Skip to content

Latest commit

 

History

History
103 lines (75 loc) · 2.84 KB

QUICKSTART.md

File metadata and controls

103 lines (75 loc) · 2.84 KB

Quick Start: Process and Query Your Documents

Process PDFs and query them interactively using the web interface.

Prerequisites

  • Go 1.21+
  • Node.js 18+
  • Python 3.10+
  • Claude API access (via ANTHROPIC_API_KEY or claude CLI authentication)

Start the Application

# Start the backend and frontend
./run.sh

# Open in browser
open http://localhost:3000

Process a Document

  1. Create a session - Click "New Session" on the home page
  2. Add a document - Click "Add Document" and enter the path to your PDF
    • Supports ~/Downloads/file.pdf paths
  3. Start processing - Click "Process" to begin extraction
  4. Watch progress - See real-time chunk progress as facts are extracted

Query Your Facts

  1. Navigate to your session
  2. Click Query in the header
  3. Ask natural language questions:
    • "Does the system implement 2-factor authentication?"
    • "What are the data retention policies?"
    • "What security controls are described?"
  4. View answers with clickable source citations
  5. Click citations to see the source context with highlighted quotes

Browse Facts

  1. Navigate to your session
  2. Click Browse Facts in the header
  3. Search and filter through extracted facts
  4. Click any fact to see its source context

What You Get

  • PDF extraction - Handles text-based and scanned PDFs
  • Structured facts - 8 metadata fields per fact
  • Source verification - Every fact linked to source text
  • Parallel processing - Fast extraction with progress visualization
  • Natural language Q&A - Query your facts conversationally

Session Structure

Each session stores:

sessions/{session_id}/
├── metadata.json      # Session metadata & document registry
├── text/              # Extracted PDF text
├── chunks/            # Source text chunks (for context panel)
├── facts/             # Extracted facts per chunk
└── summaries/         # Document summaries

Troubleshooting

Issue: Document shows "PDF not found" Solution: Use absolute paths or ~/ for home directory

Issue: No API key found Solution: Set ANTHROPIC_API_KEY or run claude login

Issue: Processing seems stuck Solution: Check the backend logs in the terminal running ./run.sh

API Usage

You can also interact with the API directly:

# List sessions
curl http://localhost:8080/api/sessions

# Create session
curl -X POST http://localhost:8080/api/sessions

# Add document
curl -X POST http://localhost:8080/api/sessions/{id}/documents \
  -H "Content-Type: application/json" \
  -d '{"path": "~/Downloads/report.pdf"}'

# Start processing
curl -X POST http://localhost:8080/api/sessions/{id}/process

# Query facts
curl -X POST http://localhost:8080/api/sessions/{id}/query \
  -H "Content-Type: application/json" \
  -d '{"query": "What security controls are implemented?"}'