore(DATA_PATH / "example.json", embedding=embedding) - - -async def search(): - store = FaissStore(EXAMPLE_PATH / "example.json") - role = Sales(profile="Sales", store=store) - query = "Which facial cleanser is good for oily skin?" - result = await role.run(query) - logger.info(result) - - -if __name__ == "__main__": - asyncio.run(search()) diff --git a/examples/search_with_specific_engine.py b/examples/search_with_specific_engine.py index 1eee762d5..276431ed8 100644 --- a/examples/search_with_specific_engine.py +++ b/examples/search_with_specific_engine.py @@ -13,7 +13,7 @@ async def main(): question = "What are the most interesting human facts?" search = Config.default().search - kwargs = {"api_key": search.api_key, "cse_id": search.cse_id, "proxy": None} + kwargs = search.model_dump() await Searcher(search_engine=SearchEngine(engine=search.api_type, **kwargs)).run(question) diff --git a/metagpt/environment/android_env/__init__.py b/examples/stanford_town/__init__.py similarity index 100% rename from metagpt/environment/android_env/__init__.py rename to examples/stanford_town/__init__.py diff --git a/examples/stanford_town/requirements.txt b/examples/stanford_town/requirements.txt new file mode 100644 index 000000000..e69de29bb diff --git a/examples/stanford_town/run_st_game.py b/examples/stanford_town/run_st_game.py new file mode 100644 index 000000000..732ae5efc --- /dev/null +++ b/examples/stanford_town/run_st_game.py @@ -0,0 +1,93 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : entry of Stanford Town(ST/st) game + +import asyncio +from typing import Optional + +import fire + +from metagpt.ext.stanford_town.roles.st_role import STRole +from metagpt.ext.stanford_town.stanford_town import StanfordTown +from metagpt.ext.stanford_town.utils.const import STORAGE_PATH +from metagpt.ext.stanford_town.utils.mg_ga_transform import ( + get_reverie_meta, + write_curr_sim_code, + write_curr_step, +) +from metagpt.ext.stanford_town.utils.utils import copy_folder +from metagpt.logs import logger + + +async def startup( + idea: str, fork_sim_code: str, sim_code: str, temp_storage_path: str, investment: float = 30.0, n_round: int = 500 +): + town = StanfordTown() + logger.info("StanfordTown init environment") + + # copy `storage/{fork_sim_code}` to `storage/{sim_code}` + copy_folder(str(STORAGE_PATH.joinpath(fork_sim_code)), str(STORAGE_PATH.joinpath(sim_code))) + + # get role names from `storage/{simulation_name}/reverie/meta.json` and then init roles + reverie_meta = get_reverie_meta(fork_sim_code) + roles = [] + sim_path = STORAGE_PATH.joinpath(sim_code) + sim_path.mkdir(exist_ok=True) + for idx, role_name in enumerate(reverie_meta["persona_names"]): + has_inner_voice = True if idx == 0 else False + role = STRole( + name=role_name, + profile=role_name, + sim_code=sim_code, + step=reverie_meta.get("step", 0), + start_time=reverie_meta.get("start_date"), + curr_time=reverie_meta.get("curr_time"), + sec_per_step=reverie_meta.get("sec_per_step"), + has_inner_voice=has_inner_voice, + ) + roles.append(role) + + # init temp_storage + write_curr_sim_code({"sim_code": sim_code}, temp_storage_path) + write_curr_step({"step": reverie_meta.get("step", 0)}, temp_storage_path) + + await town.hire(roles) + + town.invest(investment) + town.run_project(idea) + + await town.run(n_round) + + +def main( + idea: str, + fork_sim_code: str, + sim_code: str, + temp_storage_path: Optional[str] = None, + investment: float = 30.0, + n_round: int = 500, +): + """ + Args: + idea: idea works as an `inner voice` to the first agent. + fork_sim_code: old simulation name to start with, choose one inside `generative_agents/environment/frontend_server/storage/` + sim_code: new simulation name to save simulation result + temp_storage_path: generative_agents temp_storage path inside `environment/frontend_server` to interact. + investment: the investment of running agents + n_round: rounds to run agents + """ + + asyncio.run( + startup( + idea=idea, + fork_sim_code=fork_sim_code, + sim_code=sim_code, + temp_storage_path=temp_storage_path, + investment=investment, + n_round=n_round, + ) + ) + + +if __name__ == "__main__": + fire.Fire(main) diff --git a/examples/stanford_town/storage/.gitignore b/examples/stanford_town/storage/.gitignore new file mode 100644 index 000000000..962820861 --- /dev/null +++ b/examples/stanford_town/storage/.gitignore @@ -0,0 +1,4 @@ +# path to store simulation data +test_* +unittest* +July* \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/environment/0.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/environment/0.json new file mode 100644 index 000000000..6eaa46c51 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/environment/0.json @@ -0,0 +1,26 @@ +{ + "Isabella Rodriguez": { + "maze": "the_ville", + "x": 72, + "y": 14 + }, + "Klaus Mueller": { + "maze": "the_ville", + "x": 126, + "y": 46 + }, + "Maria Lopez": { + "maze": "the_ville", + "x": 123, + "y": 57 + } +} + + + + + + + + + diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/embeddings.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/embeddings.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/embeddings.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/kw_strength.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/kw_strength.json new file mode 100644 index 000000000..6dc73c1c8 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/kw_strength.json @@ -0,0 +1,2 @@ +{"kw_strength_event": {}, + "kw_strength_thought": {}} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/nodes.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/nodes.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/associative_memory/nodes.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/scratch.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/scratch.json new file mode 100644 index 000000000..dbed4b705 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/scratch.json @@ -0,0 +1,51 @@ +{ + "vision_r": 8, + "att_bandwidth": 8, + "retention": 8, + "curr_time": null, + "curr_tile": null, + "daily_plan_req": "Isabella Rodriguez opens Hobbs Cafe at 8am everyday, and works at the counter until 8pm, at which point she closes the cafe.", + "name": "Isabella Rodriguez", + "first_name": "Isabella", + "last_name": "Rodriguez", + "age": 34, + "innate": "friendly, outgoing, hospitable", + "learned": "Isabella Rodriguez is a cafe owner of Hobbs Cafe who loves to make people feel welcome. She is always looking for ways to make the cafe a place where people can come to relax and enjoy themselves.", + "currently": "Isabella Rodriguez is planning on having a Valentine's Day party at Hobbs Cafe with her customers on February 14th, 2023 at 5pm. She is gathering party material, and is telling everyone to join the party at Hobbs Cafe on February 14th, 2023, from 5pm to 7pm.", + "lifestyle": "Isabella Rodriguez goes to bed around 11pm, awakes up around 6am.", + "living_area": "the Ville:Isabella Rodriguez's apartment:main room", + "concept_forget": 100, + "daily_reflection_time": 180, + "daily_reflection_size": 5, + "overlap_reflect_th": 4, + "kw_strg_event_reflect_th": 10, + "kw_strg_thought_reflect_th": 9, + + "recency_w": 1, + "relevance_w": 1, + "importance_w": 1, + "recency_decay": 0.995, + "importance_trigger_max": 150, + "importance_trigger_curr": 150, + "importance_ele_n": 0, + "thought_count": 5, + + "daily_req": [], + "f_daily_schedule": [], + "f_daily_schedule_hourly_org": [], + "act_address": null, + "act_start_time": null, + "act_duration": null, + "act_description": null, + "act_pronunciatio": null, + "act_event": ["Isabella Rodriguez", null, null], + "act_obj_description": null, + "act_obj_pronunciatio": null, + "act_obj_event": [null, null, null], + "chatting_with": null, + "chat": null, + "chatting_with_buffer": {}, + "chatting_end_time": null, + "act_path_set": false, + "planned_path": [] +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/spatial_memory.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/spatial_memory.json new file mode 100644 index 000000000..f88157950 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Isabella Rodriguez/bootstrap_memory/spatial_memory.json @@ -0,0 +1,66 @@ +{ + "the Ville": { + "Hobbs Cafe": { + "cafe": [ + "refrigerator", + "cafe customer seating", + "cooking area", + "kitchen sink", + "behind the cafe counter", + "piano" + ] + }, + "Isabella Rodriguez's apartment": { + "main room": [ + "bed", + "desk", + "refrigerator", + "closet", + "shelf" + ] + }, + "The Rose and Crown Pub": { + "pub": [ + "shelf", + "refrigerator", + "bar customer seating", + "behind the bar counter", + "kitchen sink", + "cooking area", + "microphone" + ] + }, + "Harvey Oak Supply Store": { + "supply store": [ + "supply store product shelf", + "behind the supply store counter", + "supply store counter" + ] + }, + "The Willows Market and Pharmacy": { + "store": [ + "behind the pharmacy counter", + "pharmacy store shelf", + "pharmacy store counter", + "grocery store shelf", + "behind the grocery counter", + "grocery store counter" + ] + }, + "Dorm for Oak Hill College": { + "garden": [ + "dorm garden" + ], + "common room": [ + "common room sofa", + "pool table", + "common room table" + ] + }, + "Johnson Park": { + "park": [ + "park garden" + ] + } + } +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/embeddings.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/embeddings.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/embeddings.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/kw_strength.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/kw_strength.json new file mode 100644 index 000000000..6dc73c1c8 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/kw_strength.json @@ -0,0 +1,2 @@ +{"kw_strength_event": {}, + "kw_strength_thought": {}} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/nodes.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/nodes.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/associative_memory/nodes.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/scratch.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/scratch.json new file mode 100644 index 000000000..7b0ce7d72 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/scratch.json @@ -0,0 +1,51 @@ +{ + "vision_r": 8, + "att_bandwidth": 8, + "retention": 8, + "curr_time": null, + "curr_tile": null, + "daily_plan_req": "Klaus Mueller goes to the library at Oak Hill College early in the morning, spends his days writing, and eats at Hobbs Cafe.", + "name": "Klaus Mueller", + "first_name": "Klaus", + "last_name": "Mueller", + "age": 20, + "innate": "kind, inquisitive, passionate", + "learned": "Klaus Mueller is a student at Oak Hill College studying sociology. He is passionate about social justice and loves to explore different perspectives.", + "currently": "Klaus Mueller is writing a research paper on the effects of gentrification in low-income communities.", + "lifestyle": "Klaus Mueller goes to bed around 11pm, awakes up around 7am, eats dinner around 5pm.", + "living_area": "the Ville:Dorm for Oak Hill College:Klaus Mueller's room", + "concept_forget": 100, + "daily_reflection_time": 180, + "daily_reflection_size": 5, + "overlap_reflect_th": 4, + "kw_strg_event_reflect_th": 10, + "kw_strg_thought_reflect_th": 9, + + "recency_w": 1, + "relevance_w": 1, + "importance_w": 1, + "recency_decay": 0.99, + "importance_trigger_max": 150, + "importance_trigger_curr": 150, + "importance_ele_n": 0, + "thought_count": 5, + + "daily_req": [], + "f_daily_schedule": [], + "f_daily_schedule_hourly_org": [], + "act_address": null, + "act_start_time": null, + "act_duration": null, + "act_description": null, + "act_pronunciatio": null, + "act_event": ["Klaus Mueller", null, null], + "act_obj_description": null, + "act_obj_pronunciatio": null, + "act_obj_event": [null, null, null], + "chatting_with": null, + "chat": null, + "chatting_with_buffer": {}, + "chatting_end_time": null, + "act_path_set": false, + "planned_path": [] +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/spatial_memory.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/spatial_memory.json new file mode 100644 index 000000000..4f4168677 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Klaus Mueller/bootstrap_memory/spatial_memory.json @@ -0,0 +1,86 @@ +{ + "the Ville": { + "Oak Hill College": { + "hallway": [], + "library": [ + "library sofa", + "library table", + "bookshelf" + ], + "classroom": [ + "blackboard", + "classroom podium", + "classroom student seating" + ] + }, + "Dorm for Oak Hill College": { + "garden": [ + "dorm garden" + ], + "Klaus Mueller's room": [ + "bed", + "game console", + "closet", + "desk" + ], + "woman's bathroom": [ + "toilet", + "shower", + "bathroom sink" + ], + "common room": [ + "common room sofa", + "pool table", + "common room table" + ], + "man's bathroom": [ + "shower", + "bathroom sink", + "toilet" + ] + }, + "The Willows Market and Pharmacy": { + "store": [ + "grocery store shelf", + "behind the grocery counter", + "grocery store counter", + "pharmacy store shelf", + "pharmacy store counter", + "behind the pharmacy counter" + ] + }, + "Harvey Oak Supply Store": { + "supply store": [ + "supply store product shelf", + "behind the supply store counter", + "supply store counter" + ] + }, + "Johnson Park": { + "park": [ + "park garden" + ] + }, + "The Rose and Crown Pub": { + "pub": [ + "shelf", + "refrigerator", + "bar customer seating", + "behind the bar counter", + "kitchen sink", + "cooking area", + "microphone" + ] + }, + "Hobbs Cafe": { + "cafe": [ + "refrigerator", + "cafe customer seating", + "cooking area", + "kitchen sink", + "behind the cafe counter", + "piano" + ] + } + } +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/embeddings.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/embeddings.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/embeddings.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/kw_strength.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/kw_strength.json new file mode 100644 index 000000000..6dc73c1c8 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/kw_strength.json @@ -0,0 +1,2 @@ +{"kw_strength_event": {}, + "kw_strength_thought": {}} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/nodes.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/nodes.json new file mode 100644 index 000000000..9e26dfeeb --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/associative_memory/nodes.json @@ -0,0 +1 @@ +{} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/scratch.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/scratch.json new file mode 100644 index 000000000..c3a304952 --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/scratch.json @@ -0,0 +1,51 @@ +{ + "vision_r": 8, + "att_bandwidth": 8, + "retention": 8, + "curr_time": null, + "curr_tile": null, + "daily_plan_req": "Maria Lopez spends at least 3 hours a day Twitch streaming or gaming.", + "name": "Maria Lopez", + "first_name": "Maria", + "last_name": "Lopez", + "age": 21, + "innate": "energetic, enthusiastic, inquisitive", + "learned": "Maria Lopez is a student at Oak Hill College studying physics and a part time Twitch game streamer who loves to connect with people and explore new ideas.", + "currently": "Maria Lopez is working on her physics degree and streaming games on Twitch to make some extra money. She visits Hobbs Cafe for studying and eating just about everyday.", + "lifestyle": "Maria Lopez goes to bed around 2am, awakes up around 9am, eats dinner around 6pm. She likes to hang out at Hobbs Cafe if it's before 6pm.", + "living_area": "the Ville:Dorm for Oak Hill College:Maria Lopez's room", + "concept_forget": 100, + "daily_reflection_time": 180, + "daily_reflection_size": 5, + "overlap_reflect_th": 4, + "kw_strg_event_reflect_th": 10, + "kw_strg_thought_reflect_th": 9, + + "recency_w": 1, + "relevance_w": 1, + "importance_w": 1, + "recency_decay": 0.99, + "importance_trigger_max": 150, + "importance_trigger_curr": 150, + "importance_ele_n": 0, + "thought_count": 5, + + "daily_req": [], + "f_daily_schedule": [], + "f_daily_schedule_hourly_org": [], + "act_address": null, + "act_start_time": null, + "act_duration": null, + "act_description": null, + "act_pronunciatio": null, + "act_event": ["Maria Lopez", null, null], + "act_obj_description": null, + "act_obj_pronunciatio": null, + "act_obj_event": [null, null, null], + "chatting_with": null, + "chat": null, + "chatting_with_buffer": {}, + "chatting_end_time": null, + "act_path_set": false, + "planned_path": [] +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/spatial_memory.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/spatial_memory.json new file mode 100644 index 000000000..0a58212bd --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/personas/Maria Lopez/bootstrap_memory/spatial_memory.json @@ -0,0 +1,87 @@ +{ + "the Ville": { + "Oak Hill College": { + "hallway": [], + "library": [ + "library sofa", + "library table", + "bookshelf" + ], + "classroom": [ + "blackboard", + "classroom podium", + "classroom student seating" + ] + }, + "Dorm for Oak Hill College": { + "garden": [ + "dorm garden" + ], + "Maria Lopez's room": [ + "closet", + "desk", + "bed", + "computer", + "blackboard" + ], + "woman's bathroom": [ + "toilet", + "shower", + "bathroom sink" + ], + "common room": [ + "common room sofa", + "pool table", + "common room table" + ], + "man's bathroom": [ + "shower", + "bathroom sink", + "toilet" + ] + }, + "The Willows Market and Pharmacy": { + "store": [ + "grocery store shelf", + "behind the grocery counter", + "grocery store counter", + "pharmacy store shelf", + "pharmacy store counter", + "behind the pharmacy counter" + ] + }, + "Harvey Oak Supply Store": { + "supply store": [ + "supply store product shelf", + "behind the supply store counter", + "supply store counter" + ] + }, + "Johnson Park": { + "park": [ + "park garden" + ] + }, + "The Rose and Crown Pub": { + "pub": [ + "shelf", + "refrigerator", + "bar customer seating", + "behind the bar counter", + "kitchen sink", + "cooking area", + "microphone" + ] + }, + "Hobbs Cafe": { + "cafe": [ + "refrigerator", + "cafe customer seating", + "cooking area", + "kitchen sink", + "behind the cafe counter", + "piano" + ] + } + } +} \ No newline at end of file diff --git a/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/reverie/meta.json b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/reverie/meta.json new file mode 100644 index 000000000..1e81ec12d --- /dev/null +++ b/examples/stanford_town/storage/base_the_ville_isabella_maria_klaus/reverie/meta.json @@ -0,0 +1,13 @@ +{ + "fork_sim_code": "base_the_ville_isabella_maria_klaus", + "start_date": "February 13, 2023", + "curr_time": "February 13, 2023, 00:00:00", + "sec_per_step": 10, + "maze_name": "the_ville", + "persona_names": [ + "Isabella Rodriguez", + "Maria Lopez", + "Klaus Mueller" + ], + "step": 0 +} \ No newline at end of file diff --git a/metagpt/actions/action_node.py b/metagpt/actions/action_node.py index 09da4a988..31e4cc0fc 100644 --- a/metagpt/actions/action_node.py +++ b/metagpt/actions/action_node.py @@ -17,6 +17,7 @@ from pydantic import BaseModel, Field, create_model, model_validator from tenacity import retry, stop_after_attempt, wait_random_exponential from metagpt.actions.action_outcls_registry import register_action_outcls +from metagpt.const import USE_CONFIG_TIMEOUT from metagpt.llm import BaseLLM from metagpt.logs import logger from metagpt.provider.postprocess.llm_output_postprocess import llm_output_postprocess @@ -330,7 +331,7 @@ class ActionNode: def compile_to(self, i: Dict, schema, kv_sep) -> str: if schema == "json": - return json.dumps(i, indent=4) + return json.dumps(i, indent=4, ensure_ascii=False) elif schema == "markdown": return dict_to_markdown(i, kv_sep=kv_sep) else: @@ -339,10 +340,7 @@ class ActionNode: def tagging(self, text, schema, tag="") -> str: if not tag: return text - if schema == "json": - return f"[{tag}]\n" + text + f"\n[/{tag}]" - else: # markdown - return f"[{tag}]\n" + text + f"\n[/{tag}]" + return f"[{tag}]\n{text}\n[/{tag}]" def _compile_f(self, schema, mode, tag, format_func, kv_sep, exclude=None) -> str: nodes = self.to_dict(format_func=format_func, mode=mode, exclude=exclude) @@ -374,7 +372,7 @@ class ActionNode: schema="markdown": 编译context, example(markdown), instruction(markdown), constraint, action """ if schema == "raw": - return context + "\n\n## Actions\n" + LANGUAGE_CONSTRAINT + "\n" + self.instruction + return f"{context}\n\n## Actions\n{LANGUAGE_CONSTRAINT}\n{self.instruction}" ### 直接使用 pydantic BaseModel 生成 instruction 与 example，仅限 JSON # child_class = self._create_children_class() @@ -416,7 +414,7 @@ class ActionNode: images: Optional[Union[str, list[str]]] = None, system_msgs: Optional[list[str]] = None, schema="markdown", # compatible to original format - timeout=3, + timeout=USE_CONFIG_TIMEOUT, ) -> (str, BaseModel): """Use ActionOutput to wrap the output of aask""" content = await self.llm.aask(prompt, system_msgs, images=images, timeout=timeout) @@ -448,7 +446,9 @@ class ActionNode: def set_context(self, context): self.set_recursive("context", context) - async def simple_fill(self, schema, mode, images: Optional[Union[str, list[str]]] = None, timeout=3, exclude=None): + async def simple_fill( + self, schema, mode, images: Optional[Union[str, list[str]]] = None, timeout=USE_CONFIG_TIMEOUT, exclude=None + ): prompt = self.compile(context=self.context, schema=schema, mode=mode, exclude=exclude) if schema != "raw": @@ -473,7 +473,7 @@ class ActionNode: mode="auto", strgy="simple", images: Optional[Union[str, list[str]]] = None, - timeout=3, + timeout=USE_CONFIG_TIMEOUT, exclude=[], ): """Fill the node(s) with mode. diff --git a/metagpt/actions/di/execute_nb_code.py b/metagpt/actions/di/execute_nb_code.py index f6a8defbd..0cf16b70f 100644 --- a/metagpt/actions/di/execute_nb_code.py +++ b/metagpt/actions/di/execute_nb_code.py @@ -57,8 +57,23 @@ class ExecuteNbCode(Action): async def terminate(self): """kill NotebookClient""" - if self.nb_client.km is not None: - await self.nb_client._async_cleanup_kernel() + if self.nb_client.km is not None and await self.nb_client.km.is_alive(): + await self.nb_client.km.shutdown_kernel(now=True) + await self.nb_client.km.cleanup_resources() + + channels = [ + self.nb_client.kc.stdin_channel, # The channel for handling standard input to the kernel. + self.nb_client.kc.hb_channel, # The channel for heartbeat communication between the kernel and client. + self.nb_client.kc.control_channel, # The channel for controlling the kernel. + ] + + # Stops all the running channels for this kernel + for channel in channels: + if channel.is_alive(): + channel.stop() + + self.nb_client.kc = None + self.nb_client.km = None async def reset(self): """reset NotebookClient""" diff --git a/metagpt/actions/di/write_analysis_code.py b/metagpt/actions/di/write_analysis_code.py index 185926e31..711e56d39 100644 --- a/metagpt/actions/di/write_analysis_code.py +++ b/metagpt/actions/di/write_analysis_code.py @@ -18,7 +18,7 @@ from metagpt.prompts.di.write_analysis_code import ( STRUCTUAL_PROMPT, ) from metagpt.schema import Message, Plan -from metagpt.utils.common import CodeParser, process_message, remove_comments +from metagpt.utils.common import CodeParser, remove_comments class WriteAnalysisCode(Action): @@ -50,7 +50,7 @@ class WriteAnalysisCode(Action): ) working_memory = working_memory or [] - context = process_message([Message(content=structual_prompt, role="user")] + working_memory) + context = self.llm.format_msg([Message(content=structual_prompt, role="user")] + working_memory) # LLM call if use_reflection: diff --git a/metagpt/actions/research.py b/metagpt/actions/research.py index ce8d8a967..2a99a8d99 100644 --- a/metagpt/actions/research.py +++ b/metagpt/actions/research.py @@ -134,7 +134,7 @@ class CollectLinks(Action): break model_name = config.llm.model - prompt = reduce_message_length(gen_msg(), model_name, system_text, 4096) + prompt = reduce_message_length(gen_msg(), model_name, system_text, config.llm.max_token) logger.debug(prompt) queries = await self._aask(prompt, [system_text]) try: diff --git a/metagpt/config2.py b/metagpt/config2.py index bc6af18c6..f3273419f 100644 --- a/metagpt/config2.py +++ b/metagpt/config2.py @@ -92,7 +92,7 @@ class Config(CLIParams, YamlModel): """ default_config_paths: List[Path] = [ METAGPT_ROOT / "config/config2.yaml", - Path.home() / ".metagpt/config2.yaml", + CONFIG_ROOT / "config2.yaml", ] dicts = [dict(os.environ)] @@ -100,6 +100,20 @@ class Config(CLIParams, YamlModel): final = merge_dict(dicts) return