apunkt/rowboat
1
0
Fork 0
mirror of https://github.com/rowboatlabs/rowboat.git synced 2026-06-06 19:35:44 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge pull request #139 from rowboatlabs/dev

Browse source 
Dev changes
This commit is contained in:
Ramnique Singh 2025-05-20 20:10:32 +05:30 • committed by GitHub
commit 7400050aef
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 39 additions and 37 deletions
Download patch file Download diff file Expand all files Collapse all files

53
apps/docs/docs/using_rag.md
View file

@ -1,12 +1,12 @@
# Using RAG in Rowboat
Rowboat provides multiple ways to enhance your agents with Retrieval-Augmented Generation (RAG). This guide will help you set up and use each RAG feature.
Rowboat provides multiple ways to enhance your agents' context with Retrieval-Augmented Generation (RAG). This guide will help you set up and use each RAG features.
## Quick Start
Text RAG and local file uploads are enabled by default - no configuration needed! Just start using them right away.
## Available RAG Features
## RAG Features
### 1. Text RAG
✅ Enabled by default:
@ -21,8 +21,28 @@ Text RAG and local file uploads are enabled by default - no configuration needed
- Files are stored locally
- No configuration required
- Files are parsed using OpenAI by default
- For larger files, we recommend using Gemini models - see section below.
### 3. S3 File Uploads
#### 2.1 Using Gemini for File Parsing
To use Google's Gemini model for parsing uploaded PDFs, set the following variable:
```bash
# Enable Gemini for file parsing
export USE_GEMINI_FILE_PARSING=true
export GOOGLE_API_KEY=your_google_api_key
```
### 3. URL Scraping
Rowboat uses Firecrawl for URL scraping. To enable URL scraping, set the following variables:
```bash
export USE_RAG_SCRAPING=true
export FIRECRAWL_API_KEY=your_firecrawl_api_key
```
## Advanced RAG features
### 1. File Uploads Backed by S3
To enable S3 file uploads, set the following variables:
```bash
@ -36,20 +56,8 @@ export RAG_UPLOADS_S3_BUCKET=your_bucket_name
export RAG_UPLOADS_S3_REGION=your_region
```
### 4. URL Scraping
To enable URL scraping, set the following variables:
### 2. Changing Default Parsing Model
```bash
# Enable URL scraping
export USE_RAG_SCRAPING=true
# Firecrawl API key for web scraping
export FIRECRAWL_API_KEY=your_firecrawl_api_key
```
## File Parsing Options
### Default Parsing (OpenAI)
By default, uploaded PDF files are parsed using `gpt-4o`. You can customize this by setting the following:
```bash
@ -64,16 +72,7 @@ export FILE_PARSING_PROVIDER_BASE_URL=your-provider-base-url
export FILE_PARSING_PROVIDER_API_KEY=your-provider-api-key
```
### Using Gemini for File Parsing
To use Google's Gemini model for parsing uploaded PDFs, set the following variable:
```bash
# Enable Gemini for file parsing
export USE_GEMINI_FILE_PARSING=true
export GOOGLE_API_KEY=your_google_api_key
```
## Embedding Model options
### 3. Embedding Model Options
By default, Rowboat uses OpenAI's `text-embedding-3-small` model for generating embeddings. You can customize this by setting the following:
@ -101,4 +100,4 @@ export EMBEDDING_PROVIDER_BASE_URL=your-provider-base-url
export EMBEDDING_PROVIDER_API_KEY=your-provider-api-key
```
If you don't specify the provider settings, Rowboat will use OpenAI as the default provider.
If you don't specify the provider settings, Rowboat will use OpenAI as the default provider.

2
apps/rowboat/app/projects/[projectId]/config/app.tsx
View file

@ -305,7 +305,7 @@ function McpServersSection({
/>
<Input
label="SSE URL"
placeholder="https://localhost:8000/sse"
placeholder="http://host.docker.internal:8000/sse"
value={newServer.url}
onChange={(e) => {
setNewServer({ ...newServer, url: e.target.value });

2
apps/rowboat/app/projects/[projectId]/config/components/tools.tsx
View file

@ -210,7 +210,7 @@ function McpServersSection({ projectId }: { projectId: string }) {
<div className="space-y-2">
<label className="text-sm font-medium">SSE URL</label>
<Input
placeholder="https://localhost:8000/sse"
placeholder="http://host.docker.internal:8000/sse"
value={newServer.url}
onChange={(e) => {
setNewServer({ ...newServer, url: e.target.value });

19
apps/rowboat_agents/src/graph/core.py
View file

@ -153,8 +153,8 @@ async def run_turn_streamed(
# Initialize agents and get external tools
new_agents = get_agents(
agent_configs=agent_configs,
tool_configs=tool_configs,
agent_configs=agent_configs,
tool_configs=tool_configs,
complete_request=complete_request
)
new_agents = add_child_transfer_related_instructions_to_agents(new_agents)
@ -196,12 +196,15 @@ async def run_turn_streamed(
if event.type == "raw_response_event":
# Handle token usage counting
if hasattr(event.data, 'type') and event.data.type == "response.completed" and hasattr(event.data.response, 'usage'):
tokens_used["total"] += event.data.response.usage.total_tokens
tokens_used["prompt"] += event.data.response.usage.input_tokens
tokens_used["completion"] += event.data.response.usage.output_tokens
print('-'*50)
print(f"Found usage information. Updated cumulative tokens: {tokens_used}")
print('-'*50)
try:
tokens_used["total"] += event.data.response.usage.total_tokens
tokens_used["prompt"] += event.data.response.usage.input_tokens
tokens_used["completion"] += event.data.response.usage.output_tokens
print('-'*50)
print(f"Found usage information. Updated cumulative tokens: {tokens_used}")
print('-'*50)
except Exception as e:
print(f"Warning: Tokens used is likely not available for your chosen model: {e}")
web_search_messages = handle_web_search_event(event, current_agent)
for message in web_search_messages: