Merge pull request #628 from iorisa/fixbug/role/assistant

fixbug: 修复通用智能体role及其相关的TalkAction和SkillAction

tests/metagpt/actions/test_skill_action.py

@@ -0,0 +1,65 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+@Time    : 2023/9/19
+@Author  : mashenquan
+@File    : test_skill_action.py
+@Desc    : Unit tests.
+"""
+import pytest
+
+from metagpt.actions.skill_action import ArgumentsParingAction, SkillAction
+from metagpt.learn.skill_loader import Example, Parameter, Returns, Skill
+
+
+class TestSkillAction:
+    skill = Skill(
+        name="text_to_image",
+        description="Create a drawing based on the text.",
+        id="text_to_image.text_to_image",
+        x_prerequisite={
+            "configurations": {
+                "OPENAI_API_KEY": {
+                    "type": "string",
+                    "description": "OpenAI API key, For more details, checkout: `https://platform.openai.com/account/api-keys`",
+                },
+                "METAGPT_TEXT_TO_IMAGE_MODEL_URL": {"type": "string", "description": "Model url."},
+            },
+            "required": {"oneOf": ["OPENAI_API_KEY", "METAGPT_TEXT_TO_IMAGE_MODEL_URL"]},
+        },
+        parameters={
+            "text": Parameter(type="string", description="The text used for image conversion."),
+            "size_type": Parameter(type="string", description="size type"),
+        },
+        examples=[
+            Example(ask="Draw a girl", answer='text_to_image(text="Draw a girl", size_type="512x512")'),
+            Example(ask="Draw an apple", answer='text_to_image(text="Draw an apple", size_type="512x512")'),
+        ],
+        returns=Returns(type="string", format="base64"),
+    )
+
+    @pytest.mark.asyncio
+    async def test_parser(self):
+        args = ArgumentsParingAction.parse_arguments(
+            skill_name="text_to_image", txt='`text_to_image(text="Draw an apple", size_type="512x512")`'
+        )
+        assert args.get("text") == "Draw an apple"
+        assert args.get("size_type") == "512x512"
+
+    @pytest.mark.asyncio
+    async def test_parser_action(self):
+        parser_action = ArgumentsParingAction(skill=self.skill, ask="Draw an apple")
+        rsp = await parser_action.run()
+        assert rsp
+        assert parser_action.args
+        assert parser_action.args.get("text") == "Draw an apple"
+        assert parser_action.args.get("size_type") == "512x512"
+
+        action = SkillAction(skill=self.skill, args=parser_action.args)
+        rsp = await action.run()
+        assert rsp
+        assert "image/png;base64," in rsp.content
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-s"])

tests/metagpt/learn/test_skill_loader.py

@@ -6,12 +6,14 @@
 @File    : test_skill_loader.py
 @Desc    : Unit tests.
 """
+import pytest
 
 from metagpt.config import CONFIG
-from metagpt.learn.skill_loader import SkillLoader
+from metagpt.learn.skill_loader import SkillsDeclaration
 
 
-def test_suite():
+@pytest.mark.asyncio
+async def test_suite():
     CONFIG.agent_skills = [
         {"id": 1, "name": "text_to_speech", "type": "builtin", "config": {}, "enabled": True},
         {"id": 2, "name": "text_to_image", "type": "builtin", "config": {}, "enabled": True},
@@ -21,7 +23,7 @@ def test_suite():
         {"id": 6, "name": "knowledge", "type": "builtin", "config": {}, "enabled": True},
         {"id": 6, "name": "web_search", "type": "builtin", "config": {}, "enabled": True},
     ]
-    loader = SkillLoader()
+    loader = await SkillsDeclaration.load()
     skills = loader.get_skill_list()
     assert skills
     assert len(skills) >= 3
@@ -29,7 +31,7 @@ def test_suite():
         assert desc
         assert name
 
-    entity = loader.get_entity("Assistant")
+    entity = loader.entities.get("Assistant")
     assert entity
     assert entity.skills
     for sk in entity.skills:
@@ -38,4 +40,4 @@ def test_suite():
 
 
 if __name__ == "__main__":
-    test_suite()
+    pytest.main([__file__, "-s"])

tests/metagpt/roles/test_assistant.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+@Time    : 2023/12/25
+@Author  : mashenquan
+@File    : test_asssistant.py
+@Desc    : Used by AgentStore.
+"""
+import pytest
+from pydantic import BaseModel
+
+from metagpt.actions.skill_action import SkillAction
+from metagpt.actions.talk_action import TalkAction
+from metagpt.config import CONFIG
+from metagpt.logs import logger
+from metagpt.memory.brain_memory import BrainMemory
+from metagpt.roles.assistant import Assistant
+from metagpt.schema import Message
+from metagpt.utils.common import any_to_str
+
+
+@pytest.mark.asyncio
+async def test_run():
+    CONFIG.language = "Chinese"
+
+    class Input(BaseModel):
+        memory: BrainMemory
+        language: str
+        agent_description: str
+        cause_by: str
+
+    inputs = [
+        {
+            "memory": {
+                "history": [
+                    {
+                        "content": "who is tulin",
+                        "role": "user",
+                        "id": 1,
+                    },
+                    {"content": "The one who eaten a poison apple.", "role": "assistant"},
+                ],
+                "knowledge": [{"content": "tulin is a scientist."}],
+                "last_talk": "what's apple?",
+            },
+            "language": "English",
+            "agent_description": "chatterbox",
+            "cause_by": any_to_str(TalkAction),
+        },
+        {
+            "memory": {
+                "history": [
+                    {
+                        "content": "can you draw me an picture?",
+                        "role": "user",
+                        "id": 1,
+                    },
+                    {"content": "Yes, of course. What do you want me to draw", "role": "assistant"},
+                ],
+                "knowledge": [{"content": "tulin is a scientist."}],
+                "last_talk": "Draw me an apple.",
+            },
+            "language": "English",
+            "agent_description": "painter",
+            "cause_by": any_to_str(SkillAction),
+        },
+    ]
+    CONFIG.agent_skills = [
+        {"id": 1, "name": "text_to_speech", "type": "builtin", "config": {}, "enabled": True},
+        {"id": 2, "name": "text_to_image", "type": "builtin", "config": {}, "enabled": True},
+        {"id": 3, "name": "ai_call", "type": "builtin", "config": {}, "enabled": True},
+        {"id": 3, "name": "data_analysis", "type": "builtin", "config": {}, "enabled": True},
+        {"id": 5, "name": "crawler", "type": "builtin", "config": {"engine": "ddg"}, "enabled": True},
+        {"id": 6, "name": "knowledge", "type": "builtin", "config": {}, "enabled": True},
+        {"id": 6, "name": "web_search", "type": "builtin", "config": {}, "enabled": True},
+    ]
+
+    for i in inputs:
+        seed = Input(**i)
+        CONFIG.language = seed.language
+        CONFIG.agent_description = seed.agent_description
+        role = Assistant(language="Chinese")
+        role.memory = seed.memory  # Restore historical conversation content.
+        while True:
+            has_action = await role.think()
+            if not has_action:
+                break
+            msg: Message = await role.act()
+            logger.info(msg)
+            assert msg
+            assert msg.cause_by == seed.cause_by
+            assert msg.content
+            # # Retrieve user terminal input.
+            # logger.info("Enter prompt")
+            # talk = input("You: ")
+            # await role.talk(talk)
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-s"])