Config(**final) + @classmethod + def from_llm_config(cls, llm_config: dict): + """user config llm + example: + llm_config = {"api_type": "xxx", "api_key": "xxx", "model": "xxx"} + gpt4 = Config.from_llm_config(llm_config) + A = Role(name="A", profile="Democratic candidate", goal="Win the election", actions=[a1], watch=[a2], config=gpt4) + """ + llm_config = LLMConfig.model_validate(llm_config) + dicts = [dict(os.environ)] + dicts += [{"llm": llm_config}] + final = merge_dict(dicts) + return Config(**final) + def update_via_cli(self, project_path, project_name, inc, reqa_file, max_auto_summarize_code): """update config via cli""" diff --git a/metagpt/configs/llm_config.py b/metagpt/configs/llm_config.py index 77bfc8466..af8f56372 100644 --- a/metagpt/configs/llm_config.py +++ b/metagpt/configs/llm_config.py @@ -10,6 +10,7 @@ from typing import Optional from pydantic import field_validator +from metagpt.const import LLM_API_TIMEOUT from metagpt.utils.yaml_model import YamlModel @@ -29,6 +30,7 @@ class LLMType(Enum): DASHSCOPE = "dashscope" # Aliyun LingJi DashScope MOONSHOT = "moonshot" MISTRAL = "mistral" + YI = "yi" # lingyiwanwu def __missing__(self, key): return self.OPENAI @@ -73,7 +75,7 @@ class LLMConfig(YamlModel): stream: bool = False logprobs: Optional[bool] = None # https://cookbook.openai.com/examples/using_logprobs top_logprobs: Optional[int] = None - timeout: int = 60 + timeout: int = 600 # For Network proxy: Optional[str] = None @@ -87,3 +89,8 @@ class LLMConfig(YamlModel): if v in ["", None, "YOUR_API_KEY"]: raise ValueError("Please set your API key in config2.yaml") return v + + @field_validator("timeout") + @classmethod + def check_timeout(cls, v): + return v or LLM_API_TIMEOUT diff --git a/metagpt/configs/search_config.py b/metagpt/configs/search_config.py index af928b02a..e28b14c99 100644 --- a/metagpt/configs/search_config.py +++ b/metagpt/configs/search_config.py @@ -7,6 +7,8 @@ """ from typing import Callable, Optional +from pydantic import Field + from metagpt.tools import SearchEngineType from metagpt.utils.yaml_model import YamlModel @@ -18,3 +20,11 @@ class SearchConfig(YamlModel): api_key: str = "" cse_id: str = "" # for google search_func: Optional[Callable] = None + params: dict = Field( + default_factory=lambda: { + "engine": "google", + "google_domain": "google.com", + "gl": "us", + "hl": "en", + } + ) diff --git a/metagpt/const.py b/metagpt/const.py index a5e3ea9c2..e4cebfd96 100644 --- a/metagpt/const.py +++ b/metagpt/const.py @@ -49,6 +49,7 @@ METAGPT_ROOT = get_metagpt_root() # Dependent on METAGPT_PROJECT_ROOT DEFAULT_WORKSPACE_ROOT = METAGPT_ROOT / "workspace" EXAMPLE_PATH = METAGPT_ROOT / "examples" +EXAMPLE_DATA_PATH = EXAMPLE_PATH / "data" DATA_PATH = METAGPT_ROOT / "data" TEST_DATA_PATH = METAGPT_ROOT / "tests/data" RESEARCH_PATH = DATA_PATH / "research" @@ -122,7 +123,6 @@ BASE64_FORMAT = "base64" # REDIS REDIS_KEY = "REDIS_KEY" -LLM_API_TIMEOUT = 300 # Message id IGNORED_MESSAGE_ID = "0" @@ -131,3 +131,7 @@ IGNORED_MESSAGE_ID = "0" GENERALIZATION = "Generalize" COMPOSITION = "Composite" AGGREGATION = "Aggregate" + +# Timeout +USE_CONFIG_TIMEOUT = 0 # Using llm.timeout configuration. +LLM_API_TIMEOUT = 300 diff --git a/metagpt/document.py b/metagpt/document.py index 1a3c4b2f8..4a8bb68d5 100644 --- a/metagpt/document.py +++ b/metagpt/document.py @@ -11,12 +11,9 @@ from pathlib import Path from typing import Optional, Union import pandas as pd -from langchain.text_splitter import CharacterTextSplitter -from langchain_community.document_loaders import ( - TextLoader, - UnstructuredPDFLoader, - UnstructuredWordDocumentLoader, -) +from llama_index.core import Document, SimpleDirectoryReader +from llama_index.core.node_parser import SimpleNodeParser +from llama_index.readers.file import PDFReader from pydantic import BaseModel, ConfigDict, Field from tqdm import tqdm @@ -29,7 +26,7 @@ def validate_cols(content_col: str, df: pd.DataFrame): raise ValueError("Content column not found in DataFrame.") -def read_data(data_path: Path): +def read_data(data_path: Path) -> Union[pd.DataFrame, list[Document]]: suffix = data_path.suffix if ".xlsx" == suffix: data = pd.read_excel(data_path) @@ -38,14 +35,13 @@ def read_data(data_path: Path): elif ".json" == suffix: data = pd.read_json(data_path) elif suffix in (".docx", ".doc"): - data = UnstructuredWordDocumentLoader(str(data_path), mode="elements").load() + data = SimpleDirectoryReader(input_files=[str(data_path)]).load_data() elif ".txt" == suffix: - data = TextLoader(str(data_path)).load() - text_splitter = CharacterTextSplitter(separator="\n", chunk_size=256, chunk_overlap=0) - texts = text_splitter.split_documents(data) - data = texts + data = SimpleDirectoryReader(input_files=[str(data_path)]).load_data() + node_parser = SimpleNodeParser.from_defaults(separator="\n", chunk_size=256, chunk_overlap=0) + data = node_parser.get_nodes_from_documents(data) elif ".pdf" == suffix: - data = UnstructuredPDFLoader(str(data_path), mode="elements").load() + data = PDFReader.load_data(str(data_path)) else: raise NotImplementedError("File format not supported.") return data @@ -150,9 +146,9 @@ class IndexableDocument(Document): metadatas.append({}) return docs, metadatas - def _get_docs_and_metadatas_by_langchain(self) -> (list, list): + def _get_docs_and_metadatas_by_llamaindex(self) -> (list, list): data = self.data - docs = [i.page_content for i in data] + docs = [i.text for i in data] metadatas = [i.metadata for i in data] return docs, metadatas @@ -160,7 +156,7 @@ class IndexableDocument(Document): if isinstance(self.data, pd.DataFrame): return self._get_docs_and_metadatas_by_df() elif isinstance(self.data, list): - return self._get_docs_and_metadatas_by_langchain() + return self._get_docs_and_metadatas_by_llamaindex() else: raise NotImplementedError("Data type not supported for metadata extraction.") diff --git a/metagpt/document_store/base_store.py b/metagpt/document_store/base_store.py index ddc1d626b..6aafc57bb 100644 --- a/metagpt/document_store/base_store.py +++ b/metagpt/document_store/base_store.py @@ -38,9 +38,9 @@ class LocalStore(BaseStore, ABC): if not self.store: self.store = self.write() - def _get_index_and_store_fname(self, index_ext=".index", pkl_ext=".pkl"): - index_file = self.cache_dir / f"{self.fname}{index_ext}" - store_file = self.cache_dir / f"{self.fname}{pkl_ext}" + def _get_index_and_store_fname(self, index_ext=".json", docstore_ext=".json"): + index_file = self.cache_dir / "default__vector_store" / index_ext + store_file = self.cache_dir / "docstore" / docstore_ext return index_file, store_file @abstractmethod diff --git a/metagpt/document_store/chromadb_store.py b/metagpt/document_store/chromadb_store.py index d7344d41b..1d3a014ee 100644 --- a/metagpt/document_store/chromadb_store.py +++ b/metagpt/document_store/chromadb_store.py @@ -11,9 +11,9 @@ import chromadb class ChromaStore: """If inherited from BaseStore, or importing other modules from metagpt, a Python exception occurs, which is strange.""" - def __init__(self, name): + def __init__(self, name: str, get_or_create: bool = False): client = chromadb.Client() - collection = client.create_collection(name) + collection = client.create_collection(name, get_or_create=get_or_create) self.client = client self.collection = collection diff --git a/metagpt/document_store/faiss_store.py b/metagpt/document_store/faiss_store.py index 2359917d5..b196bef27 100644 --- a/metagpt/document_store/faiss_store.py +++ b/metagpt/document_store/faiss_store.py @@ -7,10 +7,14 @@ """ import asyncio from pathlib import Path -from typing import Optional +from typing import Any, Optional -from langchain.vectorstores import FAISS -from langchain_core.embeddings import Embeddings +import faiss +from llama_index.core import VectorStoreIndex, load_index_from_storage +from llama_index.core.embeddings import BaseEmbedding +from llama_index.core.schema import Document, QueryBundle, TextNode +from llama_index.core.storage import StorageContext +from llama_index.vector_stores.faiss import FaissVectorStore from metagpt.document import IndexableDocument from metagpt.document_store.base_store import LocalStore @@ -20,36 +24,50 @@ from metagpt.utils.embedding import get_embedding class FaissStore(LocalStore): def __init__( - self, raw_data: Path, cache_dir=None, meta_col="source", content_col="output", embedding: Embeddings = None + self, raw_data: Path, cache_dir=None, meta_col="source", content_col="output", embedding: BaseEmbedding = None ): self.meta_col = meta_col self.content_col = content_col self.embedding = embedding or get_embedding() + self.store: VectorStoreIndex super().__init__(raw_data, cache_dir) - def _load(self) -> Optional["FaissStore"]: - index_file, store_file = self._get_index_and_store_fname(index_ext=".faiss") # langchain FAISS using .faiss + def _load(self) -> Optional["VectorStoreIndex"]: + index_file, store_file = self._get_index_and_store_fname() if not (index_file.exists() and store_file.exists()): logger.info("Missing at least one of index_file/store_file, load failed and return None") return None + vector_store = FaissVectorStore.from_persist_dir(persist_dir=self.cache_dir) + storage_context = StorageContext.from_defaults(persist_dir=self.cache_dir, vector_store=vector_store) + index = load_index_from_storage(storage_context, embed_model=self.embedding) - return FAISS.load_local(self.raw_data_path.parent, self.embedding, self.fname) + return index - def _write(self, docs, metadatas): - store = FAISS.from_texts(docs, self.embedding, metadatas=metadatas) - return store + def _write(self, docs: list[str], metadatas: list[dict[str, Any]]) -> VectorStoreIndex: + assert len(docs) == len(metadatas) + documents = [Document(text=doc, metadata=metadatas[idx]) for idx, doc in enumerate(docs)] + + vector_store = FaissVectorStore(faiss_index=faiss.IndexFlatL2(1536)) + storage_context = StorageContext.from_defaults(vector_store=vector_store) + index = VectorStoreIndex.from_documents( + documents=documents, storage_context=storage_context, embed_model=self.embedding + ) + + return index def persist(self): - self.store.save_local(self.raw_data_path.parent, self.fname) + self.store.storage_context.persist(self.cache_dir) + + def search(self, query: str, expand_cols=False, sep="\n", *args, k=5, **kwargs): + retriever = self.store.as_retriever(similarity_top_k=k) + rsp = retriever.retrieve(QueryBundle(query_str=query, embedding=self.embedding.get_text_embedding(query))) - def search(self, query, expand_cols=False, sep="\n", *args, k=5, **kwargs): - rsp = self.store.similarity_search(query, k=k, **kwargs) logger.debug(rsp) if expand_cols: - return str(sep.join([f"{x.page_content}: {x.metadata}" for x in rsp])) + return str(sep.join([f"{x.node.text}: {x.node.metadata}" for x in rsp])) else: - return str(sep.join([f"{x.page_content}" for x in rsp])) + return str(sep.join([f"{x.node.text}" for x in rsp])) async def asearch(self, *args, **kwargs): return await asyncio.to_thread(self.search, *args, **kwargs) @@ -67,8 +85,12 @@ class FaissStore(LocalStore): def add(self, texts: list[str], *args, **kwargs) -> list[str]: """FIXME: Currently, the store is not updated after adding.""" - return self.store.add_texts(texts) + texts_embeds = self.embedding.get_text_embedding_batch(texts) + nodes = [TextNode(text=texts[idx], embedding=embed) for idx, embed in enumerate(texts_embeds)] + self.store.insert_nodes(nodes) + + return [] def delete(self, *args, **kwargs): - """Currently, langchain does not provide a delete interface.""" + """Currently, faiss does not provide a delete interface.""" raise NotImplementedError diff --git a/metagpt/environment/README.md b/metagpt/environment/README.md index 9476ac75a..bb7d50d50 100644 --- a/metagpt/environment/README.md +++ b/metagpt/environment/README.md @@ -21,7 +21,7 @@ ## Usage from metagpt.environment.api.env_api import EnvAPIAbstract # get screenshot from ExtEnv -screenshot_path: Path = env.observe( +screenshot_path: Path = await env.observe( EnvAPIAbstract( api_name="get_screenshot", kwargs={"ss_name": f"{round_count}_before", "local_save_dir": task_dir} ) @@ -34,5 +34,5 @@ # do a `tap` action on the screen ## TODO - add android app operation assistant under `examples/android_assistant` - migrate roles/actions of werewolf game from old version into current version -- migrate roles/actions of mincraft game from old version into current version +- migrate roles/actions of minecraft game from old version into current version - migrate roles/actions of stanford_town game from old version into current version diff --git a/metagpt/environment/__init__.py b/metagpt/environment/__init__.py index 692672fa7..0621fff16 100644 --- a/metagpt/environment/__init__.py +++ b/metagpt/environment/__init__.py @@ -3,11 +3,10 @@ # @Desc : from metagpt.environment.base_env import Environment -from metagpt.environment.android_env.android_env import AndroidEnv -from metagpt.environment.mincraft_env.mincraft_env import MincraftExtEnv -from metagpt.environment.werewolf_env.werewolf_env import WerewolfEnv -from metagpt.environment.stanford_town_env.stanford_town_env import StanfordTownEnv -from metagpt.environment.software_env.software_env import SoftwareEnv +from metagpt.environment.android.android_env import AndroidEnv +from metagpt.environment.werewolf.werewolf_env import WerewolfEnv +from metagpt.environment.stanford_town.stanford_town_env import StanfordTownEnv +from metagpt.environment.software.software_env import SoftwareEnv -__all__ = ["AndroidEnv", "MincraftExtEnv", "WerewolfEnv", "StanfordTownEnv", "SoftwareEnv", "Environment"] +__all__ = ["AndroidEnv", "WerewolfEnv", "StanfordTownEnv", "SoftwareEnv", "Environment"] diff --git a/metagpt/environment/mincraft_env/__init__.py b/metagpt/environment/android/__init__.py similarity index 100% rename from metagpt/environment/mincraft_env/__init__.py rename to metagpt/environment/android/__init__.py diff --git a/metagpt/environment/android_env/android_env.py b/metagpt/environment/android/android_env.py similarity index 83% rename from metagpt/environment/android_env/android_env.py rename to metagpt/environment/android/android_env.py index c27e20541..b3227df0f 100644 --- a/metagpt/environment/android_env/android_env.py +++ b/metagpt/environment/android/android_env.py @@ -4,7 +4,7 @@ from pydantic import Field -from metagpt.environment.android_env.android_ext_env import AndroidExtEnv +from metagpt.environment.android.android_ext_env import AndroidExtEnv from metagpt.environment.base_env import Environment diff --git a/metagpt/environment/android_env/android_ext_env.py b/metagpt/environment/android/android_ext_env.py similarity index 91% rename from metagpt/environment/android_env/android_ext_env.py rename to metagpt/environment/android/android_ext_env.py index b81b2cd26..d2344fa1f 100644 --- a/metagpt/environment/android_env/android_ext_env.py +++ b/metagpt/environment/android/android_ext_env.py @@ -8,8 +8,9 @@ from typing import Any, Optional from pydantic import Field -from metagpt.environment.android_env.const import ADB_EXEC_FAIL +from metagpt.environment.android.const import ADB_EXEC_FAIL from metagpt.environment.base_env import ExtEnv, mark_as_readable, mark_as_writeable +from metagpt.environment.base_env_space import BaseEnvAction, BaseEnvObsParams class AndroidExtEnv(ExtEnv): @@ -19,6 +20,20 @@ class AndroidExtEnv(ExtEnv): width: int = Field(default=720, description="device screen width") height: int = Field(default=1080, description="device screen height") + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, Any], dict[str, Any]]: + pass + + def observe(self, obs_params: Optional[BaseEnvObsParams] = None) -> Any: + pass + + def step(self, action: BaseEnvAction) -> tuple[dict[str, Any], float, bool, bool, dict[str, Any]]: + pass + def __init__(self, **data: Any): super().__init__(**data) if data.get("device_id"): diff --git a/metagpt/environment/android_env/const.py b/metagpt/environment/android/const.py similarity index 100% rename from metagpt/environment/android_env/const.py rename to metagpt/environment/android/const.py diff --git a/metagpt/environment/base_env.py b/metagpt/environment/base_env.py index 0e583ffb3..024c46877 100644 --- a/metagpt/environment/base_env.py +++ b/metagpt/environment/base_env.py @@ -3,9 +3,12 @@ # @Desc : base env of executing environment import asyncio +from abc import abstractmethod from enum import Enum from typing import TYPE_CHECKING, Any, Dict, Iterable, Optional, Set, Union +from gymnasium import spaces +from gymnasium.core import ActType, ObsType from pydantic import BaseModel, ConfigDict, Field, SerializeAsAny, model_validator from metagpt.context import Context @@ -14,6 +17,7 @@ from metagpt.environment.api.env_api import ( ReadAPIRegistry, WriteAPIRegistry, ) +from metagpt.environment.base_env_space import BaseEnvAction, BaseEnvObsParams from metagpt.logs import logger from metagpt.schema import Message from metagpt.utils.common import get_function_schema, is_coroutine_func, is_send_to @@ -26,7 +30,7 @@ class EnvType(Enum): ANDROID = "Android" GYM = "Gym" WEREWOLF = "Werewolf" - MINCRAFT = "Mincraft" + MINECRAFT = "Minecraft" STANFORDTOWN = "StanfordTown" @@ -47,7 +51,12 @@ def mark_as_writeable(func): class ExtEnv(BaseModel): - """External Env to intergate actual game environment""" + """External Env to integrate actual game environment""" + + model_config = ConfigDict(arbitrary_types_allowed=True) + + action_space: spaces.Space[ActType] = Field(default_factory=spaces.Space, exclude=True) + observation_space: spaces.Space[ObsType] = Field(default_factory=spaces.Space, exclude=True) def _check_api_exist(self, rw_api: Optional[str] = None): if not rw_api: @@ -61,39 +70,56 @@ class ExtEnv(BaseModel): else: return env_write_api_registry.get_apis() - async def observe(self, env_action: Union[str, EnvAPIAbstract]): + async def read_from_api(self, env_action: Union[str, EnvAPIAbstract]): """get observation from particular api of ExtEnv""" if isinstance(env_action, str): - read_api = env_read_api_registry.get(api_name=env_action)["func"] - self._check_api_exist(read_api) - if is_coroutine_func(read_api): - res = await read_api(self) + env_read_api = env_read_api_registry.get(api_name=env_action)["func"] + self._check_api_exist(env_read_api) + if is_coroutine_func(env_read_api): + res = await env_read_api(self) else: - res = read_api(self) + res = env_read_api(self) elif isinstance(env_action, EnvAPIAbstract): - read_api = env_read_api_registry.get(api_name=env_action.api_name)["func"] - self._check_api_exist(read_api) - if is_coroutine_func(read_api): - res = await read_api(self, *env_action.args, **env_action.kwargs) + env_read_api = env_read_api_registry.get(api_name=env_action.api_name)["func"] + self._check_api_exist(env_read_api) + if is_coroutine_func(env_read_api): + res = await env_read_api(self, *env_action.args, **env_action.kwargs) else: - res = read_api(self, *env_action.args, **env_action.kwargs) + res = env_read_api(self, *env_action.args, **env_action.kwargs) return res - async def step(self, env_action: Union[str, Message, EnvAPIAbstract, list[EnvAPIAbstract]]): + async def write_thru_api(self, env_action: Union[str, Message, EnvAPIAbstract, list[EnvAPIAbstract]]): """execute through particular api of ExtEnv""" res = None if isinstance(env_action, Message): self.publish_message(env_action) elif isinstance(env_action, EnvAPIAbstract): - write_api = env_write_api_registry.get(env_action.api_name)["func"] - self._check_api_exist(write_api) - if is_coroutine_func(write_api): - res = await write_api(self, *env_action.args, **env_action.kwargs) + env_write_api = env_write_api_registry.get(env_action.api_name)["func"] + self._check_api_exist(env_write_api) + if is_coroutine_func(env_write_api): + res = await env_write_api(self, *env_action.args, **env_action.kwargs) else: - res = write_api(self, *env_action.args, **env_action.kwargs) + res = env_write_api(self, *env_action.args, **env_action.kwargs) return res + @abstractmethod + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, Any], dict[str, Any]]: + """Implement this to get init observation""" + + @abstractmethod + def observe(self, obs_params: Optional[BaseEnvObsParams] = None) -> Any: + """Implement this if you want to get partial observation from the env""" + + @abstractmethod + def step(self, action: BaseEnvAction) -> tuple[dict[str, Any], float, bool, bool, dict[str, Any]]: + """Implement this to feed a action and then get new observation from the env""" + class Environment(ExtEnv): """环境，承载一批角色，角色可以向环境发布消息，可以被其他角色观察到 @@ -108,6 +134,20 @@ class Environment(ExtEnv): history: str = "" # For debug context: Context = Field(default_factory=Context, exclude=True) + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, Any], dict[str, Any]]: + pass + + def observe(self, obs_params: Optional[BaseEnvObsParams] = None) -> Any: + pass + + def step(self, action: BaseEnvAction) -> tuple[dict[str, Any], float, bool, bool, dict[str, Any]]: + pass + @model_validator(mode="after") def init_roles(self): self.add_roles(self.roles.values()) @@ -129,8 +169,8 @@ class Environment(ExtEnv): self.roles[role.profile] = role for role in roles: # setup system message with roles - role.set_env(self) role.context = self.context + role.set_env(self) def publish_message(self, message: Message, peekable: bool = True) -> bool: """ diff --git a/metagpt/environment/base_env_space.py b/metagpt/environment/base_env_space.py new file mode 100644 index 000000000..fd0cfa399 --- /dev/null +++ b/metagpt/environment/base_env_space.py @@ -0,0 +1,33 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : + +from enum import IntEnum + +from pydantic import BaseModel, ConfigDict, Field + + +class BaseEnvActionType(IntEnum): + # # NONE = 0 # no action to run, just get observation + pass + + +class BaseEnvAction(BaseModel): + """env action type and its related params of action functions/apis""" + + model_config = ConfigDict(arbitrary_types_allowed=True) + + action_type: int = Field(default=0, description="action type") + + +class BaseEnvObsType(IntEnum): + # # NONE = 0 # get whole observation from env + pass + + +class BaseEnvObsParams(BaseModel): + """observation params for different EnvObsType to get its observe result""" + + model_config = ConfigDict(arbitrary_types_allowed=True) + + obs_type: int = Field(default=0, description="observation type") diff --git a/metagpt/environment/software_env/__init__.py b/metagpt/environment/minecraft/__init__.py similarity index 100% rename from metagpt/environment/software_env/__init__.py rename to metagpt/environment/minecraft/__init__.py diff --git a/metagpt/environment/mincraft_env/const.py b/metagpt/environment/minecraft/const.py similarity index 92% rename from metagpt/environment/mincraft_env/const.py rename to metagpt/environment/minecraft/const.py index a7222f9cd..8ac15decc 100644 --- a/metagpt/environment/mincraft_env/const.py +++ b/metagpt/environment/minecraft/const.py @@ -4,8 +4,8 @@ from metagpt.const import METAGPT_ROOT -# For Mincraft Game Agent -MC_CKPT_DIR = METAGPT_ROOT / "data/mincraft/ckpt" +# For Minecraft Game Agent +MC_CKPT_DIR = METAGPT_ROOT / "data/minecraft/ckpt" MC_LOG_DIR = METAGPT_ROOT / "logs" MC_DEFAULT_WARMUP = { "context": 15, diff --git a/metagpt/environment/mincraft_env/mincraft_env.py b/metagpt/environment/minecraft/minecraft_env.py similarity index 93% rename from metagpt/environment/mincraft_env/mincraft_env.py rename to metagpt/environment/minecraft/minecraft_env.py index 6327aa3f4..0f39c9ccd 100644 --- a/metagpt/environment/mincraft_env/mincraft_env.py +++ b/metagpt/environment/minecraft/minecraft_env.py @@ -1,6 +1,6 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- -# @Desc : MG Mincraft Env +# @Desc : MG Minecraft Env # refs to `voyager voyager.py` import json @@ -8,20 +8,19 @@ import re import time from typing import Any, Iterable -from langchain.embeddings.openai import OpenAIEmbeddings -from langchain.vectorstores import Chroma +from llama_index.vector_stores.chroma import ChromaVectorStore from pydantic import ConfigDict, Field from metagpt.config2 import config as CONFIG from metagpt.environment.base_env import Environment -from metagpt.environment.mincraft_env.const import MC_CKPT_DIR -from metagpt.environment.mincraft_env.mincraft_ext_env import MincraftExtEnv +from metagpt.environment.minecraft.const import MC_CKPT_DIR +from metagpt.environment.minecraft.minecraft_ext_env import MinecraftExtEnv from metagpt.logs import logger from metagpt.utils.common import load_mc_skills_code, read_json_file, write_json_file -class MincraftEnv(Environment, MincraftExtEnv): - """MincraftEnv, including shared memory of cache and infomation between roles""" +class MinecraftEnv(Environment, MinecraftExtEnv): + """MinecraftEnv, including shared memory of cache and information between roles""" model_config = ConfigDict(arbitrary_types_allowed=True) @@ -48,9 +47,9 @@ class MincraftEnv(Environment, MincraftExtEnv): runtime_status: bool = False # equal to action execution status: success or failed - vectordb: Chroma = Field(default_factory=Chroma) + vectordb: ChromaVectorStore = Field(default_factory=ChromaVectorStore) - qa_cache_questions_vectordb: Chroma = Field(default_factory=Chroma) + qa_cache_questions_vectordb: ChromaVectorStore = Field(default_factory=ChromaVectorStore) @property def progress(self): @@ -73,16 +72,14 @@ class MincraftEnv(Environment, MincraftExtEnv): self.set_mc_resume() def set_mc_resume(self): - self.qa_cache_questions_vectordb = Chroma( + self.qa_cache_questions_vectordb = ChromaVectorStore( collection_name="qa_cache_questions_vectordb", - embedding_function=OpenAIEmbeddings(), - persist_directory=f"{MC_CKPT_DIR}/curriculum/vectordb", + persist_dir=f"{MC_CKPT_DIR}/curriculum/vectordb", ) - self.vectordb = Chroma( + self.vectordb = ChromaVectorStore( collection_name="skill_vectordb", - embedding_function=OpenAIEmbeddings(), - persist_directory=f"{MC_CKPT_DIR}/skill/vectordb", + persist_dir=f"{MC_CKPT_DIR}/skill/vectordb", ) if CONFIG.resume: @@ -285,7 +282,7 @@ class MincraftEnv(Environment, MincraftExtEnv): position = event["status"]["position"] blocks.append(block) positions.append(position) - new_events = self.step( + new_events = self._step( f"await givePlacedItemBack(bot, {json.dumps(blocks)}, {json.dumps(positions)})", programs=self.programs, ) @@ -326,7 +323,7 @@ class MincraftEnv(Environment, MincraftExtEnv): Exception: If there is an issue retrieving events. """ try: - self.reset( + self._reset( options={ "mode": "soft", "wait_ticks": 20, @@ -335,13 +332,13 @@ class MincraftEnv(Environment, MincraftExtEnv): # difficulty = "easy" if len(self.completed_tasks) > 15 else "peaceful" difficulty = "peaceful" - events = self.step("bot.chat(`/time set ${getNextTime()}`);\n" + f"bot.chat('/difficulty {difficulty}');") + events = self._step("bot.chat(`/time set ${getNextTime()}`);\n" + f"bot.chat('/difficulty {difficulty}');") self.update_event(events) return events except Exception as e: time.sleep(3) # wait for mineflayer to exit # reset bot status here - events = self.reset( + events = self._reset( options={ "mode": "hard", "wait_ticks": 20, @@ -368,7 +365,7 @@ class MincraftEnv(Environment, MincraftExtEnv): Exception: If there is an issue retrieving events. """ try: - events = self.step( + events = self._step( code=self.code, programs=self.programs, ) @@ -377,7 +374,7 @@ class MincraftEnv(Environment, MincraftExtEnv): except Exception as e: time.sleep(3) # wait for mineflayer to exit # reset bot status here - events = self.reset( + events = self._reset( options={ "mode": "hard", "wait_ticks": 20, diff --git a/metagpt/environment/mincraft_env/mincraft_ext_env.py b/metagpt/environment/minecraft/minecraft_ext_env.py similarity index 87% rename from metagpt/environment/mincraft_env/mincraft_ext_env.py rename to metagpt/environment/minecraft/minecraft_ext_env.py index b86250d8c..0436bc3aa 100644 --- a/metagpt/environment/mincraft_env/mincraft_ext_env.py +++ b/metagpt/environment/minecraft/minecraft_ext_env.py @@ -1,28 +1,29 @@ #!