Seeker

A local research assistant that performs iterative, deep-dive investigations using search engines, web scraping, and LLMs. Combines breadth-first exploration with depth-first analysis for comprehensive topic coverage.

Clariciation of Search Intent

Research Process

End of Task

Features

• Iterative Research - Self-refining research direction based on initial findings
• Follow-up Question System - Clarifies research goals through interactive dialogue
• Configurable Parameters - Control research breadth (3-10) and depth (1-5)
• Multi-format Output - Generates detailed reports or concise answers
• LLM Agnostic - Supports OpenAI, Fireworks.ai, and local LLMs via custom endpoints
• Concurrent Processing - Parallelizes research tasks for faster results

Quick Start 🚀

Prerequisites

• Node.js v18+
• API keys for:
  • Firecrawl (web scraping)
  • OpenAI or Fireworks.ai

# Clone repository
git clone https://github.com/iBz-04/Seeker.git
cd Seeker

# Install dependencies
npm install

# Configure environment
cp .env.example .env.local
# Edit .env.local with your API keys

Basic Usage

npm start

? What would you like to research? Effects of climate change on coffee production
? Research breadth (3-10): 5
? Research depth (1-5): 2

Advanced Configuration ⚙️

Environment Variables

# Required
FIRECRAWL_API_KEY="your_firecrawl_key"
OPENAI_API_KEY="your_openai_key"  # OR
FIREWORKS_API_KEY="your_fireworks_key"

# Optional
RESEARCH_CONCURRENCY=5  # Parallel requests
MAX_TOKENS=4096         # LLM context window

Docker Deployment

docker compose up -d --build
docker exec -it deep-research npm run docker

Research Process Flow 🔄

1. Initial Query Analysis

   • Parse user's research question
   • Generate clarifying follow-up questions
   • Collect additional context through interactive Q&A

2. Search Strategy Development

   • Generate multiple search queries based on research goals
   • Prioritize queries using relevance scoring
   • Execute parallel web searches using Firecrawl

3. Content Analysis

   • Extract key insights from web content
   • Maintain context-aware knowledge graph
   • Identify promising research directions for deep-dive

4. Recursive Exploration

   • Repeat search/analysis process at specified depth
   • Prune irrelevant branches
   • Merge findings across iterations

5. Report Generation

   • Synthesize findings into structured markdown
   • Include sources and confidence ratings
   • Generate both technical and executive summaries

Customization Options 🛠️

Adjust Research Parameters

// deep-research.ts
const config = {
  maxConcurrency: 5,      // Parallel requests
  minRelevanceScore: 0.7, // Content filtering threshold
  maxPagesPerQuery: 10,   // Resource limits
  timeout: 30000          // Per-request timeout
};

Supported LLM Providers

# Local AI
OPENAI_BASE_URL="http://localhost:1234/v1"
OPENAI_MODEL="your-local-model"

# OpenRouter
OPENAI_BASE_URL="https://openrouter.ai/api/v1"
OPENAI_MODEL="anthropic/claude-3-haiku"

Example Output 📄

# Research Report: Climate Change Impact on Coffee Production

## Key Findings
- **Yield Reductions**: 50% projected decrease in suitable growing areas by 2050
- **Quality Impacts**: Increased temperatures correlate with flavor profile degradation
- **Economic Costs**: $2.5B annual losses anticipated in major producing regions

## Recommended Actions
1. Develop heat-resistant coffee cultivars
2. Implement agroforestry practices
3. Diversify geographic production areas

## Sources
1. [Journal of Agricultural Science] (https://example.com/source1)
2. [FAO Climate Report] (https://example.com/source2)

Troubleshooting ⚠️

Common Issues:

• Rate Limiting: Reduce RESEARCH_CONCURRENCY in free tier
• Timeouts: Increase TIMEOUT duration for complex queries
• Relevance Filtering: Adjust MIN_RELEVANCE_SCORE for sensitive content