Quick Start Guide

1. Get Your OpenRouter API Key

Visit OpenRouter
Sign up for a free account
Go to Keys section
Create a new API key
Copy your API key

2. Configure the Tool

# Copy the environment template
cp .env.example .env

# Edit .env and add your API key
echo "OPENROUTER_API_KEY=your_key_here" >> .env

Or set it directly in your shell:

export OPENROUTER_API_KEY=your_key_here

3. Run Your First Analysis

# Basic usage
uv run pdf-to-kcf your_document.pdf

# This will create a file: your_document_insights.json

Example Output

The tool will generate a JSON file with structured insights like:

{
  "insights": [
    {
      "type": "fact",
      "insight": "The company revenue increased by 25% in Q4",
      "content": "According to the financial report, Q4 revenue reached $2.5M, up 25% from Q3...",
      "attributes": [
        {"attribute": "source", "value": "Financial Report"},
        {"attribute": "quarter", "value": "Q4"},
        {"attribute": "confidence", "value": "high"}
      ]
    },
    {
      "type": "opinion",
      "insight": "The author recommends investing in AI infrastructure",
      "content": "We strongly believe that investing in AI infrastructure is critical...",
      "attributes": [
        {"attribute": "sentiment", "value": "positive"},
        {"attribute": "section", "value": "recommendations"}
      ]
    }
  ]
}

Advanced Usage

Try Different Models

# Use GPT-4
uv run pdf-to-kcf document.pdf -m openai/gpt-4o

# Use Gemini
uv run pdf-to-kcf document.pdf -m google/gemini-pro-1.5

# Use Llama 3.1
uv run pdf-to-kcf document.pdf -m meta-llama/llama-3.1-70b-instruct

Custom Output Location

uv run pdf-to-kcf document.pdf -o /path/to/output.json

Start from Specific Page

# Start analysis from page 5 (0-indexed, so this is the 6th page)
uv run pdf-to-kcf document.pdf -s 5

How It Works

The tool loads your PDF and extracts text
An AI agent analyzes the content starting from page 0 (or your specified page)
The agent autonomously decides if it needs to read more pages
It extracts structured insights classified as facts, opinions, or comments
Each insight includes the original content and relevant metadata
Results are saved to a JSON file

Pricing

OpenRouter charges based on the model you use:

Claude 3.5 Sonnet (default): ~$3 per million input tokens
GPT-4o: ~$2.50 per million input tokens
Llama 3.1 70B: ~$0.35 per million input tokens

Most PDFs will cost just a few cents to analyze.

Troubleshooting

"No API key found"

Make sure you've set OPENROUTER_API_KEY in your .env file or environment.

"Model not found"

Check the model name format: <provider>/<model-name> (e.g., anthropic/claude-3.5-sonnet) See available models at https://openrouter.ai/models

"PDF not found"

Use the full path to your PDF file, or navigate to the directory containing it first.

Next Steps

Read the full README.md for more details
Check CLAUDE.md for architecture details
See OpenRouter models for all available models

3.2 KiB Raw Blame History