/usr/bin/env python # -*- coding: utf-8 -*- -# @Desc : The Mincraft external environment to integrate with Mincraft game +# @Desc : The Minecraft external environment to integrate with Minecraft game # refs to `voyager bridge.py` import json import time -from typing import Optional +from typing import Any, Optional import requests from pydantic import ConfigDict, Field, model_validator from metagpt.environment.base_env import ExtEnv, mark_as_writeable -from metagpt.environment.mincraft_env.const import ( +from metagpt.environment.base_env_space import BaseEnvAction, BaseEnvObsParams +from metagpt.environment.minecraft.const import ( MC_CKPT_DIR, MC_CORE_INVENTORY_ITEMS, MC_CURRICULUM_OB, MC_DEFAULT_WARMUP, METAGPT_ROOT, ) -from metagpt.environment.mincraft_env.process_monitor import SubprocessMonitor +from metagpt.environment.minecraft.process_monitor import SubprocessMonitor from metagpt.logs import logger -class MincraftExtEnv(ExtEnv): +class MinecraftExtEnv(ExtEnv): model_config = ConfigDict(arbitrary_types_allowed=True) mc_port: Optional[int] = Field(default=None) @@ -38,6 +39,20 @@ class MincraftExtEnv(ExtEnv): server_paused: bool = Field(default=False) warm_up: dict = Field(default=dict()) + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, Any], dict[str, Any]]: + pass + + def observe(self, obs_params: Optional[BaseEnvObsParams] = None) -> Any: + pass + + def step(self, action: BaseEnvAction) -> tuple[dict[str, Any], float, bool, bool, dict[str, Any]]: + pass + @property def server(self) -> str: return f"{self.server_host}:{self.server_port}" @@ -48,7 +63,7 @@ class MincraftExtEnv(ExtEnv): self.mineflayer = SubprocessMonitor( commands=[ "node", - METAGPT_ROOT.joinpath("metagpt", "environment", "mincraft_env", "mineflayer", "index.js"), + METAGPT_ROOT.joinpath("metagpt", "environment", "minecraft", "mineflayer", "index.js"), str(self.server_port), ], name="mineflayer", @@ -115,7 +130,7 @@ class MincraftExtEnv(ExtEnv): return res.json() @mark_as_writeable - def reset(self, *, seed=None, options=None) -> dict: + def _reset(self, *, seed=None, options=None) -> dict: if options is None: options = {} if options.get("inventory", {}) and options.get("mode", "hard") != "hard": @@ -145,7 +160,7 @@ class MincraftExtEnv(ExtEnv): return json.loads(returned_data) @mark_as_writeable - def step(self, code: str, programs: str = "") -> dict: + def _step(self, code: str, programs: str = "") -> dict: if not self.has_reset: raise RuntimeError("Environment has not been reset yet") self.check_process() diff --git a/metagpt/environment/mincraft_env/mineflayer/.gitignore b/metagpt/environment/minecraft/mineflayer/.gitignore similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/.gitignore rename to metagpt/environment/minecraft/mineflayer/.gitignore diff --git a/metagpt/environment/mincraft_env/mineflayer/.prettierignore b/metagpt/environment/minecraft/mineflayer/.prettierignore similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/.prettierignore rename to metagpt/environment/minecraft/mineflayer/.prettierignore diff --git a/metagpt/environment/mincraft_env/mineflayer/.prettierrc.json b/metagpt/environment/minecraft/mineflayer/.prettierrc.json similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/.prettierrc.json rename to metagpt/environment/minecraft/mineflayer/.prettierrc.json diff --git a/metagpt/environment/mincraft_env/mineflayer/index.js b/metagpt/environment/minecraft/mineflayer/index.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/index.js rename to metagpt/environment/minecraft/mineflayer/index.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/base.js b/metagpt/environment/minecraft/mineflayer/lib/observation/base.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/base.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/base.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/chests.js b/metagpt/environment/minecraft/mineflayer/lib/observation/chests.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/chests.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/chests.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/inventory.js b/metagpt/environment/minecraft/mineflayer/lib/observation/inventory.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/inventory.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/inventory.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/onChat.js b/metagpt/environment/minecraft/mineflayer/lib/observation/onChat.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/onChat.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/onChat.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/onError.js b/metagpt/environment/minecraft/mineflayer/lib/observation/onError.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/onError.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/onError.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/onSave.js b/metagpt/environment/minecraft/mineflayer/lib/observation/onSave.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/onSave.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/onSave.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/status.js b/metagpt/environment/minecraft/mineflayer/lib/observation/status.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/status.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/status.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/observation/voxels.js b/metagpt/environment/minecraft/mineflayer/lib/observation/voxels.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/observation/voxels.js rename to metagpt/environment/minecraft/mineflayer/lib/observation/voxels.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/skillLoader.js b/metagpt/environment/minecraft/mineflayer/lib/skillLoader.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/skillLoader.js rename to metagpt/environment/minecraft/mineflayer/lib/skillLoader.js diff --git a/metagpt/environment/mincraft_env/mineflayer/lib/utils.js b/metagpt/environment/minecraft/mineflayer/lib/utils.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/lib/utils.js rename to metagpt/environment/minecraft/mineflayer/lib/utils.js diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/.gitignore b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/.gitignore similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/.gitignore rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/.gitignore diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/LICENSE b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/LICENSE similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/LICENSE rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/LICENSE diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/README.md b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/README.md similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/README.md rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/README.md diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/_config.yml b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/_config.yml similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/_config.yml rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/_config.yml diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/docs/api.md b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/docs/api.md similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/docs/api.md rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/docs/api.md diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/collector.js b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/collector.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/collector.js rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/collector.js diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/oreMiner.js b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/oreMiner.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/oreMiner.js rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/oreMiner.js diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/storageBot.js b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/storageBot.js similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/examples/storageBot.js rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/examples/storageBot.js diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/package.json b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/package.json similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/package.json rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/package.json diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/BlockVeins.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/BlockVeins.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/BlockVeins.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/BlockVeins.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/CollectBlock.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/CollectBlock.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/CollectBlock.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/CollectBlock.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Inventory.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Inventory.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Inventory.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Inventory.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Targets.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Targets.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Targets.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Targets.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/TaskQueue.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/TaskQueue.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/TaskQueue.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/TaskQueue.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/TemporarySubscriber.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/TemporarySubscriber.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/TemporarySubscriber.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/TemporarySubscriber.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Util.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Util.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/Util.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/Util.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/index.ts b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/index.ts similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/src/index.ts rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/src/index.ts diff --git a/metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/tsconfig.json b/metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/tsconfig.json similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/mineflayer-collectblock/tsconfig.json rename to metagpt/environment/minecraft/mineflayer/mineflayer-collectblock/tsconfig.json diff --git a/metagpt/environment/mincraft_env/mineflayer/package.json b/metagpt/environment/minecraft/mineflayer/package.json similarity index 100% rename from metagpt/environment/mincraft_env/mineflayer/package.json rename to metagpt/environment/minecraft/mineflayer/package.json diff --git a/metagpt/environment/mincraft_env/process_monitor.py b/metagpt/environment/minecraft/process_monitor.py similarity index 100% rename from metagpt/environment/mincraft_env/process_monitor.py rename to metagpt/environment/minecraft/process_monitor.py diff --git a/metagpt/environment/stanford_town_env/__init__.py b/metagpt/environment/software/__init__.py similarity index 100% rename from metagpt/environment/stanford_town_env/__init__.py rename to metagpt/environment/software/__init__.py diff --git a/metagpt/environment/software_env/software_env.py b/metagpt/environment/software/software_env.py similarity index 100% rename from metagpt/environment/software_env/software_env.py rename to metagpt/environment/software/software_env.py diff --git a/metagpt/environment/werewolf_env/__init__.py b/metagpt/environment/stanford_town/__init__.py similarity index 100% rename from metagpt/environment/werewolf_env/__init__.py rename to metagpt/environment/stanford_town/__init__.py diff --git a/metagpt/environment/stanford_town/env_space.py b/metagpt/environment/stanford_town/env_space.py new file mode 100644 index 000000000..e100a2952 --- /dev/null +++ b/metagpt/environment/stanford_town/env_space.py @@ -0,0 +1,105 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : + +from typing import Any, Optional, Union + +import numpy as np +import numpy.typing as npt +from gymnasium import spaces +from pydantic import ConfigDict, Field, field_validator + +from metagpt.environment.base_env_space import ( + BaseEnvAction, + BaseEnvActionType, + BaseEnvObsParams, + BaseEnvObsType, +) + + +class EnvActionType(BaseEnvActionType): + NONE = 0 # no action to run, just get observation + + ADD_TILE_EVENT = 1 # Add an event triple to a tile + RM_TILE_EVENT = 2 # Remove an event triple from a tile + TURN_TILE_EVENT_IDLE = 3 # Turn an event triple from a tile into idle + RM_TITLE_SUB_EVENT = 4 # Remove an event triple that has the input subject from a tile + + +class EnvAction(BaseEnvAction): + """env action type and its related params of action functions/apis""" + + model_config = ConfigDict(arbitrary_types_allowed=True) + + action_type: int = Field(default=EnvActionType.NONE, description="action type") + coord: npt.NDArray[np.int64] = Field( + default_factory=lambda: np.zeros(2, dtype=np.int64), description="tile coordinate" + ) + subject: str = Field(default="", description="subject name of first element in event") + event: tuple[str, Optional[str], Optional[str], Optional[str]] = Field( + default=["", None, None, None], description="tile event" + ) + + @field_validator("coord", mode="before") + @classmethod + def check_coord(cls, coord) -> npt.NDArray[np.int64]: + if not isinstance(coord, np.ndarray): + return np.array(coord) + + +class EnvObsType(BaseEnvObsType): + """get part observation with specific params""" + + NONE = 0 # get whole observation from env + + GET_TITLE = 1 # get the tile detail dictionary with given tile coord + TILE_PATH = 2 # get the tile address with given tile coord + TILE_NBR = 3 # get the neighbors of given tile coord and its vision radius + + +class EnvObsParams(BaseEnvObsParams): + """observation params for different EnvObsType""" + + model_config = ConfigDict(arbitrary_types_allowed=True) + + obs_type: int = Field(default=EnvObsType.NONE, description="observation type") + coord: npt.NDArray[np.int64] = Field( + default_factory=lambda: np.zeros(2, dtype=np.int64), description="tile coordinate" + ) + level: str = Field(default="", description="different level of title") + vision_radius: int = Field(default=0, description="the vision radius of current tile") + + @field_validator("coord", mode="before") + @classmethod + def check_coord(cls, coord) -> npt.NDArray[np.int64]: + if not isinstance(coord, np.ndarray): + return np.array(coord) + + +EnvObsValType = Union[list[list[str]], dict[str, set[tuple[int, int]]], list[list[dict[str, Any]]]] + + +def get_observation_space() -> spaces.Dict: + # it's a + space = spaces.Dict( + {"collision_maze": spaces.Discrete(2), "tiles": spaces.Discrete(2), "address_tiles": spaces.Discrete(2)} + ) + + return space + + +def get_action_space(maze_shape: tuple[int, int]) -> spaces.Dict: + """The fields defined by the space correspond to the input parameters of the action except `action_type`""" + space = spaces.Dict( + { + "action_type": spaces.Discrete(len(EnvActionType)), + "coord": spaces.Box( + np.array([0, 0], dtype=np.int64), np.array([maze_shape[0], maze_shape[1]], dtype=np.int64) + ), # coord of the tile + "subject": spaces.Text(256), # the first element of an tile event + "event": spaces.Tuple( + (spaces.Text(256), spaces.Text(256), spaces.Text(256), spaces.Text(256)) + ), # event is a tuple of four str + } + ) + return space diff --git a/metagpt/environment/stanford_town/stanford_town_env.py b/metagpt/environment/stanford_town/stanford_town_env.py new file mode 100644 index 000000000..af8a882b2 --- /dev/null +++ b/metagpt/environment/stanford_town/stanford_town_env.py @@ -0,0 +1,10 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : MG StanfordTown Env + +from metagpt.environment.base_env import Environment +from metagpt.environment.stanford_town.stanford_town_ext_env import StanfordTownExtEnv + + +class StanfordTownEnv(StanfordTownExtEnv, Environment): + pass diff --git a/metagpt/environment/stanford_town_env/stanford_town_ext_env.py b/metagpt/environment/stanford_town/stanford_town_ext_env.py similarity index 83% rename from metagpt/environment/stanford_town_env/stanford_town_ext_env.py rename to metagpt/environment/stanford_town/stanford_town_ext_env.py index 8a9a65965..30a02d4db 100644 --- a/metagpt/environment/stanford_town_env/stanford_town_ext_env.py +++ b/metagpt/environment/stanford_town/stanford_town_ext_env.py @@ -5,11 +5,20 @@ import math from pathlib import Path -from typing import Optional, Tuple +from typing import Any, Optional from pydantic import ConfigDict, Field, model_validator from metagpt.environment.base_env import ExtEnv, mark_as_readable, mark_as_writeable +from metagpt.environment.stanford_town.env_space import ( + EnvAction, + EnvActionType, + EnvObsParams, + EnvObsType, + EnvObsValType, + get_action_space, + get_observation_space, +) from metagpt.utils.common import read_csv_to_list, read_json_file @@ -197,15 +206,82 @@ class StanfordTownExtEnv(ExtEnv): else: address_tiles[add] = set([(j, i)]) values["address_tiles"] = address_tiles + + values["action_space"] = get_action_space((maze_width, maze_height)) + values["observation_space"] = get_observation_space() return values + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, EnvObsValType], dict[str, Any]]: + """reset env and get the init observation + Return results corresponding to `observation, info` + """ + super().reset(seed=seed, options=options) + + obs = self._get_obs() + + return obs, {} + + def _get_obs(self) -> dict[str, EnvObsValType]: + """Get observation""" + return { + "collision_maze": self.get_collision_maze(), + "tiles": self.tiles, + "address_tiles": self.get_address_tiles(), + } + + def observe(self, obs_params: Optional[EnvObsParams] = None) -> Any: + """Get partial or full observation from the env""" + obs_type = obs_params.obs_type if obs_params else EnvObsType.NONE + if obs_type == EnvObsType.NONE: + obs = self._get_obs() + elif obs_type == EnvObsType.GET_TITLE: + obs = self.access_tile(tile=obs_params.coord) + elif obs_type == EnvObsType.TILE_PATH: + obs = self.get_tile_path(tile=obs_params.coord, level=obs_params.level) + elif obs_type == EnvObsType.TILE_NBR: + obs = self.get_nearby_tiles(tile=obs_params.coord, vision_r=obs_params.vision_radius) + return obs + + def step(self, action: EnvAction) -> tuple[dict[str, EnvObsValType], float, bool, bool, dict[str, Any]]: + """Execute action and then return observation + Return results corresponding to `observation, reward, terminated, truncated, info` + """ + terminated = False + try: + self._execute_env_action(action) + except Exception: + terminated = True + + obs = self._get_obs() + + ret = (obs, 1.0, terminated, False, {}) + return ret + + def _execute_env_action(self, action: EnvAction): + action_type = action.action_type + if action_type == EnvActionType.NONE: + pass + elif action_type == EnvActionType.ADD_TILE_EVENT: + self.add_event_from_tile(curr_event=action.event, tile=action.coord) + elif action_type == EnvActionType.RM_TILE_EVENT: + self.remove_event_from_tile(curr_event=action.event, tile=action.coord) + elif action_type == EnvActionType.TURN_TILE_EVENT_IDLE: + self.turn_event_from_tile_idle(curr_event=action.event, tile=action.coord) + elif action_type == EnvActionType.RM_TITLE_SUB_EVENT: + self.remove_subject_events_from_tile(subject=action.subject, tile=action.coord) + def turn_coordinate_to_tile(self, px_coordinate: tuple[int, int]) -> tuple[int, int]: """ Turns a pixel coordinate to a tile coordinate. """ x = math.ceil(px_coordinate[0] / self.sq_tile_size) y = math.ceil(px_coordinate[1] / self.sq_tile_size) - return (x, y) + return x, y @mark_as_readable def get_collision_maze(self) -> list: @@ -316,10 +392,6 @@ class StanfordTownExtEnv(ExtEnv): nearby_tiles += [(i, j)] return nearby_tiles - @mark_as_writeable - def add_tiles_event(self, pt_y: int, pt_x: int, event: Tuple[str, str, str, str]): - self.tiles[pt_y][pt_x]["events"].add(event) - @mark_as_writeable def add_event_from_tile(self, curr_event: tuple[str], tile: tuple[int, int]) -> None: """ diff --git a/metagpt/environment/stanford_town_env/stanford_town_env.py b/metagpt/environment/stanford_town_env/stanford_town_env.py deleted file mode 100644 index 8721d6cd1..000000000 --- a/metagpt/environment/stanford_town_env/stanford_town_env.py +++ /dev/null @@ -1,12 +0,0 @@ -#!