A Very simple offline documentation context provider for LLM via MCP
Find a file
2025-08-20 00:47:38 +02:00
autodoc fix: allow_backward_links = false 2025-08-20 00:47:38 +02:00
docker feat: add support for firecrawl simple 2025-08-13 00:27:37 +02:00
.gitignore feat: add debug urls file 2025-08-20 00:31:43 +02:00
docs.yml feat: add numpy 2025-08-20 00:27:07 +02:00
README.md feat: add support for firecrawl simple 2025-08-13 00:27:37 +02:00
requirements.txt feat: initial commit 2025-08-10 05:51:18 +02:00

AutoDoc

A simple offline documentation context provider for LLMs via MCP, designed as an alternative to context7

AutoDoc extracts and processes documentation to enhance LLM interactions through vector-based retrieval.

How It Works

  1. Web crawlers extract documentation
  2. Content is embedded using Ollama models
  3. Embeddings are stored in a vector database
  4. MCP retrieves relevant information based on natural language prompts

Tech Stack

  • llama-index: For document processing
  • chroma: Vector database
  • firecrawl: Web crawling
  • ollama: Embedding models

Quick Start

Windows

# Setup environment
copy docker\.env.example docker\.env

# Start services
$env:COMPOSE_FILE="docker\compose-fire.yml;docker\compose-chroma.yml"
# or with fire simple
$env:COMPOSE_FILE="docker/compose-fire-simple.yml;docker/compose-chroma.yml"
docker compose up -d

# Install and run
python -m pip install -r requirements.txt
python autodoc\main.py

Linux

# Setup environment
cp docker/.env.example docker/.env

# Start services
COMPOSE_FILE="docker/compose-fire.yml:docker/compose-chroma.yml"
# or with fire simple
COMPOSE_FILE="docker/compose-fire-simple.yml:docker/compose-chroma.yml"
docker compose up -d

# Install and run
pip install -r requirements.txt
python autodoc/main.py