GUIDE.md

Features & Usage Guide

Overview

The AI Research Agent is a powerful tool designed to automate comprehensive city and topic research, fill data gaps, and store structured data.

Features

1. Research

• Purpose: Perform multi-step deep research on defined criteria.
• Usage:
  1. Enter Job Keywords (e.g., "San Francisco Tech Scene").
  2. Define Research Criteria (e.g., "Top Startups", "Venture Capital Activity").
  3. Select an AI Model (e.g., Groq Compound).
  4. Click Start Research.
  5. Review real-time logs and results.
  6. Stop Research: Click the Stop button at any time to halt the process.

2. Criteria Builder

• Purpose: Manage reusable research criteria templates.
• Usage:
  1. Add new criteria with a Name and Description.
  2. Define the expected Output Schema (JSON format).
  3. Save for use in the Research tab.

3. Vector Store

• Purpose: Embed and store text data for semantic search.
• Usage:
  1. Enter Mixedbread API Key (for embeddings).
  2. Enter Pinecone API Key & Index Name.
  3. Paste large text content.
  4. Click Embed & Store.
  5. The system automatically chunks and stores the data.

4. HITL Deep Research (Agent Mode)

• Purpose: Let an AI architect design a custom research plan for complex queries.
• Usage:
  1. Go to the Agent tab (Purple Sparkle icon).
  2. Describe your goal (e.g., "Find top 5 AI startups in SF and save their CEO's LinkedIn").
  3. Click Generate Plan.
  4. Review the proposed Flowchart.
  5. Edit steps if needed (remove unwanted steps).
  6. Click Execute Plan (Coming Soon).

5. Settings

• Purpose: Configure global API keys and model preferences.
• Usage:
  • Set Groq API Key (Primary LLM).
  • Set Tavily API Key (Search Engine).
  • Select default models and fallback behavior.

5. Analytics

• Purpose: Track usage and research performance.
• Usage: View charts on job completion rates, token usage, and successful criteria extractions.