/usr/bin/env python -# -*- coding: utf-8 -*- -# @Desc : MG StanfordTown Env - -from metagpt.environment.base_env import Environment -from metagpt.environment.stanford_town_env.stanford_town_ext_env import ( - StanfordTownExtEnv, -) - - -class StanfordTownEnv(Environment, StanfordTownExtEnv): - pass diff --git a/tests/metagpt/environment/mincraft_env/__init__.py b/metagpt/environment/werewolf/__init__.py similarity index 100% rename from tests/metagpt/environment/mincraft_env/__init__.py rename to metagpt/environment/werewolf/__init__.py diff --git a/metagpt/environment/werewolf_env/werewolf_env.py b/metagpt/environment/werewolf/werewolf_env.py similarity index 93% rename from metagpt/environment/werewolf_env/werewolf_env.py rename to metagpt/environment/werewolf/werewolf_env.py index d174f322c..e22d9c46f 100644 --- a/metagpt/environment/werewolf_env/werewolf_env.py +++ b/metagpt/environment/werewolf/werewolf_env.py @@ -5,7 +5,7 @@ from pydantic import Field from metagpt.environment.base_env import Environment -from metagpt.environment.werewolf_env.werewolf_ext_env import WerewolfExtEnv +from metagpt.environment.werewolf.werewolf_ext_env import WerewolfExtEnv from metagpt.logs import logger from metagpt.schema import Message diff --git a/metagpt/environment/werewolf_env/werewolf_ext_env.py b/metagpt/environment/werewolf/werewolf_ext_env.py similarity index 96% rename from metagpt/environment/werewolf_env/werewolf_ext_env.py rename to metagpt/environment/werewolf/werewolf_ext_env.py index 7c4b4c475..3f2508b06 100644 --- a/metagpt/environment/werewolf_env/werewolf_ext_env.py +++ b/metagpt/environment/werewolf/werewolf_ext_env.py @@ -5,11 +5,12 @@ import random from collections import Counter from enum import Enum -from typing import Callable, Optional +from typing import Any, Callable, Optional from pydantic import ConfigDict, Field from metagpt.environment.base_env import ExtEnv, mark_as_readable, mark_as_writeable +from metagpt.environment.base_env_space import BaseEnvAction, BaseEnvObsParams from metagpt.logs import logger @@ -128,6 +129,20 @@ class WerewolfExtEnv(ExtEnv): player_poisoned: Optional[str] = Field(default=None) player_current_dead: list[str] = Field(default=[]) + def reset( + self, + *, + seed: Optional[int] = None, + options: Optional[dict[str, Any]] = None, + ) -> tuple[dict[str, Any], dict[str, Any]]: + pass + + def observe(self, obs_params: Optional[BaseEnvObsParams] = None) -> Any: + pass + + def step(self, action: BaseEnvAction) -> tuple[dict[str, Any], float, bool, bool, dict[str, Any]]: + pass + @property def living_players(self) -> list[str]: player_names = [] diff --git a/metagpt/ext/__init__.py b/metagpt/ext/__init__.py new file mode 100644 index 000000000..2bcf8efd0 --- /dev/null +++ b/metagpt/ext/__init__.py @@ -0,0 +1,3 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : diff --git a/metagpt/ext/stanford_town/README.md b/metagpt/ext/stanford_town/README.md new file mode 100644 index 000000000..bfcb1c818 --- /dev/null +++ b/metagpt/ext/stanford_town/README.md @@ -0,0 +1,36 @@ +## Stanford Town Game + +### Pre-Description +In order to facilitate GA( [generative_agents](https://github.com/joonspk-research/generative_agents) )'s frontend docking data (to avoid changing its code), you can set the value `temp_storage_path` to `temp_storage` of `generative_agents` when start `run_st_game.py`. like + +`python3 run_st_game.py --temp_storage_path path/to/ga/temp_storage xxx` + +Or change the path under `const.py` like beflow + +``` +STORAGE_PATH = EXAMPLE_PATH.joinpath("storage") +TEMP_STORAGE_PATH = EXAMPLE_PATH.joinpath("temp_storage") +# updated +STORAGE_PATH = Path("{path/to/ga/storage}") +TEMP_STORAGE_PATH = Path("{path/to/ga/temp_storage}") +``` + +This can be used to achieve docking of simulation data without changing the GA code. Otherwise, the GA code must be modified to adapt to the MG output path. + +If you don't want to start from 0, copy other simulation directories under `generative_agents/environment/frontend_server/storage/` to `examples/stanford_town/storage`, and select a directory named `fork_sim_code`. + +### Backend service startup +The execution entry is `python3 run_st_game.py "Host a open lunch party at 13:00 pm" "base_the_ville_isabella_maria_klaus" "test_sim" 10` +or +`python3 run_st_game.py "Host a open lunch party at 13:00 pm" "base_the_ville_isabella_maria_klaus" "test_sim" 10 --temp_storage_path path/to/ga/temp_storage` + +`idea` is the user's voice to the first Agent, and it is disseminated through this voice to see whether the final multi-agents achieve the goal of hosting or participating in the event. + +### Frontend service startup +Enter project folder `generative_agents` + +Enter `environment/frontend_server` and use `python3 manage.py runserver` to start the front-end service. +Visit `http://localhost:8000/simulator_home` to enter the current simulation interface. + +## Appreciation +The reproduction work has referred the `https://github.com/joonspk-research/generative_agents`, let's make a general statement here. diff --git a/metagpt/ext/stanford_town/README_CN.md b/metagpt/ext/stanford_town/README_CN.md new file mode 100644 index 000000000..c55a742fc --- /dev/null +++ b/metagpt/ext/stanford_town/README_CN.md @@ -0,0 +1,35 @@ +## Stanford Town Game + +### 前置 +为了方便GA（ [generative_agents](https://github.com/joonspk-research/generative_agents) ）的前端对接数据（避免改动它那块的代码），可在启动`run_st_game.py`加上`temp_storage_path`指向`generative_agents`对应的`temp_storage`路径。比如 + +`python3 run_st_game.py --temp_storage_path path/to/ga/temp_storage xxx` + +或将`const.py`下的 + +``` +STORAGE_PATH = EXAMPLE_PATH.joinpath("storage") +TEMP_STORAGE_PATH = EXAMPLE_PATH.joinpath("temp_storage") +# 更新为 +STORAGE_PATH = Path("{path/to/ga/storage}") +TEMP_STORAGE_PATH = Path("{path/to/ga/temp_storage}") +``` +这样可用实现不改变GA代码情况下，实现仿真数据的对接。不然得修改GA的代码来适配MG的输出路径。 + +如果你不想从0开始启动，拷贝`generative_agents/environment/frontend_server/storage/`下的其他仿真目录到`examples/stanford_town/storage`，并选择一个目录名作为`fork_sim_code`。 + +### 后端服务启动 +执行入口为：`python3 run_st_game.py "Host a open lunch party at 13:00 pm" "base_the_ville_isabella_maria_klaus" "test_sim" 10` +或者 +`python3 run_st_game.py "Host a open lunch party at 13:00 pm" "base_the_ville_isabella_maria_klaus" "test_sim" 10 --temp_storage_path path/to/ga/temp_storage` + +`idea`为用户给第一个Agent的用户心声，并通过这个心声进行传播，看最后多智能体是否达到举办、参加活动的目标。 + +### 前端服务启动 +进入`generative_agents`项目目录 + +进入`environment/frontend_server`，使用`python3 manage.py runserver`启动前端服务。 +访问`http://localhost:8000/simulator_home` 进入当前的仿真界面。 + +## Appreciation +The reproduction work has referred the `https://github.com/joonspk-research/generative_agents`, let's make a general statement here. diff --git a/metagpt/ext/stanford_town/__init__.py b/metagpt/ext/stanford_town/__init__.py new file mode 100644 index 000000000..56ea35c9f --- /dev/null +++ b/metagpt/ext/stanford_town/__init__.py @@ -0,0 +1,3 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : stanford town implement diff --git a/metagpt/ext/stanford_town/actions/__init__.py b/metagpt/ext/stanford_town/actions/__init__.py new file mode 100644 index 000000000..2bcf8efd0 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/__init__.py @@ -0,0 +1,3 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : diff --git a/metagpt/ext/stanford_town/actions/agent_chat_sum_rel.py b/metagpt/ext/stanford_town/actions/agent_chat_sum_rel.py new file mode 100644 index 000000000..98d370bb0 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/agent_chat_sum_rel.py @@ -0,0 +1,39 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : summarize relationship in a agent chat + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class AgentChatSumRel(STAction): + name: str = "AgentChatSumRel" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + resp = False + try: + _ = llm_resp.split('"')[0].strip() + resp = True + except Exception: + pass + return resp + + def _func_cleanup(self, llm_resp: str, prompt: str) -> str: + return llm_resp.split('"')[0].strip() + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, init_role: "STRole", target_role: "STRole", statements: str) -> str: + def create_prompt_input(init_role: "STRole", target_role: "STRole", statements: str) -> str: + prompt_input = [statements, init_role.name, target_role.name] + return prompt_input + + prompt_input = create_prompt_input(init_role, target_role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "summarize_chat_relationship_v2.txt") + + example_output = "Jane Doe is working on a project" + special_instruction = "The output should be a string that responds to the question." + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {init_role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/decide_to_talk.py b/metagpt/ext/stanford_town/actions/decide_to_talk.py new file mode 100644 index 000000000..a393f31af --- /dev/null +++ b/metagpt/ext/stanford_town/actions/decide_to_talk.py @@ -0,0 +1,97 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : device to talk to another role, return yes or no + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class DecideToTalk(STAction): + name: str = "DecideToTalk" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + resp = False + try: + if llm_resp.split("Answer in yes or no:")[-1].strip().lower() in ["yes", "no"]: + resp = True + except ValueError: + pass + return resp + + def _func_cleanup(self, llm_resp: str, prompt: str) -> str: + return llm_resp.split("Answer in yes or no:")[-1].strip().lower() + + def _func_fail_default_resp(self) -> str: + return "yes" + + async def run(self, init_role: "STRole", target_role: "STRole", retrieved: dict, *args, **kwargs) -> bool: + """Run action""" + + def create_prompt_input(init_role: "STRole", target_role: "STRole", retrieved: dict) -> str: + scratch = init_role.rc.scratch + target_scratch = target_role.rc.scratch + last_chat = init_role.rc.memory.get_last_chat(target_role.name) + last_chatted_time = "" + last_chat_about = "" + if last_chat: + last_chatted_time = last_chat.created.strftime("%B %d, %Y, %H:%M:%S") + last_chat_about = last_chat.description + + context = "" + for c_node in retrieved["events"]: + curr_desc = c_node.description.split(" ") + curr_desc[2:3] = ["was"] + curr_desc = " ".join(curr_desc) + context += f"{curr_desc}. " + context += "\n" + for c_node in retrieved["thoughts"]: + context += f"{c_node.description}. " + + curr_time = scratch.curr_time.strftime("%B %d, %Y, %H:%M:%S %p") + init_act_desc = scratch.act_description + if "(" in init_act_desc: + init_act_desc = init_act_desc.split("(")[-1][:-1] + + if len(scratch.planned_path) == 0 and "waiting" not in init_act_desc: + init_p_desc = f"{init_role.name} is already {init_act_desc}" + elif "waiting" in init_act_desc: + init_p_desc = f"{init_role.name} is {init_act_desc}" + else: + init_p_desc = f"{init_role.name} is on the way to {init_act_desc}" + + target_act_desc = scratch.act_description + if "(" in target_act_desc: + target_act_desc = target_act_desc.split("(")[-1][:-1] + + if len(target_scratch.planned_path) == 0 and "waiting" not in init_act_desc: + target_p_desc = f"{target_role.name} is already {target_act_desc}" + elif "waiting" in init_act_desc: + target_p_desc = f"{init_role.name} is {init_act_desc}" + else: + target_p_desc = f"{target_role.name} is on the way to {target_act_desc}" + + prompt_input = [] + prompt_input += [context] + + prompt_input += [curr_time] + + prompt_input += [init_role.name] + prompt_input += [target_role.name] + prompt_input += [last_chatted_time] + prompt_input += [last_chat_about] + + prompt_input += [init_p_desc] + prompt_input += [target_p_desc] + prompt_input += [init_role.name] + prompt_input += [target_role.name] + return prompt_input + + prompt_input = create_prompt_input(init_role, target_role, retrieved) + prompt = self.generate_prompt_with_tmpl_filename( + prompt_input=prompt_input, tmpl_filename="decide_to_talk_v2.txt" + ) + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=20) # yes or no + result = True if output == "yes" else False + logger.info(f"Role: {init_role.name} Action: {self.cls_name} output: {result}") + return result diff --git a/metagpt/ext/stanford_town/actions/dummy_action.py b/metagpt/ext/stanford_town/actions/dummy_action.py new file mode 100644 index 000000000..a5004d5ef --- /dev/null +++ b/metagpt/ext/stanford_town/actions/dummy_action.py @@ -0,0 +1,20 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : dummy action to make every STRole can deal DummyMessage which is caused by DummyAction + +from metagpt.actions import Action +from metagpt.schema import Message + + +class DummyAction(Action): + async def run(self, *args, **kwargs): + raise NotImplementedError + + +class DummyMessage(Message): + """ + dummy message to pass to role and make them to have a execution every round + """ + + content: str = "dummy" + cause_by: str = "DummyAction" diff --git a/metagpt/ext/stanford_town/actions/gen_action_details.py b/metagpt/ext/stanford_town/actions/gen_action_details.py new file mode 100644 index 000000000..8e268a723 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/gen_action_details.py @@ -0,0 +1,401 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : gen_action_details + +import random + +from metagpt.environment.stanford_town.env_space import EnvObsParams, EnvObsType +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class GenActionSector(STAction): + name: str = "GenActionSector" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cleaned_response = llm_resp.split("}")[0] + return cleaned_response + + def _func_validate(self, llm_resp: str, prompt: str): + if len(llm_resp.strip()) < 1: + return False + if "}" not in llm_resp: + return False + if "," in llm_resp: + return False + return True + + def _func_fail_default_resp(self): + fs = "kitchen" + return fs + + async def run(self, role: "STRole", access_tile: dict[str, str], act_desp: str): + def create_prompt_input(role, access_tile: dict[str, str], act_desp): + act_world = f"{access_tile['world']}" + + prompt_input = [] + + prompt_input += [role.scratch.get_str_name()] + prompt_input += [role.scratch.living_area.split(":")[1]] + x = f"{act_world}:{role.scratch.living_area.split(':')[1]}" + prompt_input += [role.s_mem.get_str_accessible_sector_arenas(x)] + + prompt_input += [role.scratch.get_str_name()] + prompt_input += [f"{access_tile['sector']}"] + x = f"{act_world}:{access_tile['sector']}" + prompt_input += [role.s_mem.get_str_accessible_sector_arenas(x)] + + if role.scratch.get_str_daily_plan_req() != "": + prompt_input += [f"\n{role.scratch.get_str_daily_plan_req()}"] + else: + prompt_input += [""] + + # MAR 11 TEMP + prompt_input = [] + act_world = access_tile["world"] + accessible_sector_str = role.s_mem.get_str_accessible_sectors(act_world) + curr = accessible_sector_str.split(", ") + fin_accessible_sectors = [] + for i in curr: + if "'s house" in i: + if role.scratch.last_name in i: + fin_accessible_sectors += [i] + else: + fin_accessible_sectors += [i] + accessible_sector_str = ", ".join(fin_accessible_sectors) + # END MAR 11 TEMP + + prompt_input += [accessible_sector_str] + + act_desp_1 = act_desp + act_desp_2 = act_desp + if "(" in act_desp: + act_desp_1 = act_desp.split("(")[0].strip() + act_desp_2 = act_desp.split("(")[-1][:-1] + prompt_input += [role.scratch.get_str_name()] + prompt_input += [act_desp_1] + + prompt_input += [act_desp_2] + prompt_input += [role.scratch.get_str_name()] + return prompt_input + + prompt_template = "action_location_sector_v1.txt" + prompt_input = create_prompt_input(role, access_tile, act_desp) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=15) + y = f"{access_tile['world']}" + x = [i.strip() for i in role.s_mem.get_str_accessible_sectors(y).split(",")] + if output not in x: + # output = random.choice(x) + output = role.scratch.living_area.split(":")[1] + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenActionArena(STAction): + name: str = "GenActionArena" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cleaned_response = llm_resp.split("}")[0] + return cleaned_response + + def _func_validate(self, llm_resp: str, prompt: str): + if len(llm_resp.strip()) < 1: + return False + if "}" not in llm_resp: + return False + if "," in llm_resp: + return False + return True + + def _func_fail_default_resp(self): + fs = "kitchen" + return fs + + async def run(self, role: "STRole", act_desp: str, act_world: str, act_sector: str): + def create_prompt_input(role, act_desp, act_world, act_sector): + prompt_input = [] + prompt_input += [role.scratch.get_str_name()] + x = f"{act_world}:{act_sector}" + prompt_input += [act_sector] + + # MAR 11 TEMP + accessible_arena_str = role.s_mem.get_str_accessible_sector_arenas(x) + curr = accessible_arena_str.split(", ") + fin_accessible_arenas = [] + for i in curr: + if "'s room" in i: + if role.scratch.last_name in i: + fin_accessible_arenas += [i] + else: + fin_accessible_arenas += [i] + accessible_arena_str = ", ".join(fin_accessible_arenas) + # END MAR 11 TEMP + prompt_input += [accessible_arena_str] + act_desp_1 = act_desp + act_desp_2 = act_desp + if "(" in act_desp: + act_desp_1 = act_desp.split("(")[0].strip() + act_desp_2 = act_desp.split("(")[-1][:-1] + prompt_input += [role.scratch.get_str_name()] + prompt_input += [act_desp_1] + + prompt_input += [act_desp_2] + prompt_input += [role.scratch.get_str_name()] + + prompt_input += [act_sector] + prompt_input += [accessible_arena_str] + return prompt_input + + prompt_template = "action_location_object_vMar11.txt" + prompt_input = create_prompt_input(role, act_desp, act_world, act_sector) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=15) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenActionObject(STAction): + name: str = "GenActionObject" + + def _func_validate(self, llm_resp: str, prompt: str): + if len(llm_resp.strip()) < 1: + return False + return True + + def _func_cleanup(self, llm_resp: str, prompt: str): + cleaned_response = llm_resp.strip() + return cleaned_response + + def _func_fail_default_resp(self): + fs = "bed" + return fs + + async def run(self, role: "STRole", act_desp: str, temp_address: str): + def create_prompt_input(role, act_desp, temp_address): + prompt_input = [] + if "(" in act_desp: + act_desp = act_desp.split("(")[-1][:-1] + + prompt_input += [act_desp] + prompt_input += [role.s_mem.get_str_accessible_arena_game_objects(temp_address)] + return prompt_input + + prompt_template = "action_object_v2.txt" + prompt_input = create_prompt_input(role, act_desp, temp_address) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=15) + x = [i.strip() for i in role.s_mem.get_str_accessible_arena_game_objects(temp_address).split(",")] + if output not in x: + output = random.choice(x) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenPronunciatio(STAction): + name: str = "GenPronunciatio" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cr = llm_resp.strip() + if len(cr) > 3: + cr = cr[:3] + return cr + + def _func_validate(self, llm_resp: str, prompt: str): + try: + self._func_cleanup(llm_resp, prompt="") + if len(llm_resp) == 0: + return False + except Exception: + return False + return True + + def _func_fail_default_resp(self): + fs = "😋" + return fs + + async def run(self, role: "STRole", act_desp: str): + def create_prompt_input(act_desp): + if "(" in act_desp: + act_desp = act_desp.split("(")[-1].split(")")[0] + prompt_input = [act_desp] + return prompt_input + + prompt_template = "generate_pronunciatio_v1.txt" + prompt_input = create_prompt_input(act_desp) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + example_output = "🛁🧖‍♀️" + special_instruction = "The value for the output must ONLY contain the emojis." + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenEventTriple(STAction): + name: str = "GenEventTriple" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cr = llm_resp.strip() + cr = [i.strip() for i in cr.split(")")[0].split(",")] + return cr + + def _func_validate(self, llm_resp: str, prompt: str): + try: + llm_resp = self._func_cleanup(llm_resp, prompt="") + if len(llm_resp) != 2: + return False + except Exception: + return False + return True + + def _func_fail_default_resp(self, role): + fs = (role.name, "is", "idle") + return fs + + async def run(self, role: "STRole", act_desp: str): + def create_prompt_input(role, act_desp): + if "(" in act_desp: + act_desp = act_desp.split("(")[-1].split(")")[0] + prompt_input = [role.name, act_desp, role.name] + return prompt_input + + prompt_template = "generate_event_triple_v1.txt" + prompt_input = create_prompt_input(role, act_desp) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp(role) + output = await self._run_gpt35_max_tokens(prompt, max_tokens=30) + output = (role.name, output[0], output[1]) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenActObjDescription(STAction): + name: str = "GenActObjDescription" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cr = llm_resp.strip() + if cr[-1] == ".": + cr = cr[:-1] + return cr + + def _func_validate(self, llm_resp: str, prompt: str): + try: + llm_resp = self._func_cleanup(llm_resp, prompt="") + except Exception: + return False + return True + + def _func_fail_default_resp(self, act_game_object): + fs = f"{act_game_object} is idle" + return fs + + async def run(self, role: "STRole", act_game_object: str, act_desp: str): + def create_prompt_input(act_game_object, act_desp, role): + prompt_input = [act_game_object, role.name, act_desp, act_game_object, act_game_object] + return prompt_input + + prompt_template = "generate_obj_event_v1.txt" + prompt_input = create_prompt_input(act_game_object, act_desp, role) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + example_output = "being fixed" + special_instruction = "The output should ONLY contain the phrase that should go in ." + self.fail_default_resp = self._func_fail_default_resp(act_game_object) + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenObjEventTriple(STAction): + name: str = "GenObjEventTriple" + + def _func_cleanup(self, llm_resp: str, prompt: str): + cr = llm_resp.strip() + cr = [i.strip() for i in cr.split(")")[0].split(",")] + return cr + + def _func_validate(self, llm_resp: str, prompt: str): + try: + llm_resp = self._func_cleanup(llm_resp, prompt="") + if len(llm_resp) != 2: + return False + except Exception: + return False + return True + + def _func_fail_default_resp(self, act_game_object: str): + fs = (act_game_object, "is", "idle") + return fs + + async def run(self, role: "STRole", act_game_object, act_obj_desp): + def create_prompt_input(act_game_object, act_obj_desp): + prompt_input = [act_game_object, act_obj_desp, act_game_object] + return prompt_input + + prompt_template = "generate_event_triple_v1.txt" + prompt_input = create_prompt_input(act_game_object, act_obj_desp) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp(act_game_object) + output = await self._run_gpt35_max_tokens(prompt, max_tokens=30) + output = (act_game_object, output[0], output[1]) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +class GenActionDetails(STAction): + name: str = "GenActionDetails" + + def _func_cleanup(self, llm_resp: str, prompt: str) -> list: + pass + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + # TODO -- this sometimes generates error + try: + self._func_cleanup(llm_resp) + except Exception: + return False + return True + + def _func_fail_default_resp(self): + fs = {} + return fs + + async def run(self, role: "STRole", act_desp: str, act_dura): + access_tile = role.rc.env.observe( + obs_params=EnvObsParams(obs_type=EnvObsType.GET_TITLE, coord=role.scratch.curr_tile) + ) + act_world = access_tile["world"] + act_sector = await GenActionSector().run(role, access_tile, act_desp) + act_arena = await GenActionArena().run(role, act_desp, act_world, act_sector) + act_address = f"{act_world}:{act_sector}:{act_arena}" + if not role.s_mem.get_str_accessible_arena_game_objects(act_address): + act_game_object = "" + else: + act_game_object = await GenActionObject().run(role, act_desp, act_address) + new_address = f"{act_world}:{act_sector}:{act_arena}:{act_game_object}" + act_pron = await GenPronunciatio().run(role, act_desp) + act_event = await GenEventTriple().run(role, act_desp) + # Persona's actions also influence the object states. We set those up here. + act_obj_desp = await GenActObjDescription().run(role, act_game_object, act_desp) + act_obj_pron = await GenPronunciatio().run(role, act_obj_desp) + act_obj_event = await GenObjEventTriple().run(role, act_game_object, act_obj_desp) + result_dict = { + "action_address": new_address, + "action_duration": int(act_dura), + "action_description": act_desp, + "action_pronunciatio": act_pron, + "action_event": act_event, + "chatting_with": None, + "chat": None, + "chatting_with_buffer": None, + "chatting_end_time": None, + "act_obj_description": act_obj_desp, + "act_obj_pronunciatio": act_obj_pron, + "act_obj_event": act_obj_event, + } + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {result_dict}") + return result_dict diff --git a/metagpt/ext/stanford_town/actions/gen_daily_schedule.py b/metagpt/ext/stanford_town/actions/gen_daily_schedule.py new file mode 100644 index 000000000..5dffa8995 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/gen_daily_schedule.py @@ -0,0 +1,60 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : gen_daily_schedule + + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class GenDailySchedule(STAction): + name: str = "GenDailySchedule" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt="") + except Exception: + return False + return True + + def _func_cleanup(self, llm_resp: str, prompt: str) -> list: + cr = [] + _cr = llm_resp.split(")") + for i in _cr: + if i[-1].isdigit(): + i = i[:-1].strip() + if i[-1] == "." or i[-1] == ",": + cr += [i[:-1].strip()] + return cr + + def _func_fail_default_resp(self) -> int: + fs = [ + "wake up and complete the morning routine at 6:00 am", + "eat breakfast at 7:00 am", + "read a book from 8:00 am to 12:00 pm", + "have lunch at 12:00 pm", + "take a nap from 1:00 pm to 4:00 pm", + "relax and watch TV from 7:00 pm to 8:00 pm", + "go to bed at 11:00 pm", + ] + return fs + + async def run(self, role: "STRole", wake_up_hour: str): + def create_prompt_input(role, wake_up_hour): + prompt_input = [] + prompt_input += [role.scratch.get_str_iss()] + prompt_input += [role.scratch.get_str_lifestyle()] + prompt_input += [role.scratch.get_str_curr_date_str()] + prompt_input += [role.scratch.get_str_firstname()] + prompt_input += [f"{str(wake_up_hour)}:00 am"] + return prompt_input + + wake_up_hour = int(wake_up_hour) + prompt_template = "daily_planning_v6.txt" + prompt_input = create_prompt_input(role, wake_up_hour) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=500) + output = [f"wake up and complete the morning routine at {wake_up_hour}:00 am"] + output + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/gen_hourly_schedule.py b/metagpt/ext/stanford_town/actions/gen_hourly_schedule.py new file mode 100644 index 000000000..5d59f96dd --- /dev/null +++ b/metagpt/ext/stanford_town/actions/gen_hourly_schedule.py @@ -0,0 +1,181 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : gen_hourly_schedule + +import random +import string + +from metagpt.logs import logger + +from .st_action import STAction + + +def get_random_alphanumeric(i=6, j=6): + """ + Returns a random alpha numeric strength that has the length of somewhere + between i and j. + + INPUT: + i: min_range for the length + j: max_range for the length + OUTPUT: + an alpha numeric str with the length of somewhere between i and j. + """ + k = random.randint(i, j) + x = "".join(random.choices(string.ascii_letters + string.digits, k=k)) + return x + + +class GenHourlySchedule(STAction): + name: str = "GenHourlySchedule" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt="") + except Exception: + return False + return True + + def _func_cleanup(self, llm_resp: str, prompt: str) -> list: + cr = llm_resp.strip() + if cr[-1] == ".": + cr = cr[:-1] + # to only use the first line of output + cr = cr.split("\n")[0] + return cr + + def _func_fail_default_resp(self) -> int: + fs = "asleep" + return fs + + async def _generate_schedule_for_given_hour( + self, role: "STRole", curr_hour_str, p_f_ds_hourly_org, hour_str, intermission2=None + ): + def create_prompt_input(persona, curr_hour_str, p_f_ds_hourly_org, hour_str, intermission2=None): + schedule_format = "" + for i in hour_str: + schedule_format += f"[{persona.scratch.get_str_curr_date_str()} -- {i}]" + schedule_format += " Activity: [Fill in]\n" + schedule_format = schedule_format[:-1] + + intermission_str = "Here the originally intended hourly breakdown of" + intermission_str += f" {persona.scratch.get_str_firstname()}'s schedule today: " + for count, i in enumerate(persona.scratch.daily_req): + intermission_str += f"{str(count + 1)}) {i}, " + intermission_str = intermission_str[:-2] + + prior_schedule = "" + if p_f_ds_hourly_org: + prior_schedule = "\n" + for count, i in enumerate(p_f_ds_hourly_org): + prior_schedule += f"[(ID:{get_random_alphanumeric()})" + prior_schedule += f" {persona.scratch.get_str_curr_date_str()} --" + prior_schedule += f" {hour_str[count]}] Activity:" + prior_schedule += f" {persona.scratch.get_str_firstname()}" + prior_schedule += f" is {i}\n" + + prompt_ending = f"[(ID:{get_random_alphanumeric()})" + prompt_ending += f" {persona.scratch.get_str_curr_date_str()}" + prompt_ending += f" -- {curr_hour_str}] Activity:" + prompt_ending += f" {persona.scratch.get_str_firstname()} is" + + if intermission2: + intermission2 = f"\n{intermission2}" + + prompt_input = [] + prompt_input += [schedule_format] + prompt_input += [persona.scratch.get_str_iss()] + + prompt_input += [prior_schedule + "\n"] + prompt_input += [intermission_str] + if intermission2: + prompt_input += [intermission2] + else: + prompt_input += [""] + prompt_input += [prompt_ending] + + return prompt_input + + prompt_template = "generate_hourly_schedule_v2.txt" + prompt_input = create_prompt_input(role, curr_hour_str, p_f_ds_hourly_org, hour_str, intermission2) + prompt_input_str = "\n".join(prompt_input) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, prompt_template) + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=50) + logger.info( + f"Role: {role.name} _generate_schedule_for_given_hour prompt_input: {prompt_input_str}, " + f"output: {output}" + ) + return output + + async def run(self, role: "STRole", wake_up_hour: int): + hour_str = [ + "00:00 AM", + "01:00 AM", + "02:00 AM", + "03:00 AM", + "04:00 AM", + "05:00 AM", + "06:00 AM", + "07:00 AM", + "08:00 AM", + "09:00 AM", + "10:00 AM", + "11:00 AM", + "12:00 PM", + "01:00 PM", + "02:00 PM", + "03:00 PM", + "04:00 PM", + "05:00 PM", + "06:00 PM", + "07:00 PM", + "08:00 PM", + "09:00 PM", + "10:00 PM", + "11:00 PM", + ] + n_m1_activity = [] + diversity_repeat_count = 1 # TODO mg 1->3 + for i in range(diversity_repeat_count): + logger.info(f"diversity_repeat_count idx: {i}") + n_m1_activity_set = set(n_m1_activity) + if len(n_m1_activity_set) < 5: + n_m1_activity = [] + for count, curr_hour_str in enumerate(hour_str): + if wake_up_hour > 0: + n_m1_activity += ["sleeping"] + wake_up_hour -= 1 + else: + logger.info(f"_generate_schedule_for_given_hour idx: {count}, n_m1_activity: {n_m1_activity}") + n_m1_activity += [ + await self._generate_schedule_for_given_hour(role, curr_hour_str, n_m1_activity, hour_str) + ] + + # Step 1. Compressing the hourly schedule to the following format: + # The integer indicates the number of hours. They should add up to 24. + # [['sleeping', 6], ['waking up and starting her morning routine', 1], + # ['eating breakfast', 1], ['getting ready for the day', 1], + # ['working on her painting', 2], ['taking a break', 1], + # ['having lunch', 1], ['working on her painting', 3], + # ['taking a break', 2], ['working on her painting', 2], + # ['relaxing and watching TV', 1], ['going to bed', 1], ['sleeping', 2]] + _n_m1_hourly_compressed = [] + prev = None + prev_count = 0 + for i in n_m1_activity: + if i != prev: + prev_count = 1 + _n_m1_hourly_compressed += [[i, prev_count]] + prev = i + elif _n_m1_hourly_compressed: + _n_m1_hourly_compressed[-1][1] += 1 + + # Step 2. Expand to min scale (from hour scale) + # [['sleeping', 360], ['waking up and starting her morning routine', 60], + # ['eating breakfast', 60],.. + n_m1_hourly_compressed = [] + for task, duration in _n_m1_hourly_compressed: + n_m1_hourly_compressed += [[task, duration * 60]] + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {n_m1_hourly_compressed}") + return n_m1_hourly_compressed diff --git a/metagpt/ext/stanford_town/actions/gen_iter_chat_utt.py b/metagpt/ext/stanford_town/actions/gen_iter_chat_utt.py new file mode 100644 index 000000000..40f6d3af0 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/gen_iter_chat_utt.py @@ -0,0 +1,125 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : generate_iterative_chat_utt + +from metagpt.environment.stanford_town.env_space import EnvObsParams, EnvObsType +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.ext.stanford_town.utils.utils import extract_first_json_dict +from metagpt.logs import logger + + +class GenIterChatUTT(STAction): + name: str = "GenIterChatUTT" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + resp = False + try: + _ = extract_first_json_dict(llm_resp) + resp = True + except Exception: + pass + return resp + + def _func_cleanup(self, llm_resp: str, prompt: str) -> dict: + gpt_response = extract_first_json_dict(llm_resp) + + cleaned_dict = dict() + cleaned = [] + for key, val in gpt_response.items(): + cleaned += [val] + cleaned_dict["utterance"] = cleaned[0] + cleaned_dict["end"] = True + if "f" in str(cleaned[1]) or "F" in str(cleaned[1]): + cleaned_dict["end"] = False + + return cleaned_dict + + def _func_fail_default_resp(self) -> dict: + cleaned_dict = dict() + cleaned_dict["utterance"] = "..." + cleaned_dict["end"] = False + return cleaned_dict + + async def run( + self, + init_role: "STRole", + target_role: "STRole", + retrieved: dict, + curr_context: str, + curr_chat: list[str], + *args, + **kwargs, + ) -> dict: + def create_prompt_input( + access_tile: dict[str, str], + init_role: "STRole", + target_role: "STRole", + retrieved: dict, + curr_context: str, + curr_chat: list[str], + ): + role = init_role + scratch = role.rc.scratch + target_scratch = target_role.rc.scratch + prev_convo_insert = "\n" + if role.rc.memory.chat_list: + for i in role.rc.memory.chat_list: + if i.object == target_role.name: + v1 = int((scratch.curr_time - i.created).total_seconds() / 60) + prev_convo_insert += ( + f"{str(v1)} minutes ago, {scratch.name} and " + f"{target_scratch.name} were already {i.description} " + f"This context takes place after that conversation." + ) + break + if prev_convo_insert == "\n": + prev_convo_insert = "" + if role.rc.memory.chat_list: + if int((scratch.curr_time - role.rc.memory.chat_list[-1].created).total_seconds() / 60) > 480: + prev_convo_insert = "" + logger.info(f"prev_convo_insert: {prev_convo_insert}") + + curr_sector = f"{access_tile['sector']}" + curr_arena = f"{access_tile['arena']}" + curr_location = f"{curr_arena} in {curr_sector}" + + retrieved_str = "" + for key, vals in retrieved.items(): + for v in vals: + retrieved_str += f"- {v.description}\n" + + convo_str = "" + for i in curr_chat: + convo_str += ": ".join(i) + "\n" + if convo_str == "": + convo_str = "[The conversation has not started yet -- start it!]" + + init_iss = f"Here is Here is a brief description of {scratch.name}.\n{scratch.get_str_iss()}" + prompt_input = [ + init_iss, + scratch.name, + retrieved_str, + prev_convo_insert, + curr_location, + curr_context, + scratch.name, + target_scratch.name, + convo_str, + scratch.name, + target_scratch.name, + scratch.name, + scratch.name, + scratch.name, + ] + return prompt_input + + access_tile = init_role.rc.env.observe( + obs_params=EnvObsParams(obs_type=EnvObsType.GET_TITLE, coord=init_role.scratch.curr_tile) + ) + prompt_input = create_prompt_input(access_tile, init_role, target_role, retrieved, curr_context, curr_chat) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "iterative_convo_v1.txt") + # original using `ChatGPT_safe_generate_response_OLD` + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_wo_extra_prompt(prompt) + logger.info(f"Role: {init_role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/inner_voice_action.py b/metagpt/ext/stanford_town/actions/inner_voice_action.py new file mode 100644 index 000000000..83cfa037b --- /dev/null +++ b/metagpt/ext/stanford_town/actions/inner_voice_action.py @@ -0,0 +1,35 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class AgentWhisperThoughtAction(STAction): + name: str = "AgentWhisperThoughtAction" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> list: + return llm_resp.split('"')[0].strip() + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, test_input=None, verbose=False) -> str: + def create_prompt_input(role: "STRole", statements, test_input=None): + prompt_input = [role.scratch.name, statements] + return prompt_input + + prompt_input = create_prompt_input(role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "whisper_inner_thought_v1.txt") + + output = await self._run_gpt35_max_tokens(prompt, max_tokens=50) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/new_decomp_schedule.py b/metagpt/ext/stanford_town/actions/new_decomp_schedule.py new file mode 100644 index 000000000..759ec170f --- /dev/null +++ b/metagpt/ext/stanford_town/actions/new_decomp_schedule.py @@ -0,0 +1,154 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : new_decomp_schedule + +import datetime + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class NewDecompSchedule(STAction): + name: str = "NewDecompSchedule" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + resp = False + try: + llm_resp = self._func_cleanup(llm_resp, prompt) + dur_sum = 0 + for act, dur in llm_resp: + dur_sum += dur + if isinstance(act, str): + return False + if isinstance(dur, int): + return False + x = prompt.split("\n")[0].split("originally planned schedule from")[-1].strip()[:-1] + x = [datetime.datetime.strptime(i.strip(), "%H:%M %p") for i in x.split(" to ")] + delta_min = int((x[1] - x[0]).total_seconds() / 60) + + if int(dur_sum) != int(delta_min): + return False + except Exception: + pass + return resp + + def _func_cleanup(self, llm_resp: str, prompt: str) -> list: + new_schedule = prompt + " " + llm_resp.strip() + new_schedule = new_schedule.split("The revised schedule:")[-1].strip() + new_schedule = new_schedule.split("\n") + + ret_temp = [] + for i in new_schedule: + ret_temp += [i.split(" -- ")] + + ret = [] + for time_str, action in ret_temp: + start_time = time_str.split(" ~ ")[0].strip() + end_time = time_str.split(" ~ ")[1].strip() + delta = datetime.datetime.strptime(end_time, "%H:%M") - datetime.datetime.strptime(start_time, "%H:%M") + delta_min = int(delta.total_seconds() / 60) + if delta_min < 0: + delta_min = 0 + ret += [[action, delta_min]] + + return ret + + def _func_fail_default_resp(self, main_act_dur: int, truncated_act_dur: int) -> int: + dur_sum = 0 + for act, dur in main_act_dur: + dur_sum += dur + + ret = truncated_act_dur[:] + ret += main_act_dur[len(ret) - 1 :] + + # If there are access, we need to trim... + ret_dur_sum = 0 + count = 0 + over = None + for act, dur in ret: + ret_dur_sum += dur + if ret_dur_sum == dur_sum: + break + if ret_dur_sum > dur_sum: + over = ret_dur_sum - dur_sum + break + count += 1 + + if over: + ret = ret[: count + 1] + ret[-1][1] -= over + + return ret + + async def run( + self, + role: "STRole", + main_act_dur: int, + truncated_act_dur: int, + start_time_hour: datetime, + end_time_hour: datetime, + inserted_act: str, + inserted_act_dur: int, + *args, + **kwargs, + ): + def create_prompt_input( + role: "STRole", + main_act_dur: int, + truncated_act_dur: int, + start_time_hour: datetime, + end_time_hour: datetime, + inserted_act: str, + inserted_act_dur: int, + ): + persona_name = role.name + start_hour_str = start_time_hour.strftime("%H:%M %p") + end_hour_str = end_time_hour.strftime("%H:%M %p") + + original_plan = "" + for_time = start_time_hour + for i in main_act_dur: + original_plan += ( + f'{for_time.strftime("%H:%M")} ~ ' + f'{(for_time + datetime.timedelta(minutes=int(i[1]))).strftime("%H:%M")} -- ' + i[0] + ) + original_plan += "\n" + for_time += datetime.timedelta(minutes=int(i[1])) + + new_plan_init = "" + for_time = start_time_hour + for count, i in enumerate(truncated_act_dur): + new_plan_init += ( + f'{for_time.strftime("%H:%M")} ~ ' + f'{(for_time + datetime.timedelta(minutes=int(i[1]))).strftime("%H:%M")} -- ' + i[0] + ) + new_plan_init += "\n" + if count < len(truncated_act_dur) - 1: + for_time += datetime.timedelta(minutes=int(i[1])) + + new_plan_init += (for_time + datetime.timedelta(minutes=int(i[1]))).strftime("%H:%M") + " ~" + + prompt_input = [ + persona_name, + start_hour_str, + end_hour_str, + original_plan, + persona_name, + inserted_act, + inserted_act_dur, + persona_name, + start_hour_str, + end_hour_str, + end_hour_str, + new_plan_init, + ] + return prompt_input + + prompt_input = create_prompt_input( + role, main_act_dur, truncated_act_dur, start_time_hour, end_time_hour, inserted_act, inserted_act_dur + ) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "new_decomp_schedule_v1.txt") + self.fail_default_resp = self._func_fail_default_resp(main_act_dur, truncated_act_dur) + output = await self._run_gpt35_max_tokens(prompt, max_tokens=1000) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/run_reflect_action.py b/metagpt/ext/stanford_town/actions/run_reflect_action.py new file mode 100644 index 000000000..895f6828f --- /dev/null +++ b/metagpt/ext/stanford_town/actions/run_reflect_action.py @@ -0,0 +1,277 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : Integration Reflect Action + +import re + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +# Run GPT Prompt Focal Point method +class AgentFocusPt(STAction): + name: str = "AgentFocusPt" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> str: + try: + """ + Cleanup handling has been completed for run_v2 + """ + return llm_resp + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, n: int, test_input=None) -> str: + def create_prompt_input(role: "STRole", statements, n, test_input=None): + prompt_input = [statements, str(n)] + return prompt_input + + prompt_input = create_prompt_input(role, statements, n) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "generate_focal_pt_v1.txt") + + example_output = '["What should Jane do for lunch", "Does Jane like strawberry", "Who is Jane"]' + special_instruction = "Output must be a list of str." + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Insight and Guidance +class AgentInsightAndGuidance(STAction): + name: str = "AgentInsightAndGuidance" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> dict: + try: + llm_resp = "1. " + llm_resp.strip() + ret = dict() + for i in llm_resp.split("\n"): + row = " ".join(i.split(". ")[1:]) + if "(because of " not in row: + continue + thought = row.split("(because of ")[0].strip() + if ")" not in row.split("(because of ")[1]: + continue + evi_raw = row.split("(because of ")[1].split(")")[0].strip() + evi_raw = re.findall(r"\d+", evi_raw) + evi_raw = [int(i.strip()) for i in evi_raw] + ret[thought] = evi_raw + return ret + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self, n: int) -> str: + return ["I am hungry"] * n + + async def run(self, role: "STRole", statements: str, n: int, test_input=None) -> dict: + def create_prompt_input(role, statements, n, test_input=None): + prompt_input = [statements, str(n)] + return prompt_input + + prompt_input = create_prompt_input(role, statements, n) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "insight_and_evidence_v1.txt") + + self.fail_default_resp = self._func_fail_default_resp(n) + output = await self._run_gpt35_max_tokens(prompt, max_tokens=150) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Event Triple +class AgentEventTriple(STAction): + name: str = "AgentEventTriple" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + llm_resp = self._func_cleanup(llm_resp, prompt="") + if len(llm_resp) != 2: + return False + except Exception: + return False + return True + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> list: + try: + cr = llm_resp.strip() + cr = [i.strip() for i in cr.split(")")[0].split(",")] + if len(cr) != 2: + return cr[-2:] + return cr + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, statements: str, role: "STRole", verbose=False) -> tuple: + def create_prompt_input(statements, role): + if "(" in statements: + statements = statements.split("(")[-1].split(")")[0] + prompt_input = [role.scratch.name, statements, role.scratch.name] + return prompt_input + + prompt_input = create_prompt_input(statements, role) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "generate_event_triple_v1.txt") + + output = await self._run_gpt35_max_tokens(prompt, max_tokens=30) + output = (role.scratch.name, output[0], output[1]) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Event Poignancy +class AgentEventPoignancy(STAction): + name: str = "AgentEventPoignancy" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> int: + try: + llm_resp = int(llm_resp.strip()) + return llm_resp + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, test_input=None, verbose=False) -> str: + def create_prompt_input(role: "STRole", statements: str, test_input=None): + prompt_input = [role.scratch.name, role.scratch.get_str_iss(), role.scratch.name, statements] + return prompt_input + + prompt_input = create_prompt_input(role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "poignancy_event_v1.txt") + + example_output = "5" # ######## + special_instruction = "The output should ONLY contain ONE integer value on the scale of 1 to 10." + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Chat Poignancy +class AgentChatPoignancy(STAction): + name: str = "AgentChatPoignancy" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> int: + try: + llm_resp = int(llm_resp.strip()) + return llm_resp + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, test_input=None, verbose=False) -> str: + def create_prompt_input(role: "STRole", statements, test_input=None): + prompt_input = [role.scratch.name, role.scratch.get_str_iss(), role.scratch.name, statements] + return prompt_input + + prompt_input = create_prompt_input(role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "poignancy_chat_v1.txt") + + example_output = "5" # ######## + special_instruction = "The output should ONLY contain ONE integer value on the scale of 1 to 10." + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Planning Thought on Convo +class AgentPlanThoughtOnConvo(STAction): + name: str = "AgentPlanThoughtOnConvo" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> str: + try: + return llm_resp.split('"')[0].strip() + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, test_input=None, verbose=False) -> str: + def create_prompt_input(role, statements, test_input=None): + prompt_input = [statements, role.scratch.name, role.scratch.name, role.scratch.name] + return prompt_input + + prompt_input = create_prompt_input(role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "planning_thought_on_convo_v1.txt") + + output = await self._run_gpt35_max_tokens(prompt, max_tokens=50) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output + + +# Run GPT Prompt Memory on Convo +class AgentMemoryOnConvo(STAction): + name: str = "AgentMemoryOnConvo" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + try: + self._func_cleanup(llm_resp, prompt) + return True + except Exception: + return False + + def _func_cleanup(self, llm_resp: str, prompt: str = "") -> str: + try: + return llm_resp.split('"')[0].strip() + except Exception as exp: + logger.error(f"{self.cls_name} with error {exp}") + + def _func_fail_default_resp(self) -> str: + pass + + async def run(self, role: "STRole", statements: str, test_input=None, verbose=False) -> str: + def create_prompt_input(role, statements, test_input=None): + prompt_input = [statements, role.scratch.name, role.scratch.name, role.scratch.name] + return prompt_input + + prompt_input = create_prompt_input(role, statements) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "memo_on_convo_v1.txt") + example_output = "Jane Doe was interesting to talk to." + special_instruction = ( + "The output should ONLY contain a string that summarizes anything interesting " + "that the agent may have noticed" + ) + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/st_action.py b/metagpt/ext/stanford_town/actions/st_action.py new file mode 100644 index 000000000..321676374 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/st_action.py @@ -0,0 +1,119 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : StanfordTown Action +import json +import time +from abc import abstractmethod +from pathlib import Path +from typing import Any, Optional, Union + +from metagpt.actions.action import Action +from metagpt.config2 import config +from metagpt.ext.stanford_town.utils.const import PROMPTS_DIR +from metagpt.logs import logger + + +class STAction(Action): + name: str = "STAction" + prompt_dir: Path = PROMPTS_DIR + fail_default_resp: Optional[str] = None + + @property + def cls_name(self): + return self.__class__.