🔬 Local Deep Research Agent Platform

A local deep research agentic platform with Ollama and Local LLMs

✨ Features

User Interface:

🔍 Core Research Capabilities

• Deep Research Workflow: Multi-loop research process with query generation, web search, summarization, and reflection
• Local LLM Integration: Works with Ollama and LMStudio for complete privacy
• Multiple Search APIs: DuckDuckGo, Tavily, Perplexity, etc.
• Real-time Progress: Track progress in realtime step-by-step visualization
• Smart Source Management: Automatic deduplication and relevance scoring
• Vector Embeddings: Semantic search across research history using sentence transformers
• Research History: Persistent storage with similarity search
• Configuration Management: Easy LLM and search API configuration
• Export Capabilities: Download results in Markdown or text format

🚀 Quick Start

📋 Prerequisites

Requirement	Version	Description
🐍 Python	3.11+	Core runtime environment
🦙 Ollama/LMStudio	Latest	Local LLM provider
📦 Git	Latest	Version control

🛠️ Installation Steps

# 1. Clone and navigate to the directory
cd streamlit-deep-researcher

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure environment
cp .env.example .env
# Edit .env with your settings

# 4. Run the application
streamlit run app.py

# 5. Open your browser at http://localhost:8501

🎯 First Steps

1. Navigate to http://localhost:8501
2. Configure your LLM settings in the sidebar
3. Start researching!

⚙️ Configuration

🦙 LLM Setup

Ollama (Recommended)

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull a model
ollama pull gemma3:latest
# or
ollama pull llama3.2

# Start Ollama (usually runs automatically)
ollama serve

LMStudio

1. Download and install LMStudio
2. Load a model in LMStudio
3. Start the local server
4. Configure the base URL in the sidebar

🔧 Environment Variables

Configuration File (.env)

# LLM Settings
LOCAL_LLM="gemma3:latest"
LLM_PROVIDER="ollama"
OLLAMA_BASE_URL="http://localhost:11434/"
LMSTUDIO_BASE_URL="http://localhost:1234/v1"

# Research Settings
MAX_WEB_RESEARCH_LOOPS=3
SEARCH_API="duckduckgo"
FETCH_FULL_PAGE=true

# API Keys (optional)
TAVILY_API_KEY=""
PERPLEXITY_API_KEY=""

# Vector Embeddings
EMBEDDING_MODEL="all-MiniLM-L6-v2"

🎯 Usage Guide

🚀 Starting Research

Step	Action	Description
1️⃣	Configure LLM	Use the sidebar to set up your local LLM
2️⃣	Test Connection	Click "Test LLM Connection" to verify setup
3️⃣	Enter Research Topic	Type your research question in the main input
4️⃣	Start Research	Click "🚀 Start Research" and watch the progress
5️⃣	View Results	Explore the comprehensive research summary

💡 Research Examples

Sample Research Topics

• 🔬 "Latest developments in quantum computing 2024"
• 🌱 "Climate change impact on agriculture"
• 🤖 "Artificial intelligence safety research"
• ⚡ "Renewable energy storage technologies"
• 🧬 "CRISPR gene editing recent advances"

🔧 Advanced Features

Research History

• Automatic Saving: All research sessions are saved automatically
• Semantic Search: Find similar research using vector embeddings
• Session Management: Load, view, and manage previous research

Configuration Options

• Research Depth: Adjust number of research loops (1-10)
• Search APIs: Choose between different search providers
• LLM Models: Switch between different local models
• Embedding Models: Select sentence transformer models

📊 Architecture

📁 Project Structure

streamlit-deep-researcher/
├── 📱 app.py                          # Main Streamlit application
├── 📋 requirements.txt                # Python dependencies
├── ⚙️ .env                           # Configuration file
├── 📂 config/
│   ├── __init__.py
│   └── ⚙️ settings.py                # Configuration management
├── 📂 research/
│   ├── __init__.py
│   ├── 🔄 graph.py                   # LangGraph research workflow
│   ├── 📊 state.py                   # Research state management
│   ├── 🔍 utils.py                   # Search utilities
│   ├── 💬 prompts.py                 # LLM prompts
│   └── 🤖 llm_providers.py           # Ollama/LMStudio integration
├── 📂 storage/
│   ├── __init__.py
│   └── 🗄️ vector_store.py            # Vector embeddings storage
├── 📂 components/
│   ├── __init__.py
│   ├── 📈 progress_display.py        # Progress visualization
│   └── 🎛️ sidebar.py                 # Configuration sidebar
└── 📂 assets/
    └── 🗃️ research_history.db        # SQLite database (auto-created)

🏗️ System Components

Component	Purpose	Technology
Frontend	User interface	Streamlit
Research Engine	Core logic	LangGraph
LLM Integration	AI processing	Ollama/LMStudio
Vector Store	Embeddings	Sentence Transformers
Search APIs	Web research	Multiple providers

🔧 Troubleshooting

❗ Common Issues

🔌 LLM Connection Failed

# Check if Ollama is running
ollama list

# Restart Ollama
ollama serve

Solution: Ensure Ollama service is running and accessible on the configured port.

📦 Module Import Errors

# Reinstall dependencies
pip install -r requirements.txt --force-reinstall

Solution: Clean reinstall of all Python dependencies.

🐌 Slow Research Performance

Quick Fixes:

• Reduce research depth in sidebar
• Use lighter LLM models
• Disable "Fetch Full Page Content"

💾 Memory Issues

Optimization Steps:

• Use smaller embedding models
• Reduce research loops
• Clear old research sessions

⚡ Performance Optimization

Optimization	Recommendation	Impact
Model Size	Use gemma3:2b instead of larger models	🚀 Faster
Embeddings	Use all-MiniLM-L6-v2	🚀 Faster
Search Settings	Disable full page content	🚀 Faster
Research Depth	Start with 2-3 loops	⚖️ Balanced

🚀 Recent Updates

✨ Latest Enhancements

• 🔧 Enhanced research capabilities
• ⚡ Improved performance and reliability
• 🎨 Better user interface design
• 🔍 Advanced search functionality

📈 Performance Improvements

• Faster LLM response times
• Optimized vector embeddings
• Reduced memory usage
• Better error handling

Happy Researching! 🔬✨