Research Assistant Agent

Modular Multi-Provider Research System

A scalable research assistant with modular architecture supporting multiple LLM providers and extensible workflow nodes.

Features

• Modular Architecture - Pluggable components for maximum flexibility
• Multi-LLM Support - Easy switching between Ollama, OpenAI, Anthropic
• Interactive Workflow - Human review and approval steps
• Extensible Nodes - Add new research capabilities easily
• Structured Generation - JSON-constrained llm output for reliability
• Web Interface - Streamlit UI for easy interaction
• Configuration-Driven - YAML-based workflow management

Agent Graph

Quick Start

1. Install dependencies:

   pip install -r requirements.txt

2. Setup LLM Provider:

   # For Ollama (default)
   ollama serve
   ollama pull llama3.2
   
   # For OpenAI (coming soon)
   export OPENAI_API_KEY="your-key"
   
   # For Anthropic (coming soon)  
   export ANTHROPIC_API_KEY="your-key"

3. Run the application:

   # Web interface
   streamlit run main.py

Configuration

Provider Settings (config/settings.yaml)

llm_provider:
  type: "ollama"  # or "openai", "anthropic"
  
ollama:
  model: "llama3.2"
  base_url: "http://localhost:11434"
  
openai:
  model: "gpt-4"
  temperature: 0.7
  
anthropic:
  model: "claude-3-sonnet"
  temperature: 0.7

workflow:
  enable_human_review: true
  max_retries: 3
  timeout: 300

Adding Custom Nodes

from agents.nodes.base_node import BaseNode

class CustomAnalysisNode(BaseNode):
    def __init__(self, llm_provider):
        super().__init__(
            name="custom_analysis",
            llm_provider=llm_provider
        )
    
    def execute(self, state):
        # Your custom logic here
        return {"custom_results": "analysis_data"}

Extending the System

Add New LLM Provider

1. Implement BaseLLMProvider interface
2. Add to LLMProviderFactory
3. Update configuration schema

Add New Node Type

1. Extend BaseNode or ConditionalNode
2. Implement execute() method
3. Register in workflow configuration

Further Details

For a deeper dive into the system's architecture, refer to Architecture Details. This document provides an in-depth explanation of the modular design, workflow orchestration, and integration with multiple LLM providers.

Troubleshooting

Common Issues

LLM Provider Connection

# Check Ollama status
ollama list
curl http://localhost:11434/api/version

Configuration Issues

• Verify config/settings.yaml syntax
• Check provider-specific environment variables
• Ensure model names match available models