__name__ + + @abstractmethod + def _func_validate(self, llm_resp: str, prompt: str): + raise NotImplementedError + + @abstractmethod + def _func_cleanup(self, llm_resp: str, prompt: str): + raise NotImplementedError + + @abstractmethod + def _func_fail_default_resp(self): + raise NotImplementedError + + def generate_prompt_with_tmpl_filename(self, prompt_input: Union[str, list], tmpl_filename) -> str: + """ + same with `generate_prompt` + Args: + prompt_input: the input we want to feed in (IF THERE ARE MORE THAN ONE INPUT, THIS CAN BE A LIST.) + tmpl_filename: prompt template filename + Returns: + a str prompt that will be sent to LLM server. + """ + if isinstance(prompt_input, str): + prompt_input = [prompt_input] + prompt_input = [str(i) for i in prompt_input] + + f = open(str(self.prompt_dir.joinpath(tmpl_filename)), "r") + prompt = f.read() + f.close() + for count, i in enumerate(prompt_input): + prompt = prompt.replace(f"!!", i) + if "###" in prompt: + prompt = prompt.split("###")[1] + return prompt.strip() + + async def _aask(self, prompt: str) -> str: + return await self.llm.aask(prompt) + + async def _run_gpt35_max_tokens(self, prompt: str, max_tokens: int = 50, retry: int = 3): + for idx in range(retry): + try: + tmp_max_tokens_rsp = getattr(config.llm, "max_token", 1500) + setattr(config.llm, "max_token", max_tokens) + self.llm.use_system_prompt = False # to make it behave like a non-chat completions + + llm_resp = await self._aask(prompt) + + setattr(config.llm, "max_token", tmp_max_tokens_rsp) + logger.info(f"Action: {self.cls_name} llm _run_gpt35_max_tokens raw resp: {llm_resp}") + if self._func_validate(llm_resp, prompt): + return self._func_cleanup(llm_resp, prompt) + except Exception as exp: + logger.warning(f"Action: {self.cls_name} _run_gpt35_max_tokens exp: {exp}") + time.sleep(5) + return self.fail_default_resp + + async def _run_gpt35( + self, prompt: str, example_output: str, special_instruction: str, retry: int = 3 + ) -> Union[bool, Any]: + """same with `gpt_structure.ChatGPT_safe_generate_response`""" + prompt = '"""\n' + prompt + '\n"""\n' + prompt += f"Output the response to the prompt above in json. {special_instruction}\n" + prompt += "Example output json:\n" + prompt += '{"output": "' + str(example_output) + '"}' + + for idx in range(retry): + try: + llm_resp = await self._aask(prompt) + logger.info(f"Action: {self.cls_name} llm _run_gpt35 raw resp: {llm_resp}") + end_idx = llm_resp.strip().rfind("}") + 1 + llm_resp = llm_resp[:end_idx] + llm_resp = json.loads(llm_resp)["output"] + + if self._func_validate(llm_resp, prompt): + return self._func_cleanup(llm_resp, prompt) + except Exception as exp: + logger.warning(f"Action: {self.cls_name} _run_gpt35 exp: {exp}") + time.sleep(5) # usually avoid `Rate limit` + return False + + async def _run_gpt35_wo_extra_prompt(self, prompt: str, retry: int = 3) -> str: + for idx in range(retry): + try: + llm_resp = await self._aask(prompt) + llm_resp = llm_resp.strip() + logger.info(f"Action: {self.cls_name} llm _run_gpt35_wo_extra_prompt raw resp: {llm_resp}") + if self._func_validate(llm_resp, prompt): + return self._func_cleanup(llm_resp, prompt) + except Exception as exp: + logger.warning(f"Action: {self.cls_name} _run_gpt35_wo_extra_prompt exp: {exp}") + time.sleep(5) # usually avoid `Rate limit` + return self.fail_default_resp + + async def run(self, *args, **kwargs): + """Run action""" + raise NotImplementedError("The run method should be implemented in a subclass.") diff --git a/metagpt/ext/stanford_town/actions/summarize_conv.py b/metagpt/ext/stanford_town/actions/summarize_conv.py new file mode 100644 index 000000000..5be5fcaa4 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/summarize_conv.py @@ -0,0 +1,47 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : summarize the content of agents' conversation + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class SummarizeConv(STAction): + name: str = "SummarizeConv" + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + resp = False + try: + _ = self._func_cleanup(llm_resp, prompt) + resp = True + except Exception: + pass + return resp + + def _func_cleanup(self, llm_resp: str, prompt: str) -> str: + ret = "conversing about " + llm_resp.strip() + return ret + + def _func_fail_default_resp(self) -> str: + return "conversing with a housemate about morning greetings" + + async def run(self, conv: list): + def create_prompt_input(conversation: list): + convo_str = "" + for row in conversation: + convo_str += f'{row[0]}: "{row[1]}"\n' + prompt_input = [convo_str] + return prompt_input + + prompt_input = create_prompt_input(conv) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "summarize_conversation_v1.txt") + + example_output = "conversing about what to eat for lunch" + special_instruction = ( + "The output must continue the sentence above by filling in the tag. " + "Don't start with 'this is a conversation about...' Just finish the sentence " + "but do not miss any important details (including who are chatting)." + ) + output = await self._run_gpt35(prompt, example_output, special_instruction) + logger.info(f"Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/task_decomp.py b/metagpt/ext/stanford_town/actions/task_decomp.py new file mode 100644 index 000000000..3a23a7345 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/task_decomp.py @@ -0,0 +1,173 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : task_decomp + +import datetime + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class TaskDecomp(STAction): + name: str = "TaskDecomp" + + def _func_cleanup(self, llm_resp: str, prompt: str) -> list: + # TODO SOMETHING HERE sometimes fails... See screenshot + temp = [i.strip() for i in llm_resp.split("\n")] + _cr = [] + cr = [] + for count, i in enumerate(temp): + if count != 0: + _cr += [" ".join([j.strip() for j in i.split(" ")][3:])] + else: + _cr += [i] + for count, i in enumerate(_cr): + k = [j.strip() for j in i.split("(duration in minutes:")] + task = k[0] + if task[-1] == ".": + task = task[:-1] + duration = int(k[1].split(",")[0].strip()) + cr += [[task, duration]] + + total_expected_min = int(prompt.split("(total duration in minutes")[-1].split("):")[0].strip()) + + # TODO -- now, you need to make sure that this is the same as the sum of + # the current action sequence. + curr_min_slot = [ + ["dummy", -1], + ] # (task_name, task_index) + for count, i in enumerate(cr): + i_task = i[0] + i_duration = i[1] + + i_duration -= i_duration % 5 + if i_duration > 0: + for j in range(i_duration): + curr_min_slot += [(i_task, count)] + curr_min_slot = curr_min_slot[1:] + + if len(curr_min_slot) > total_expected_min: + last_task = curr_min_slot[60] + for i in range(1, 6): + curr_min_slot[-1 * i] = last_task + elif len(curr_min_slot) < total_expected_min: + last_task = curr_min_slot[-1] + for i in range(total_expected_min - len(curr_min_slot)): + curr_min_slot += [last_task] + + cr_ret = [ + ["dummy", -1], + ] + for task, task_index in curr_min_slot: + if task != cr_ret[-1][0]: + cr_ret += [[task, 1]] + else: + cr_ret[-1][1] += 1 + cr = cr_ret[1:] + + return cr + + def _func_validate(self, llm_resp: str, prompt: str) -> bool: + # TODO -- this sometimes generates error + try: + self._func_cleanup(llm_resp, prompt) + except Exception: + return False + return True + + def _func_fail_default_resp(self) -> int: + fs = [["asleep", 0]] + return fs + + async def run(self, role: "STRole", task_desc: int, truncated_act_dur: int, *args, **kwargs): + def create_prompt_input(role, task, duration): + """ + Today is Saturday June 25. From 00:00 ~ 06:00am, Maeve is + planning on sleeping, 06:00 ~ 07:00am, Maeve is + planning on waking up and doing her morning routine, + and from 07:00am ~08:00am, Maeve is planning on having breakfast. + """ + + curr_f_org_index = role.scratch.get_f_daily_schedule_hourly_org_index() + all_indices = [] + # if curr_f_org_index > 0: + # all_indices += [curr_f_org_index-1] + all_indices += [curr_f_org_index] + if curr_f_org_index + 1 <= len(role.scratch.f_daily_schedule_hourly_org): + all_indices += [curr_f_org_index + 1] + if curr_f_org_index + 2 <= len(role.scratch.f_daily_schedule_hourly_org): + all_indices += [curr_f_org_index + 2] + + curr_time_range = "" + + logger.debug("DEBUG") + logger.debug(role.scratch.f_daily_schedule_hourly_org) + logger.debug(all_indices) + + summ_str = f'Today is {role.scratch.curr_time.strftime("%B %d, %Y")}. ' + summ_str += "From " + for index in all_indices: + logger.debug(f"index {index}") + if index < len(role.scratch.f_daily_schedule_hourly_org): + start_min = 0 + for i in range(index): + start_min += role.scratch.f_daily_schedule_hourly_org[i][1] + end_min = start_min + role.scratch.f_daily_schedule_hourly_org[index][1] + start_time = datetime.datetime.strptime("00:00:00", "%H:%M:%S") + datetime.timedelta( + minutes=start_min + ) + end_time = datetime.datetime.strptime("00:00:00", "%H:%M:%S") + datetime.timedelta( + minutes=end_min + ) + start_time_str = start_time.strftime("%H:%M%p") + end_time_str = end_time.strftime("%H:%M%p") + summ_str += ( + f"{start_time_str} ~ {end_time_str}, {role.name} is planning " + f"on {role.scratch.f_daily_schedule_hourly_org[index][0]}, " + ) + if curr_f_org_index + 1 == index: + curr_time_range = f"{start_time_str} ~ {end_time_str}" + summ_str = summ_str[:-2] + "." + + prompt_input = [] + prompt_input += [role.scratch.get_str_iss()] + prompt_input += [summ_str] + # prompt_input += [role.scratch.get_str_curr_date_str()] + prompt_input += [role.scratch.get_str_firstname()] + prompt_input += [role.scratch.get_str_firstname()] + prompt_input += [task] + prompt_input += [curr_time_range] + prompt_input += [duration] + prompt_input += [role.scratch.get_str_firstname()] + return prompt_input + + prompt_input = create_prompt_input(role, task_desc, truncated_act_dur) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "task_decomp_v3.txt") + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=1000) + logger.info(f"Role: {role.name} {self.cls_name} output: {output}") + + fin_output = [] + time_sum = 0 + for i_task, i_duration in output: + time_sum += i_duration + # HM????????? + # if time_sum < duration: + if time_sum <= truncated_act_dur: + fin_output += [[i_task, i_duration]] + else: + break + ftime_sum = 0 + for fi_task, fi_duration in fin_output: + ftime_sum += fi_duration + + fin_output[-1][1] += truncated_act_dur - ftime_sum + output = fin_output + + task_decomp = output + ret = [] + for decomp_task, duration in task_decomp: + ret += [[f"{task_desc} ({decomp_task})", duration]] + output = ret + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/actions/wake_up.py b/metagpt/ext/stanford_town/actions/wake_up.py new file mode 100644 index 000000000..ea44cd3a4 --- /dev/null +++ b/metagpt/ext/stanford_town/actions/wake_up.py @@ -0,0 +1,42 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : wake_up + + +from metagpt.ext.stanford_town.actions.st_action import STAction +from metagpt.logs import logger + + +class WakeUp(STAction): + name: str = "WakeUp" + + def _func_validate(self, llm_resp: str, prompt: str = None) -> bool: + try: + self._func_cleanup(llm_resp, prompt="") + except Exception: + return False + return True + + def _func_cleanup(self, llm_resp: str, prompt: str) -> int: + cr = int(llm_resp.strip().lower().split("am")[0]) + return cr + + def _func_fail_default_resp(self) -> int: + fs = 8 + return fs + + async def run(self, role: "STRole"): + def create_prompt_input(role): + prompt_input = [ + role.scratch.get_str_iss(), + role.scratch.get_str_lifestyle(), + role.scratch.get_str_firstname(), + ] + return prompt_input + + prompt_input = create_prompt_input(role) + prompt = self.generate_prompt_with_tmpl_filename(prompt_input, "wake_up_hour_v1.txt") + self.fail_default_resp = self._func_fail_default_resp() + output = await self._run_gpt35_max_tokens(prompt, max_tokens=5) + logger.info(f"Role: {role.name} Action: {self.cls_name} output: {output}") + return output diff --git a/metagpt/ext/stanford_town/memory/__init__.py b/metagpt/ext/stanford_town/memory/__init__.py new file mode 100644 index 000000000..e69de29bb diff --git a/metagpt/ext/stanford_town/memory/agent_memory.py b/metagpt/ext/stanford_town/memory/agent_memory.py new file mode 100644 index 000000000..d212232f4 --- /dev/null +++ b/metagpt/ext/stanford_town/memory/agent_memory.py @@ -0,0 +1,378 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : BasicMemory,AgentMemory实现 + +from datetime import datetime +from pathlib import Path +from typing import Optional + +from pydantic import Field, field_serializer, model_validator + +from metagpt.logs import logger +from metagpt.memory.memory import Memory +from metagpt.schema import Message +from metagpt.utils.common import read_json_file, write_json_file + + +class BasicMemory(Message): + """ + BasicMemory继承于MG的Message类，其中content属性替代description属性 + Message类中对于Chat类型支持的非常好，对于Agent个体的Perceive,Reflection,Plan支持的并不多 + 在Type设计上，我们延续GA的三个种类，但是对于Chat种类的对话进行特别设计（具体怎么设计还没想好） + """ + + memory_id: Optional[str] = Field(default=None) # 记忆ID + memory_count: int = -1 # 第几个记忆，实际数值与Memory相等 + type_count: int = -1 # 第几种记忆，类型为整数 + memory_type: Optional[str] = Field(default=None) # 记忆类型，包含 event,thought,chat三种类型 + depth: int = -1 # 记忆深度，类型为整数 + created: Optional[datetime] = Field(default=None) # 创建时间 + expiration: Optional[datetime] = Field(default=None) # 记忆失效时间，默认为空（） + last_accessed: Optional[datetime] = Field(default=None) # 上一次调用的时间，初始化时候与self.created一致 + subject: Optional[str] = Field(default=None) # 主语 + predicate: Optional[str] = Field(default=None) # 谓语 + object: Optional[str] = Field(default=None) # 宾语 + + description: Optional[str] = Field(default=None) + embedding_key: Optional[str] = Field(default=None) # 内容与self.content一致 + poignancy: int = -1 # importance值 + keywords: list[str] = Field(default=[]) # keywords + filling: list = Field(default=[]) # 装的与之相关联的memory_id的列表 + + __hash__ = object.__hash__ # support hash in AgentMemory + + @model_validator(mode="before") + @classmethod + def check_values(cls, values): + if "created" in values: + values["last_accessed"] = values["created"] + if "content" in values: + values["description"] = values["content"] + if "filling" in values: + values["filling"] = values["filling"] or [] + return values + + @field_serializer("created", "expiration") + def transform_time_field(self, time_field: Optional[datetime]) -> str: + if time_field: + time_field = time_field.strftime("%Y-%m-%d %H:%M:%S") + return time_field + + def summary(self): + return self.subject, self.predicate, self.object + + def save_to_dict(self) -> dict: + """ + 将MemoryBasic类转化为字典，用于存储json文件 + 这里需要注意，cause_by跟GA不兼容，所以需要做一个格式转换 + """ + memory_dict = dict() + node_id = self.memory_id + basic_mem_obj = self.model_dump( + include=[ + "node_count", + "type_count", + "type", + "depth", + "created", + "expiration", + "subject", + "predicate", + "object", + "description", + "embedding_key", + "poignancy", + "keywords", + "filling", + "cause_by", + ] + ) + + memory_dict[node_id] = basic_mem_obj + return memory_dict + + +class AgentMemory(Memory): + """ + GA中主要存储三种JSON + 1. embedding.json (Dict embedding_key:embedding) + 2. Node.json (Dict Node_id:Node) + 3. kw_strength.json + """ + + storage: list[BasicMemory] = [] # 重写Storage，存储BasicMemory所有节点 + event_list: list[BasicMemory] = [] # 存储event记忆 + thought_list: list[BasicMemory] = [] # 存储thought记忆 + chat_list: list[BasicMemory] = [] # chat-related memory + + event_keywords: dict[str, list[BasicMemory]] = dict() # 存储keywords + thought_keywords: dict[str, list[BasicMemory]] = dict() + chat_keywords: dict[str, list[BasicMemory]] = dict() + + kw_strength_event: dict[str, int] = dict() + kw_strength_thought: dict[str, int] = dict() + + memory_saved: Optional[Path] = Field(default=None) + embeddings: dict[str, list[float]] = dict() + + def set_mem_path(self, memory_saved: Path): + self.memory_saved = memory_saved + self.load(memory_saved) + + def save(self, memory_saved: Path): + """ + 将MemoryBasic类存储为Nodes.json形式。复现GA中的Kw Strength.json形式 + 这里添加一个路径即可 + TODO 这里在存储时候进行倒序存储，之后需要验证（test_memory通过） + """ + memory_json = dict() + for i in range(len(self.storage)): + memory_node = self.storage[len(self.storage) - i - 1] + memory_node = memory_node.save_to_dict() + memory_json.update(memory_node) + write_json_file(memory_saved.joinpath("nodes.json"), memory_json) + write_json_file(memory_saved.joinpath("embeddings.json"), self.embeddings) + + strength_json = dict() + strength_json["kw_strength_event"] = self.kw_strength_event + strength_json["kw_strength_thought"] = self.kw_strength_thought + write_json_file(memory_saved.joinpath("kw_strength.json"), strength_json) + + def load(self, memory_saved: Path): + """ + 将GA的JSON解析，填充到AgentMemory类之中 + """ + self.embeddings = read_json_file(memory_saved.joinpath("embeddings.json")) + memory_load = read_json_file(memory_saved.joinpath("nodes.json")) + for count in range(len(memory_load.keys())): + node_id = f"node_{str(count + 1)}" + node_details = memory_load[node_id] + node_type = node_details["type"] + created = datetime.strptime(node_details["created"], "%Y-%m-%d %H:%M:%S") + expiration = None + if node_details["expiration"]: + expiration = datetime.strptime(node_details["expiration"], "%Y-%m-%d %H:%M:%S") + + s = node_details["subject"] + p = node_details["predicate"] + o = node_details["object"] + + description = node_details["description"] + embedding_pair = (node_details["embedding_key"], self.embeddings[node_details["embedding_key"]]) + poignancy = node_details["poignancy"] + keywords = set(node_details["keywords"]) + filling = node_details["filling"] + if node_type == "thought": + self.add_thought( + created, expiration, s, p, o, description, keywords, poignancy, embedding_pair, filling + ) + if node_type == "event": + self.add_event(created, expiration, s, p, o, description, keywords, poignancy, embedding_pair, filling) + if node_type == "chat": + self.add_chat(created, expiration, s, p, o, description, keywords, poignancy, embedding_pair, filling) + + strength_keywords_load = read_json_file(memory_saved.joinpath("kw_strength.json")) + if strength_keywords_load["kw_strength_event"]: + self.kw_strength_event = strength_keywords_load["kw_strength_event"] + if strength_keywords_load["kw_strength_thought"]: + self.kw_strength_thought = strength_keywords_load["kw_strength_thought"] + + def add(self, memory_basic: BasicMemory): + """ + Add a new message to storage, while updating the index + 重写add方法，修改原有的Message类为BasicMemory类，并添加不同的记忆类型添加方式 + """ + if memory_basic.memory_id in self.storage: + return + self.storage.append(memory_basic) + if memory_basic.memory_type == "chat": + self.chat_list[0:0] = [memory_basic] + return + if memory_basic.memory_type == "thought": + self.thought_list[0:0] = [memory_basic] + return + if memory_basic.memory_type == "event": + self.event_list[0:0] = [memory_basic] + return + + def add_chat( + self, created, expiration, s, p, o, content, keywords, poignancy, embedding_pair, filling, cause_by="" + ): + """ + 调用add方法，初始化chat，在创建的时候就需要调用embedding函数 + """ + memory_count = len(self.storage) + 1 + type_count = len(self.thought_list) + 1 + memory_type = "chat" + memory_id = f"node_{str(memory_count)}" + depth = 1 + + memory_node = BasicMemory( + memory_id=memory_id, + memory_count=memory_count, + type_count=type_count, + memory_type=memory_type, + depth=depth, + created=created, + expiration=expiration, + subject=s, + predicate=p, + object=o, + description=content, + embedding_key=embedding_pair[0], + poignancy=poignancy, + keywords=keywords, + filling=filling, + cause_by=cause_by, + ) + + keywords = [i.lower() for i in keywords] + for kw in keywords: + if kw in self.chat_keywords: + self.chat_keywords[kw][0:0] = [memory_node] + else: + self.chat_keywords[kw] = [memory_node] + + self.add(memory_node) + + self.embeddings[embedding_pair[0]] = embedding_pair[1] + return memory_node + + def add_thought(self, created, expiration, s, p, o, content, keywords, poignancy, embedding_pair, filling): + """ + 调用add方法，初始化thought + """ + memory_count = len(self.storage) + 1 + type_count = len(self.thought_list) + 1 + memory_type = "thought" + memory_id = f"node_{str(memory_count)}" + depth = 1 + + try: + if filling: + depth_list = [memory_node.depth for memory_node in self.storage if memory_node.memory_id in filling] + depth += max(depth_list) + except Exception as exp: + logger.warning(f"filling init occur {exp}") + pass + + memory_node = BasicMemory( + memory_id=memory_id, + memory_count=memory_count, + type_count=type_count, + memory_type=memory_type, + depth=depth, + created=created, + expiration=expiration, + subject=s, + predicate=p, + object=o, + description=content, + embedding_key=embedding_pair[0], + poignancy=poignancy, + keywords=keywords, + filling=filling, + ) + + keywords = [i.lower() for i in keywords] + for kw in keywords: + if kw in self.thought_keywords: + self.thought_keywords[kw][0:0] = [memory_node] + else: + self.thought_keywords[kw] = [memory_node] + + self.add(memory_node) + + if f"{p} {o}" != "is idle": + for kw in keywords: + if kw in self.kw_strength_thought: + self.kw_strength_thought[kw] += 1 + else: + self.kw_strength_thought[kw] = 1 + + self.embeddings[embedding_pair[0]] = embedding_pair[1] + return memory_node + + def add_event(self, created, expiration, s, p, o, content, keywords, poignancy, embedding_pair, filling): + """ + 调用add方法，初始化event + """ + memory_count = len(self.storage) + 1 + type_count = len(self.event_list) + 1 + memory_type = "event" + memory_id = f"node_{str(memory_count)}" + depth = 0 + + if "(" in content: + content = " ".join(content.split()[:3]) + " " + content.split("(")[-1][:-1] + + memory_node = BasicMemory( + memory_id=memory_id, + memory_count=memory_count, + type_count=type_count, + memory_type=memory_type, + depth=depth, + created=created, + expiration=expiration, + subject=s, + predicate=p, + object=o, + description=content, + embedding_key=embedding_pair[0], + poignancy=poignancy, + keywords=keywords, + filling=filling, + ) + + keywords = [i.lower() for i in keywords] + for kw in keywords: + if kw in self.event_keywords: + self.event_keywords[kw][0:0] = [memory_node] + else: + self.event_keywords[kw] = [memory_node] + + self.add(memory_node) + + if f"{p} {o}" != "is idle": + for kw in keywords: + if kw in self.kw_strength_event: + self.kw_strength_event[kw] += 1 + else: + self.kw_strength_event[kw] = 1 + + self.embeddings[embedding_pair[0]] = embedding_pair[1] + return memory_node + + def get_summarized_latest_events(self, retention): + ret_set = set() + for e_node in self.event_list[:retention]: + ret_set.add(e_node.summary()) + return ret_set + + def get_last_chat(self, target_role_name: str): + if target_role_name.lower() in self.chat_keywords: + return self.chat_keywords[target_role_name.lower()][0] + else: + return False + + def retrieve_relevant_thoughts(self, s_content: str, p_content: str, o_content: str) -> set: + contents = [s_content, p_content, o_content] + + ret = [] + for i in contents: + if i in self.thought_keywords: + ret += self.thought_keywords[i.lower()] + + ret = set(ret) + return ret + + def retrieve_relevant_events(self, s_content: str, p_content: str, o_content: str) -> set: + contents = [s_content, p_content, o_content] + + ret = [] + for i in contents: + if i in self.event_keywords: + ret += self.event_keywords[i] + + ret = set(ret) + return ret diff --git a/metagpt/ext/stanford_town/memory/retrieve.py b/metagpt/ext/stanford_town/memory/retrieve.py new file mode 100644 index 000000000..c4b32f965 --- /dev/null +++ b/metagpt/ext/stanford_town/memory/retrieve.py @@ -0,0 +1,180 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : Retrieve函数实现 + +import datetime + +from numpy import dot +from numpy.linalg import norm + +from metagpt.ext.stanford_town.memory.agent_memory import BasicMemory +from metagpt.ext.stanford_town.utils.utils import get_embedding + + +def agent_retrieve( + agent_memory, + curr_time: datetime.datetime, + memory_forget: float, + query: str, + nodes: list[BasicMemory], + topk: int = 4, +) -> list[BasicMemory]: + """ + Retrieve需要集合Role使用,原因在于Role才具有AgentMemory,scratch + 逻辑:Role调用该函数,self.rc.AgentMemory,self.rc.scratch.curr_time,self.rc.scratch.memory_forget + 输入希望查询的内容与希望回顾的条数,返回TopK条高分记忆，即List[BasicMemory] + + Score_lists示例 + { + "memory": memories[i], BasicMemory类 + "importance": memories[i].poignancy + "recency": 衰减因子计算结果 + "relevance": 搜索结果 + } + """ + memories = nodes + agent_memory_embedding = agent_memory.embeddings + memories = sorted(memories, key=lambda memory_node: memory_node.last_accessed, reverse=True) + + score_list = [] + score_list = extract_importance(memories, score_list) + score_list = extract_recency(curr_time, memory_forget, score_list) + score_list = extract_relevance(agent_memory_embedding, query, score_list) + score_list = normalize_score_floats(score_list, 0, 1) + + total_dict = {} + gw = [1, 1, 1] # 三个因素的权重,重要性,近因性,相关性, + for i in range(len(score_list)): + total_score = ( + score_list[i]["importance"] * gw[0] + score_list[i]["recency"] * gw[1] + score_list[i]["relevance"] * gw[2] + ) + total_dict[score_list[i]["memory"].memory_id] = total_score + + result = top_highest_x_values(total_dict, topk) + + return result # 返回的是一个BasicMemory列表 + + +def new_agent_retrieve(role, focus_points: list, n_count=30) -> dict: + """ + 输入为role，关注点列表,返回记忆数量 + 输出为字典，键为focus_point，值为对应的记忆列表 + """ + retrieved = dict() + for focal_pt in focus_points: + nodes = [ + [i.last_accessed, i] + for i in role.memory.event_list + role.memory.thought_list + if "idle" not in i.embedding_key + ] + nodes = sorted(nodes, key=lambda x: x[0]) + nodes = [i for created, i in nodes] + results = agent_retrieve( + role.memory, role.scratch.curr_time, role.scratch.recency_decay, focal_pt, nodes, n_count + ) + final_result = [] + for n in results: + for i in role.memory.storage: + if i.memory_id == n: + i.last_accessed = role.scratch.curr_time + final_result.append(i) + + retrieved[focal_pt] = final_result + + return retrieved + + +def top_highest_x_values(d, x): + """ + 输入字典，Topx + 返回以字典值排序，字典键组成的List[BasicMemory] + """ + top_v = [item[0] for item in sorted(d.items(), key=lambda item: item[1], reverse=True)[:x]] + return top_v + + +def extract_importance(memories, score_list): + """ + 抽取重要性 + """ + for i in range(len(memories)): + score = {"memory": memories[i], "importance": memories[i].poignancy} + score_list.append(score) + return score_list + + +def extract_relevance(agent_memory_embedding, query, score_list): + """ + 抽取相关性 + """ + query_embedding = get_embedding(query) + # 进行 + for i in range(len(score_list)): + node_embedding = agent_memory_embedding[score_list[i]["memory"].embedding_key] + result = cos_sim(node_embedding, query_embedding) + score_list[i]["relevance"] = result + + return score_list + + +def extract_recency(curr_time, memory_forget, score_list): + """ + 抽取近因性，目前使用的现实世界过一天走一个衰减因子 + """ + for i in range(len(score_list)): + day_count = (curr_time - score_list[i]["memory"].created).days + score_list[i]["recency"] = memory_forget**day_count + return score_list + + +def cos_sim(a, b): + """ + 计算余弦相似度 + """ + return dot(a, b) / (norm(a) * norm(b)) + + +def normalize_list_floats(single_list, target_min, target_max): + """ + 单个列表归一化 + """ + if len(single_list) == 0: + return [] + + min_val = min(single_list) + max_val = max(single_list) + range_val = max_val - min_val + + if range_val == 0: + for i in range(len(single_list)): + single_list[i] = (target_max - target_min) / 2 + else: + for i in range(len(single_list)): + single_list[i] = (single_list[i] - min_val) * (target_max - target_min) / range_val + target_min + return single_list + + +def normalize_score_floats(score_list, target_min, target_max): + """ + 整体归一化 + """ + importance_list = [] + relevance_list = [] + recency_list = [] + + for i in range(len(score_list)): + importance_list.append(score_list[i]["importance"]) + relevance_list.append(score_list[i]["relevance"]) + recency_list.append(score_list[i]["recency"]) + + # 进行归一化操作 + importance_list = normalize_list_floats(importance_list, target_min, target_max) + relevance_list = normalize_list_floats(relevance_list, target_min, target_max) + recency_list = normalize_list_floats(recency_list, target_min, target_max) + + for i in range(len(score_list)): + score_list[i]["importance"] = importance_list[i] + score_list[i]["relevance"] = relevance_list[i] + score_list[i]["recency"] = recency_list[i] + + return score_list diff --git a/metagpt/ext/stanford_town/memory/scratch.py b/metagpt/ext/stanford_town/memory/scratch.py new file mode 100644 index 000000000..b4036f839 --- /dev/null +++ b/metagpt/ext/stanford_town/memory/scratch.py @@ -0,0 +1,383 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : Scratch类实现（角色信息类） + +from datetime import datetime, timedelta +from pathlib import Path +from typing import Optional, Union + +from pydantic import BaseModel, Field, field_serializer, field_validator + +from metagpt.utils.common import read_json_file, write_json_file + + +class Scratch(BaseModel): + # 类别1:人物超参 + vision_r: int = 4 + att_bandwidth: int = 3 + retention: int = 5 + + # 类别2:世界信息 + curr_time: Optional[datetime] = Field(default=None) + curr_tile: Optional[list[int]] = Field(default=None) + daily_plan_req: Optional[str] = Field(default=None) + + # 类别3:人物角色的核心身份 + name: Optional[str] = Field(default=None) + first_name: Optional[str] = Field(default=None) + last_name: Optional[str] = Field(default=None) + age: Optional[int] = Field(default=None) + innate: Optional[str] = Field(default=None) # L0 permanent core traits. + learned: Optional[str] = Field(default=None) # L1 stable traits. + currently: Optional[str] = Field(default=None) # L2 external implementation. + lifestyle: Optional[str] = Field(default=None) + living_area: Optional[str] = Field(default=None) + + # 类别4:旧反思变量 + concept_forget: int = 100 + daily_reflection_time: int = 60 * 3 + daily_reflection_size: int = 5 + overlap_reflect_th: int = 2 + kw_strg_event_reflect_th: int = 4 + kw_strg_thought_reflect_th: int = 4 + + # 类别5:新反思变量 + recency_w: int = 1 + relevance_w: int = 1 + importance_w: int = 1 + recency_decay: float = 0.99 + importance_trigger_max: int = 150 + importance_trigger_curr: int = 150 + importance_ele_n: int = 0 + thought_count: int = 5 + + # 类别6:个人计划 + daily_req: list[str] = Field(default=[]) + f_daily_schedule: list[list[Union[int, str]]] = Field(default=[]) + f_daily_schedule_hourly_org: list[list[Union[int, str]]] = Field(default=[]) + + # 类别7:当前动作 + act_address: Optional[str] = Field(default=None) + act_start_time: Optional[datetime] = Field(default=None) + act_duration: Optional[int] = Field(default=None) + act_description: Optional[str] = Field(default=None) + act_pronunciatio: Optional[str] = Field(default=None) + act_event: list[Optional[str]] = [None, None, None] + + act_obj_description: Optional[str] = Field(default=None) + act_obj_pronunciatio: Optional[str] = Field(default=None) + act_obj_event: list[Optional[str]] = [None, None, None] + + chatting_with: Optional[str] = Field(default=None) + chat: Optional[str] = Field(default=None) + chatting_with_buffer: dict = dict() + chatting_end_time: Optional[datetime] = Field(default=None) + + act_path_set: bool = False + planned_path: list[list[int]] = Field(default=[]) + + @field_validator("curr_time", "act_start_time", "chatting_end_time", mode="before") + @classmethod + def check_time_filed(cls, time_filed): + val = datetime.strptime(time_filed, "%B %d, %Y, %H:%M:%S") if time_filed else None + return val + + @field_serializer("curr_time", "act_start_time", "chatting_end_time") + def transform_time_field(self, time_filed: Optional[datetime]) -> str: + if time_filed: + time_filed = time_filed.strftime("%B %d, %Y, %H:%M:%S") + return time_filed + + @classmethod + def init_scratch_from_path(cls, f_saved: Path): + scratch_load = read_json_file(f_saved) + scratch = Scratch(**scratch_load) + return scratch + + def save(self, out_json: Path): + """ + Save persona's scratch. + + INPUT: + out_json: The file where we wil be saving our persona's state. + OUTPUT: + None + """ + scratch = self.model_dump() + write_json_file(out_json, scratch, encoding="utf-8") + + def get_f_daily_schedule_index(self, advance=0): + """ + We get the current index of self.f_daily_schedule. + + Recall that self.f_daily_schedule stores the decomposed action sequences + up until now, and the hourly sequences of the future action for the rest + of today. Given that self.f_daily_schedule is a list of list where the + inner list is composed of [task, duration], we continue to add up the + duration until we reach "if elapsed > today_min_elapsed" condition. The + index where we stop is the index we will return. + + INPUT + advance: Integer value of the number minutes we want to look into the + future. This allows us to get the index of a future timeframe. + OUTPUT + an integer value for the current index of f_daily_schedule. + """ + # We first calculate teh number of minutes elapsed today. + today_min_elapsed = 0 + today_min_elapsed += self.curr_time.hour * 60 + today_min_elapsed += self.curr_time.minute + today_min_elapsed += advance + + x = 0 + for task, duration in self.f_daily_schedule: + x += duration + x = 0 + for task, duration in self.f_daily_schedule_hourly_org: + x += duration + + # We then calculate the current index based on that. + curr_index = 0 + elapsed = 0 + for task, duration in self.f_daily_schedule: + elapsed += duration + if elapsed > today_min_elapsed: + return curr_index + curr_index += 1 + + return curr_index + + def get_f_daily_schedule_hourly_org_index(self, advance=0): + """ + We get the current index of self.f_daily_schedule_hourly_org. + It is otherwise the same as get_f_daily_schedule_index. + + INPUT + advance: Integer value of the number minutes we want to look into the + future. This allows us to get the index of a future timeframe. + OUTPUT + an integer value for the current index of f_daily_schedule. + """ + # We first calculate teh number of minutes elapsed today. + today_min_elapsed = 0 + today_min_elapsed += self.curr_time.hour * 60 + today_min_elapsed += self.curr_time.minute + today_min_elapsed += advance + # We then calculate the current index based on that. + curr_index = 0 + elapsed = 0 + for task, duration in self.f_daily_schedule_hourly_org: + elapsed += duration + if elapsed > today_min_elapsed: + return curr_index + curr_index += 1 + return curr_index + + def get_str_iss(self): + """ + ISS stands for "identity stable set." This describes the commonset summary + of this persona -- basically, the bare minimum description of the persona + that gets used in almost all prompts that need to call on the persona. + + INPUT + None + OUTPUT + the identity stable set summary of the persona in a string form. + EXAMPLE STR OUTPUT + "Name: Dolores Heitmiller + Age: 28 + Innate traits: hard-edged, independent, loyal + Learned traits: Dolores is a painter who wants live quietly and paint + while enjoying her everyday life. + Currently: Dolores is preparing for her first solo show. She mostly + works from home. + Lifestyle: Dolores goes to bed around 11pm, sleeps for 7 hours, eats + dinner around 6pm. + Daily plan requirement: Dolores is planning to stay at home all day and + never go out." + """ + commonset = "" + commonset += f"Name: {self.name}\n" + commonset += f"Age: {self.age}\n" + commonset += f"Innate traits: {self.innate}\n" + commonset += f"Learned traits: {self.learned}\n" + commonset += f"Currently: {self.currently}\n" + commonset += f"Lifestyle: {self.lifestyle}\n" + commonset += f"Daily plan requirement: {self.daily_plan_req}\n" + commonset += f"Current Date: {self.curr_time.strftime('%A %B %d') if self.curr_time else ''}\n" + return commonset + + def get_str_name(self): + return self.name + + def get_str_firstname(self): + return self.first_name + + def get_str_lastname(self): + return self.last_name + + def get_str_age(self): + return str(self.age) + + def get_str_innate(self): + return self.innate + + def get_str_learned(self): + return self.learned + + def get_str_currently(self): + return self.currently + + def get_str_lifestyle(self): + return self.lifestyle + + def get_str_daily_plan_req(self): + return self.daily_plan_req + + def get_str_curr_date_str(self): + return self.curr_time.strftime("%A %B %d") + + def get_curr_event(self): + if not self.act_address: + return self.name, None, None + else: + return self.act_event + + def get_curr_event_and_desc(self): + if not self.act_address: + return self.name, None, None, None + else: + return self.act_event[0], self.act_event[1], self.act_event[2], self.act_description + + def get_curr_obj_event_and_desc(self): + if not self.act_address: + return "", None, None, None + else: + return self.act_address, self.act_obj_event[1], self.act_obj_event[2], self.act_obj_description + + def add_new_action( + self, + action_address, + action_duration, + action_description, + action_pronunciatio, + action_event, + chatting_with, + chat, + chatting_with_buffer, + chatting_end_time, + act_obj_description, + act_obj_pronunciatio, + act_obj_event, + act_start_time=None, + ): + self.act_address = action_address + self.act_duration = action_duration + self.act_description = action_description + self.act_pronunciatio = action_pronunciatio + self.act_event = action_event + + self.chatting_with = chatting_with + self.chat = chat + if chatting_with_buffer: + self.chatting_with_buffer.update(chatting_with_buffer) + self.chatting_end_time = chatting_end_time + + self.act_obj_description = act_obj_description + self.act_obj_pronunciatio = act_obj_pronunciatio + self.act_obj_event = act_obj_event + + self.act_start_time = self.curr_time + + self.act_path_set = False + + def act_time_str(self): + """ + Returns a string output of the current time. + + INPUT + None + OUTPUT + A string output of the current time. + EXAMPLE STR OUTPUT + "14:05 P.M." + """ + return self.act_start_time.strftime("%H:%M %p") + + def act_check_finished(self): + """ + Checks whether the self.Action instance has finished. + + INPUT + curr_datetime: Current time. If current time is later than the action's + start time + its duration, then the action has finished. + OUTPUT + Boolean [True]: Action has finished. + Boolean [False]: Action has not finished and is still ongoing. + """ + if not self.act_address: + return True + + if self.chatting_with: + end_time = self.chatting_end_time + else: + x = self.act_start_time + if x.second != 0: + x = x.replace(second=0) + x = x + timedelta(minutes=1) + end_time = x + timedelta(minutes=self.act_duration) + + if end_time.strftime("%H:%M:%S") == self.curr_time.strftime("%H:%M:%S"): + return True + return False + + def act_summarize(self): + """ + Summarize the current action as a dictionary. + + INPUT + None + OUTPUT + ret: A human readable summary of the action. + """ + exp = dict() + exp["persona"] = self.name + exp["address"] = self.act_address + exp["start_datetime"] = self.act_start_time + exp["duration"] = self.act_duration + exp["description"] = self.act_description + exp["pronunciatio"] = self.act_pronunciatio + return exp + + def act_summary_str(self): + """ + Returns a string summary of the current action. Meant to be + human-readable. + + INPUT + None + OUTPUT + ret: A human readable summary of the action. + """ + start_datetime_str = self.act_start_time.strftime("%A %B %d -- %H:%M %p") + ret = f"[{start_datetime_str}]\n" + ret += f"Activity: {self.name} is {self.act_description}\n" + ret += f"Address: {self.act_address}\n" + ret += f"Duration in minutes (e.g., x min): {str(self.act_duration)} min\n" + return ret + + def get_daily_schedule(self, daily_schedule: list[list[str]]): + ret = "" + curr_min_sum = 0 + for row in daily_schedule: + curr_min_sum += row[1] + hour = int(curr_min_sum / 60) + minute = curr_min_sum % 60 + ret += f"{hour:02}:{minute:02} || {row[0]}\n" + return ret + + def get_str_daily_schedule_summary(self): + return self.get_daily_schedule(self.f_daily_schedule) + + def get_str_daily_schedule_hourly_org_summary(self): + return self.get_daily_schedule(self.f_daily_schedule_hourly_org) diff --git a/metagpt/ext/stanford_town/memory/spatial_memory.py b/metagpt/ext/stanford_town/memory/spatial_memory.py new file mode 100644 index 000000000..71b856907 --- /dev/null +++ b/metagpt/ext/stanford_town/memory/spatial_memory.py @@ -0,0 +1,116 @@ +""" +Author: Joon Sung Park (joonspk@stanford.edu) + +File: spatial_memory.py +Description: Defines the MemoryTree class that serves as the agents' spatial +memory that aids in grounding their behavior in the game world. +""" +from pathlib import Path + +from pydantic import BaseModel, Field + +from metagpt.logs import logger +from metagpt.utils.common import read_json_file, write_json_file + + +class MemoryTree(BaseModel): + tree: dict = Field(default=dict) + + def set_mem_path(self, f_saved: Path): + self.tree = read_json_file(f_saved) + + def print_tree(self) -> None: + def _print_tree(tree, depth): + dash = " >" * depth + if isinstance(tree, list): + if tree: + logger.info(f"{dash} {tree}") + return + + for key, val in tree.items(): + if key: + logger.info(f"{dash} {tree}") + _print_tree(val, depth + 1) + + _print_tree(self.tree, 0) + + def save(self, out_json: Path) -> None: + write_json_file(out_json, self.tree) + + def get_str_accessible_sectors(self, curr_world: str) -> str: + """ + Returns a summary string of all the arenas that the persona can access + within the current sector. + + Note that there are places a given persona cannot enter. This information + is provided in the persona sheet. We account for this in this function. + + INPUT + None + OUTPUT + A summary string of all the arenas that the persona can access. + EXAMPLE STR OUTPUT + "bedroom, kitchen, dining room, office, bathroom" + """ + x = ", ".join(list(self.tree[curr_world].keys())) + return x + + def get_str_accessible_sector_arenas(self, sector: str) -> str: + """ + Returns a summary string of all the arenas that the persona can access + within the current sector. + + Note that there are places a given persona cannot enter. This information + is provided in the persona sheet. We account for this in this function. + + INPUT + None + OUTPUT + A summary string of all the arenas that the persona can access. + EXAMPLE STR OUTPUT + "bedroom, kitchen, dining room, office, bathroom" + """ + curr_world, curr_sector = sector.split(":") + if not curr_sector: + return "" + x = ", ".join(list(self.tree[curr_world][curr_sector].keys())) + return x + + def get_str_accessible_arena_game_objects(self, arena: str) -> str: + """ + Get a str list of all accessible game objects that are in the arena. If + temp_address is specified, we return the objects that are available in + that arena, and if not, we return the objects that are in the arena our + persona is currently in. + + INPUT + temp_address: optional arena address + OUTPUT + str list of all accessible game objects in the gmae arena. + EXAMPLE STR OUTPUT + "phone, charger, bed, nightstand" + """ + curr_world, curr_sector, curr_arena = arena.split(":") + + if not curr_arena: + return "" + + try: + x = ", ".join(list(self.tree[curr_world][curr_sector][curr_arena])) + except Exception: + x = ", ".join(list(self.tree[curr_world][curr_sector][curr_arena.lower()])) + return x + + def add_tile_info(self, tile_info: dict) -> None: + if tile_info["world"]: + if tile_info["world"] not in self.tree: + self.tree[tile_info["world"]] = {} + if tile_info["sector"]: + if tile_info["sector"] not in self.tree[tile_info["world"]]: + self.tree[tile_info["world"]][tile_info["sector"]] = {} + if tile_info["arena"]: + if tile_info["arena"] not in self.tree[tile_info["world"]][tile_info["sector"]]: + self.tree[tile_info["world"]][tile_info["sector"]][tile_info["arena"]] = [] + if tile_info["game_object"]: + if tile_info["game_object"] not in self.tree[tile_info["world"]][tile_info["sector"]][tile_info["arena"]]: + self.tree[tile_info["world"]][tile_info["sector"]][tile_info["arena"]] += [tile_info["game_object"]] diff --git a/metagpt/ext/stanford_town/plan/__init__.py b/metagpt/ext/stanford_town/plan/__init__.py new file mode 100644 index 000000000..2bcf8efd0 --- /dev/null +++ b/metagpt/ext/stanford_town/plan/__init__.py @@ -0,0 +1,3 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : diff --git a/metagpt/ext/stanford_town/plan/converse.py b/metagpt/ext/stanford_town/plan/converse.py new file mode 100644 index 000000000..8eefbc9b4 --- /dev/null +++ b/metagpt/ext/stanford_town/plan/converse.py @@ -0,0 +1,93 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : conversation between two agents + +from typing import Tuple + +from metagpt.ext.stanford_town.actions.agent_chat_sum_rel import AgentChatSumRel +from metagpt.ext.stanford_town.actions.gen_iter_chat_utt import GenIterChatUTT +from metagpt.ext.stanford_town.memory.retrieve import new_agent_retrieve +from metagpt.logs import logger + + +async def agent_conversation(init_role: "STRole", target_role: "STRole", conv_rounds: int = 8) -> list[list[str]]: + curr_chat = [] + logger.info(f"Role: {init_role.name} starts a conversation with Role: {target_role.name}") + + for idx in range(conv_rounds): + logger.info(f"Conv round: {idx} between {init_role.name} and {target_role.name}") + scratch = init_role.rc.scratch + target_scratch = target_role.rc.scratch + + focal_points = [f"{target_scratch.name}"] + retrieved = new_agent_retrieve(init_role, focal_points, 50) + relationship = await generate_summarize_agent_relationship(init_role, target_role, retrieved) + logger.info(f"The relationship between {init_role.name} and {target_role.name}: {relationship}") + last_chat = "" + for i in curr_chat[-4:]: + last_chat += ": ".join(i) + "\n" + if last_chat: + focal_points = [f"{relationship}", f"{target_scratch.name} is {target_scratch.act_description}", last_chat] + else: + focal_points = [f"{relationship}", f"{target_scratch.name} is {target_scratch.act_description}"] + retrieved = new_agent_retrieve(init_role, focal_points, 15) + utt, end = await generate_one_utterance(init_role, target_role, retrieved, curr_chat) + + curr_chat += [[scratch.name, utt]] + if end: + break + + focal_points = [f"{scratch.name}"] + retrieved = new_agent_retrieve(target_role, focal_points, 50) + relationship = await generate_summarize_agent_relationship(target_role, init_role, retrieved) + logger.info(f"The relationship between {target_role.name} and {init_role.name}: {relationship}") + last_chat = "" + for i in curr_chat[-4:]: + last_chat += ": ".join(i) + "\n" + if last_chat: + focal_points = [f"{relationship}", f"{scratch.name} is {scratch.act_description}", last_chat] + else: + focal_points = [f"{relationship}", f"{scratch.name} is {scratch.act_description}"] + retrieved = new_agent_retrieve(target_role, focal_points, 15) + utt, end = await generate_one_utterance(target_role, init_role, retrieved, curr_chat) + + curr_chat += [[target_scratch.name, utt]] + if end: + break + + logger.warning(f"Conversations between {target_role.name} and {init_role.name}:") + for row in curr_chat: + logger.info(row) + + return curr_chat + + +async def generate_summarize_agent_relationship(init_role: "STRole", target_role: "STRole", retrieved: dict) -> str: + all_embedding_keys = list() + for key, val in retrieved.items(): + for i in val: + all_embedding_keys += [i.embedding_key] + all_embedding_key_str = "" + for i in all_embedding_keys: + all_embedding_key_str += f"{i}\n" + + summarized_relationship = await AgentChatSumRel().run(init_role, target_role, all_embedding_key_str) + return summarized_relationship + + +async def generate_one_utterance(init_role, target_role, retrieved: dict, curr_chat: list) -> Tuple[str, str]: + # Chat version optimized for speed via batch generation + scratch = init_role.rc.scratch + target_scratch = target_role.rc.scratch + curr_context = ( + f"{scratch.name} " + + f"was {scratch.act_description} " + + f"when {scratch.name} " + + f"saw {target_scratch.name} " + + f"in the middle of {target_scratch.act_description}.\n" + ) + curr_context += f"{scratch.name} " + "is initiating a conversation with " + f"{target_scratch.name}." + + x = await GenIterChatUTT().run(init_role, target_role, retrieved, curr_context, curr_chat) + + return x["utterance"], x["end"] diff --git a/metagpt/ext/stanford_town/plan/st_plan.py b/metagpt/ext/stanford_town/plan/st_plan.py new file mode 100644 index 000000000..f63052fc5 --- /dev/null +++ b/metagpt/ext/stanford_town/plan/st_plan.py @@ -0,0 +1,706 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# @Desc : st' planning execution + +import datetime +import math +import random +from typing import Tuple, Union + +from metagpt.ext.stanford_town.actions.decide_to_talk import DecideToTalk +from metagpt.ext.stanford_town.actions.gen_action_details import GenActionDetails +from metagpt.ext.stanford_town.actions.gen_daily_schedule import GenDailySchedule +from metagpt.ext.stanford_town.actions.gen_hourly_schedule import GenHourlySchedule +from metagpt.ext.stanford_town.actions.new_decomp_schedule import NewDecompSchedule +from metagpt.ext.stanford_town.actions.summarize_conv import SummarizeConv +from metagpt.ext.stanford_town.actions.task_decomp import TaskDecomp +from metagpt.ext.stanford_town.actions.wake_up import WakeUp +from metagpt.ext.stanford_town.memory.retrieve import new_agent_retrieve +from metagpt.ext.stanford_town.plan.converse import agent_conversation +from metagpt.ext.stanford_town.utils.utils import get_embedding +from metagpt.llm import LLM +from metagpt.logs import logger + + +async def plan(role: "STRole", roles: dict["STRole"], new_day: bool, retrieved: dict) -> str: + # PART 1: Generate the hourly schedule. + if new_day: + await _long_term_planning(role, new_day) + + # PART 2: If the current action has expired, we want to create a new plan. + act_check_finished = role.scratch.act_check_finished() + logger.info(f"Role: {role.name} act_check_finished is {act_check_finished}") + if act_check_finished: + await _determine_action(role) + + # PART 3: If you perceived an event that needs to be responded to (saw + # another role), and retrieved relevant information. + # Step 1: Retrieved may have multiple events represented in it. The first + # job here is to determine which of the events we want to focus + # on for the role. + # takes the form of a dictionary like this: + # dictionary {["curr_event"] = , + # ["events"] = [, ...], + # ["thoughts"] = [, ...]} + focused_event = False + if retrieved.keys(): + focused_event = _choose_retrieved(role.name, retrieved) + + # Step 2: Once we choose an event, we need to determine whether the + # role will take any actions for the perceived event. There are + # three possible modes of reaction returned by _should_react. + # a) "chat with {target_role.name}" + # b) "react" + # c) False + logger.info(f"Role: {role.name} focused_event: {focused_event}") + if focused_event: + reaction_mode = await _should_react(role, focused_event, roles) + logger.info(f"Role: {role.name} reaction_mode: {reaction_mode}") + if reaction_mode: + # If we do want to chat, then we generate conversation + if reaction_mode[:9] == "chat with": + await _chat_react(role, reaction_mode, roles) + elif reaction_mode[:4] == "wait": + await _wait_react(role, reaction_mode) + + # Step 3: Chat-related state clean up. + # If the persona is not chatting with anyone, we clean up any of the + # chat-related states here. + if role.rc.scratch.act_event[1] != "chat with": + role.rc.scratch.chatting_with = None + role.rc.scratch.chat = None + role.rc.scratch.chatting_end_time = None + # We want to make sure that the persona does not keep conversing with each + # other in an infinite loop. So, chatting_with_buffer maintains a form of + # buffer that makes the persona wait from talking to the same target + # immediately after chatting once. We keep track of the buffer value here. + curr_persona_chat_buffer = role.rc.scratch.chatting_with_buffer + for persona_name, buffer_count in curr_persona_chat_buffer.items(): + if persona_name != role.rc.scratch.chatting_with: + role.rc.scratch.chatting_with_buffer[persona_name] -= 1 + + return role.rc.scratch.act_address + + +def _choose_retrieved(role_name: str, retrieved: dict) -> Union[None, dict]: + """ + Retrieved elements have multiple core "curr_events". We need to choose one + event to which we are going to react to. We pick that event here. + Args: + role_name: Current role instance's name whose action we are determining. + retrieved: A dictionary of that were retrieved from the + the role's associative memory. This dictionary takes the + following form: + dictionary[event.description] = + {["curr_event"] = , + ["events"] = [, ...], + ["thoughts"] = [, ...] } + """ + # Once we are done with the reflection, we might want to build a more + # complex structure here. + + # We do not want to take self events... for now + copy_retrieved = retrieved.copy() + for event_desc, rel_ctx in copy_retrieved.items(): + curr_event = rel_ctx["curr_event"] + if curr_event.subject == role_name: + del retrieved[event_desc] + + # Always choose role first. + priority = [] + for event_desc, rel_ctx in retrieved.items(): + curr_event = rel_ctx["curr_event"] + if ":" not in curr_event.subject and curr_event.subject != role_name: + priority += [rel_ctx] + if priority: + return random.choice(priority) + + # Skip idle. + for event_desc, rel_ctx in retrieved.items(): + if "is idle" not in event_desc: + priority += [rel_ctx] + if priority: + return random.choice(priority) + return None + + +async def _should_react(role: "STRole", retrieved: dict, roles: dict): + """ + Determines what form of reaction the role should exihibit given the + retrieved values. + INPUT + role: Current <"STRole"> instance whose action we are determining. + retrieved: A dictionary of that were retrieved from the + the role's associative memory. This dictionary takes the + following form: + dictionary[event.description] = + {["curr_event"] = , + ["events"] = [, ...], + ["thoughts"] = [, ...] } + roles: A dictionary that contains all role names as keys, and the + <"STRole"> instance as values. + """ + + async def lets_talk(init_role: "STRole", target_role: "STRole", retrieved: dict): + if init_role.name == target_role.name: + logger.info(f"Role: {role.name} _should_react lets_talk meet same role, return False") + return False + + scratch = init_role.rc.scratch + target_scratch = target_role.rc.scratch + if ( + not target_scratch.act_address + or not target_scratch.act_description + or not scratch.act_address + or not scratch.act_description + ): + return False + + if "sleeping" in target_scratch.act_description or "sleeping" in scratch.act_description: + return False + + if scratch.curr_time.hour == 23: + return False + + if "" in target_scratch.act_address: + return False + + if target_scratch.chatting_with or scratch.chatting_with: + return False + + if target_role.name in scratch.chatting_with_buffer: + if scratch.chatting_with_buffer[target_role.name] > 0: + return False + + if await DecideToTalk().run(init_role, target_role, retrieved): + return True + + return False + + async def lets_react(init_role: "STRole", target_role: "STRole", retrieved: dict): + if init_role.name == target_role.name: + logger.info(f"Role: {role.name} _should_react lets_react meet same role, return False") + return False + + scratch = init_role.rc.scratch + target_scratch = target_role.rc.scratch + if ( + not target_scratch.act_address + or not target_scratch.act_description + or not scratch.act_address + or not scratch.act_description + ): + return False + + if "sleeping" in target_scratch.act_description or "sleeping" in scratch.act_description: + return False + + # return False + if scratch.curr_time.hour == 23: + return False + + if "waiting" in target_scratch.act_description: + return False + if scratch.planned_path == []: + return False + + if scratch.act_address != target_scratch.act_address: + return False + + react_mode = await DecideToTalk().run(init_role, target_role, retrieved) + + if react_mode == "1": + wait_until = ( + target_scratch.act_start_time + datetime.timedelta(minutes=target_scratch.act_duration - 1) + ).strftime("%B %d, %Y, %H:%M:%S") + return f"wait: {wait_until}" + elif react_mode == "2": + return False + return "do other things" + else: + return False # "keep" + + # If the role is chatting right now, default to no reaction + scratch = role.rc.scratch + if scratch.chatting_with: + return False + if "" in scratch.act_address: + return False + + # Recall that retrieved takes the following form: + # dictionary {["curr_event"] = } + curr_event = retrieved["curr_event"] + logger.info(f"Role: {role.name} _should_react curr_event.subject: {curr_event.subject}") + + if ":" not in curr_event.subject: + # this is a role event. + if await lets_talk(role, roles[curr_event.subject], retrieved): + return f"chat with {curr_event.subject}" + react_mode = await lets_react(role, roles[curr_event.subject], retrieved) + return react_mode + return False + + +async def _chat_react(role: "STRole", reaction_mode: str, roles: dict["STRole"]): + # There are two roles -- the role who is initiating the conversation + # and the role who is the target. We get the role instances here. + init_role = role + target_role = roles[reaction_mode[9:].strip()] + + # Actually creating the conversation here. + convo, duration_min = await generate_convo(init_role, target_role) # 2222 + convo_summary = await generate_convo_summary(convo) + inserted_act = convo_summary + inserted_act_dur = duration_min + + act_start_time = target_role.rc.scratch.act_start_time + + curr_time = target_role.rc.scratch.curr_time + if curr_time.second != 0: + temp_curr_time = curr_time + datetime.timedelta(seconds=60 - curr_time.second) + chatting_end_time = temp_curr_time + datetime.timedelta(minutes=inserted_act_dur) + else: + chatting_end_time = curr_time + datetime.timedelta(minutes=inserted_act_dur) + + for role, p in [("init", init_role), ("target", target_role)]: + if role == "init": + act_address = f" {target_role.name}" + act_event = (p.name, "chat with", target_role.name) + chatting_with = target_role.name + chatting_with_buffer = {} + chatting_with_buffer[target_role.name] = 800 + elif role == "target": + act_address = f" {init_role.name}" + act_event = (p.name, "chat with", init_role.name) + chatting_with = init_role.name + chatting_with_buffer = {} + chatting_with_buffer[init_role.name] = 800 + + act_pronunciatio = "💬" + act_obj_description = None + act_obj_pronunciatio = None + act_obj_event = (None, None, None) + + await _create_react( + p, + inserted_act, + inserted_act_dur, + act_address, + act_event, + chatting_with, + convo, + chatting_with_buffer, + chatting_end_time, + act_pronunciatio, + act_obj_description, + act_obj_pronunciatio, + act_obj_event, + act_start_time, + ) + + +async def _create_react( + role: "STRole", + inserted_act: str, + inserted_act_dur: int, + act_address: str, + act_event: Tuple, + chatting_with: str, + chat: list, + chatting_with_buffer: dict, + chatting_end_time: datetime, + act_pronunciatio: str, + act_obj_description: str, + act_obj_pronunciatio: str, + act_obj_event: Tuple, + act_start_time=None, +): + p = role + scratch = role.rc.scratch + + min_sum = 0 + for i in range(scratch.get_f_daily_schedule_hourly_org_index()): + min_sum += scratch.f_daily_schedule_hourly_org[i][1] + start_hour = int(min_sum / 60) + + if scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index()][1] >= 120: + end_hour = ( + start_hour + scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index()][1] / 60 + ) + + elif ( + scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index()][1] + + scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index() + 1][1] + ): + end_hour = start_hour + ( + ( + scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index()][1] + + scratch.f_daily_schedule_hourly_org[scratch.get_f_daily_schedule_hourly_org_index() + 1][1] + ) + / 60 + ) + + else: + end_hour = start_hour + 2 + end_hour = int(end_hour) + + dur_sum = 0 + count = 0 + start_index = None + end_index = None + for act, dur in scratch.f_daily_schedule: + if dur_sum >= start_hour * 60 and start_index is None: + start_index = count + if dur_sum >= end_hour * 60 and end_index is None: + end_index = count + dur_sum += dur + count += 1 + + ret = await generate_new_decomp_schedule(p, inserted_act, inserted_act_dur, start_hour, end_hour) + scratch.f_daily_schedule[start_index:end_index] = ret + scratch.add_new_action( + act_address, + inserted_act_dur, + inserted_act, + act_pronunciatio, + act_event, + chatting_with, + chat, + chatting_with_buffer, + chatting_end_time, + act_obj_description, + act_obj_pronunciatio, + act_obj_event, + act_start_time, + ) + + +async def _wait_react(role: "STRole", reaction_mode: str): + scratch = role.rc.scratch + + inserted_act = f'waiting to start {scratch.act_description.split("(")[-1][:-1]}' + end_time = datetime.datetime.strptime(reaction_mode[6:].strip(), "%B %d, %Y, %H:%M:%S") + inserted_act_dur = ( + (end_time.minute + end_time.hour * 60) - (scratch.curr_time.minute + scratch.curr_time.hour * 60) + 1 + ) + + act_address = f" {scratch.curr_tile[0]} {scratch.curr_tile[1]}" + act_event = (role.name, "waiting to start", scratch.act_description.split("(")[-1][:-1]) + chatting_with = None + chat = None + chatting_with_buffer = None + chatting_end_time = None + + act_pronunciatio = "⌛" + act_obj_description = None + act_obj_pronunciatio = None + act_obj_event = (None, None, None) + + await _create_react( + role, + inserted_act, + inserted_act_dur, + act_address, + act_event, + chatting_with, + chat, + chatting_with_buffer, + chatting_end_time, + act_pronunciatio, + act_obj_description, + act_obj_pronunciatio, + act_obj_event, + ) + + +async def generate_convo(init_role: "STRole", target_role: "STRole") -> Union[list, int]: + convo = await agent_conversation(init_role, target_role) + all_utt = "" + + for row in convo: + speaker = row[0] + utt = row[1] + all_utt += f"{speaker}: {utt}\n" + + convo_length = math.ceil(int(len(all_utt) / 8) / 30) + + return convo, convo_length + + +async def generate_convo_summary(conv: list[list[str]]) -> str: + conv_summary = await SummarizeConv().run(conv) + return conv_summary + + +async def generate_new_decomp_schedule( + role: "STRole", inserted_act: str, inserted_act_dur: int, start_hour: int, end_hour: int +): + # Step 1: Setting up the core variables for the function. + #

is the role whose schedule we are editing right now. + scratch = role.rc.scratch + # indicates the number of minutes that have passed today. + today_min_pass = int(scratch.curr_time.hour) * 60 + int(scratch.curr_time.minute) + 1 + + # Step 2: We need to create and . + main_act_dur = [] + truncated_act_dur = [] + dur_sum = 0 # duration sum + count = 0 # enumerate count + truncated_fin = False + + logger.debug(f"DEBUG::: {scratch.name}") + for act, dur in scratch.f_daily_schedule: + if (dur_sum >= start_hour * 60) and (dur_sum < end_hour * 60): + main_act_dur += [[act, dur]] + if dur_sum <= today_min_pass: + truncated_act_dur += [[act, dur]] + elif dur_sum > today_min_pass and not truncated_fin: + # We need to insert that last act, duration list like this one: + # e.g., ['wakes up and completes her morning routine (wakes up...)', 2] + truncated_act_dur += [[scratch.f_daily_schedule[count][0], dur_sum - today_min_pass]] + truncated_act_dur[-1][-1] -= ( + dur_sum - today_min_pass + ) # DEC 7 DEBUG;.. is the +1 the right thing to do??? + # DEC 7 DEBUG;.. is the +1 the right thing to do??? + # truncated_act_dur[-1][-1] -= (dur_sum - today_min_pass + 1) + logger.debug(f"DEBUG::: {truncated_act_dur}") + + # DEC 7 DEBUG;.. is the +1 the right thing to do??? + # truncated_act_dur[-1][-1] -= (dur_sum - today_min_pass) + truncated_fin = True + dur_sum += dur + count += 1 + + main_act_dur = main_act_dur + + x = ( + truncated_act_dur[-1][0].split("(")[0].strip() + + " (on the way to " + + truncated_act_dur[-1][0].split("(")[-1][:-1] + + ")" + ) + truncated_act_dur[-1][0] = x + + if "(" in truncated_act_dur[-1][0]: + inserted_act = truncated_act_dur[-1][0].split("(")[0].strip() + " (" + inserted_act + ")" + + # To do inserted_act_dur+1 below is an important decision but I'm not sure + # if I understand the full extent of its implications. Might want to + # revisit. + truncated_act_dur += [[inserted_act, inserted_act_dur]] + start_time_hour = datetime.datetime(2022, 10, 31, 0, 0) + datetime.timedelta(hours=start_hour) + end_time_hour = datetime.datetime(2022, 10, 31, 0, 0) + datetime.timedelta(hours=end_hour) + + return await NewDecompSchedule().run( + role, main_act_dur, truncated_act_dur, start_time_hour, end_time_hour, inserted_act, inserted_act_dur + ) + + +async def _long_term_planning(role: "STRole", new_day: bool): + """ + Formulates the role's daily long-term plan if it is the start of a new + day. This basically has two components: first, we create the wake-up hour, + and second, we create the hourly schedule based on it. + INPUT + new_day: Indicates whether the current time signals a "First day", + "New day", or False (for neither). This is important because we + create the roles' long term planning on the new day. + """ + # We start by creating the wake up hour for the role. + wake_up_hour = await WakeUp().run(role) + wake_up_hour = int(wake_up_hour) + logger.info(f"Role: {role.name} long_term_planning, wake_up_hour: {wake_up_hour}") + + # When it is a new day, we start by creating the daily_req of the role. + # Note that the daily_req is a list of strings that describe the role's + # day in broad strokes. + if new_day == "First day": + # Bootstrapping the daily plan for the start of then generation: + # if this is the start of generation (so there is no previous day's + # daily requirement, or if we are on a new day, we want to create a new + # set of daily requirements. + role.scratch.daily_req = await GenDailySchedule().run(role, wake_up_hour) + logger.info(f"Role: {role.name} daily requirements: {role.scratch.daily_req}") + elif new_day == "New day": + revise_identity(role) + + # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - TODO + # We need to create a new daily_req here... + role.scratch.daily_req = role.scratch.daily_req + + # Based on the daily_req, we create an hourly schedule for the role, + # which is a list of todo items with a time duration (in minutes) that + # add up to 24 hours. + role.scratch.f_daily_schedule = await GenHourlySchedule().run(role, wake_up_hour) + logger.info(f"Role: {role.name} f_daily_schedule: {role.scratch.f_daily_schedule}") + role.scratch.f_daily_schedule_hourly_org = role.scratch.f_daily_schedule[:] + + # Added March 4 -- adding plan to the memory. + thought = f"This is {role.scratch.name}'s plan for {role.scratch.curr_time.strftime('%A %B %d')}:" + for i in role.scratch.daily_req: + thought += f" {i}," + thought = thought[:-1] + "." + created = role.scratch.curr_time + expiration = role.scratch.curr_time + datetime.timedelta(days=30) + s, p, o = (role.scratch.name, "plan", role.scratch.curr_time.strftime("%A %B %d")) + keywords = set(["plan"]) + thought_poignancy = 5 + thought_embedding_pair = (thought, get_embedding(thought)) + role.a_mem.add_thought( + created, expiration, s, p, o, thought, keywords, thought_poignancy, thought_embedding_pair, None + ) + + +async def _determine_action(role: "STRole"): + """ + Creates the next action sequence for the role. + The main goal of this function is to run "add_new_action" on the role's + scratch space, which sets up all the action related variables for the next + action. + As a part of this, the role may need to decompose its hourly schedule as + needed. + INPUT + role: Current instance whose action we are determining. + """ + + def determine_decomp(act_desp, act_dura): + """ + Given an action description and its duration, we determine whether we need + to decompose it. If the action is about the agent sleeping, we generally + do not want to decompose it, so that's what we catch here. + + INPUT: + act_desp: the description of the action (e.g., "sleeping") + act_dura: the duration of the action in minutes. + OUTPUT: + a boolean. True if we need to decompose, False otherwise. + """ + if "sleep" not in act_desp and "bed" not in act_desp: + return True + elif "sleeping" in act_desp or "asleep" in act_desp or "in bed" in act_desp: + return False + elif "sleep" in act_desp or "bed" in act_desp: + if act_dura > 60: + return False + return True + + # The goal of this function is to get us the action associated with + # . As a part of this, we may need to decompose some large + # chunk actions. + # Importantly, we try to decompose at least two hours worth of schedule at + # any given point. + curr_index = role.scratch.get_f_daily_schedule_index() + curr_index_60 = role.scratch.get_f_daily_schedule_index(advance=60) + + logger.info(f"f_daily_schedule: {role.scratch.f_daily_schedule}") + # * Decompose * + # During the first hour of the day, we need to decompose two hours + # sequence. We do that here. + if curr_index == 0: + # This portion is invoked if it is the first hour of the day. + act_desp, act_dura = role.scratch.f_daily_schedule[curr_index] + if act_dura >= 60: + # We decompose if the next action is longer than an hour, and fits the + # criteria described in determine_decomp. + if determine_decomp(act_desp, act_dura): + role.scratch.f_daily_schedule[curr_index : curr_index + 1] = await TaskDecomp().run( + role, act_desp, act_dura + ) + if curr_index_60 + 1 < len(role.scratch.f_daily_schedule): + act_desp, act_dura = role.scratch.f_daily_schedule[curr_index_60 + 1] + if act_dura >= 60: + if determine_decomp(act_desp, act_dura): + role.scratch.f_daily_schedule[curr_index_60 + 1 : curr_index_60 + 2] = await TaskDecomp().run( + role, act_desp, act_dura + ) + + if curr_index_60 < len(role.scratch.f_daily_schedule): + # If it is not the first hour of the day, this is always invoked (it is + # also invoked during the first hour of the day -- to double up so we can + # decompose two hours in one go). Of course, we need to have something to + # decompose as well, so we check for that too. + if role.scratch.curr_time.hour < 23: + # And we don't want to decompose after 11 pm. + act_desp, act_dura = role.scratch.f_daily_schedule[curr_index_60] + if act_dura >= 60: + if determine_decomp(act_desp, act_dura): + role.scratch.f_daily_schedule[curr_index_60 : curr_index_60 + 1] = await TaskDecomp().run( + role, act_desp, act_dura + ) + # * End of Decompose * + + # Generate an instance from the action description and duration. By + # this point, we assume that all the relevant actions are decomposed and + # ready in f_daily_schedule. + logger.debug("DEBUG LJSDLFSKJF") + for i in role.scratch.f_daily_schedule: + logger.debug(i) + logger.debug(curr_index) + logger.debug(len(role.scratch.f_daily_schedule)) + logger.debug(role.scratch.name) + + # 1440 + x_emergency = 0 + for i in role.scratch.f_daily_schedule: + x_emergency += i[1] + + if 1440 - x_emergency > 0: + logger.info(f"x_emergency__AAA: {x_emergency}") + role.scratch.f_daily_schedule += [["sleeping", 1440 - x_emergency]] + + act_desp, act_dura = role.scratch.f_daily_schedule[curr_index] + + new_action_details = await GenActionDetails().run(role, act_desp, act_dura) + # Adding the action to role's queue. + role.scratch.add_new_action(**new_action_details) + + +def revise_identity(role: "STRole"): + p_name = role.scratch.name + + focal_points = [ + f"{p_name}'s plan for {role.scratch.get_str_curr_date_str()}.", + f"Important recent events for {p_name}'s life.", + ] + retrieved = new_agent_retrieve(role, focal_points) + + statements = "[Statements]\n" + for key, val in retrieved.items(): + for i in val: + statements += f"{i.created.strftime('%A %B %d -- %H:%M %p')}: {i.embedding_key}\n" + + plan_prompt = statements + "\n" + plan_prompt += f"Given the statements above, is there anything that {p_name} should remember as they plan for" + plan_prompt += f" *{role.scratch.curr_time.strftime('%A %B %d')}*? " + plan_prompt += "If there is any scheduling information, be as specific as possible (include date, time, and location if stated in the statement)\n\n" + plan_prompt += f"Write the response from {p_name}'s perspective." + plan_note = LLM().ask(plan_prompt) + + thought_prompt = statements + "\n" + thought_prompt += ( + f"Given the statements above, how might we summarize {p_name}'s feelings about their days up to now?\n\n" + ) + thought_prompt += f"Write the response from {p_name}'s perspective." + thought_note = LLM().ask(thought_prompt) + + currently_prompt = ( + f"{p_name}'s status from {(role.scratch.curr_time - datetime.timedelta(days=1)).strftime('%A %B %d')}:\n" + ) + currently_prompt += f"{role.scratch.currently}\n\n" + currently_prompt += f"{p_name}'s thoughts at the end of {(role.scratch.curr_time - datetime.timedelta(days=1)).strftime('%A %B %d')}:\n" + currently_prompt += (plan_note + thought_note).replace("\n", "") + "\n\n" + currently_prompt += f"It is now {role.scratch.curr_time.strftime('%A %B %d')}. Given the above, write {p_name}'s status for {role.scratch.curr_time.strftime('%A %B %d')} that reflects {p_name}'s thoughts at the end of {(role.scratch.curr_time - datetime.timedelta(days=1)).strftime('%A %B %d')}. Write this in third-person talking about {p_name}." + currently_prompt += "If there is any scheduling information, be as specific as possible (include date, time, and location if stated in the statement).\n\n" + currently_prompt += "Follow this format below:\nStatus: " + new_currently = LLM().ask(currently_prompt) + + role.scratch.currently = new_currently + + daily_req_prompt = role.scratch.get_str_iss() + "\n" + daily_req_prompt += f"Today is {role.scratch.curr_time.strftime('%A %B %d')}. Here is {role.scratch.name}'s plan today in broad-strokes (with the time of the day. e.g., have a lunch at 12:00 pm, watch TV from 7 to 8 pm).\n\n" + daily_req_prompt += "Follow this format (the list should have 4~6 items but no more):\n" + daily_req_prompt += "1. wake up and complete the morning routine at