Dragula

Dragula (D-RAG-ula) is a Python library that indexes a Python codebase and serves a local web UI for AI-generated object documentation over pluggable LLM providers.

Features

• CLI commands: init, index, serve
• AST-based symbol extraction (module/class/function/method)
• SQLite storage for structured symbol data
• ChromaDB storage for semantic chunk vectors
• FastAPI web UI to browse symbols and request AI descriptions (also available via HTTP API)

Install

Requires Python 3.11+.

pip install dragula

Configuration

Dragula keeps its config and local databases under <project_root>/.dragula/ (fixed location, similar in spirit to an Alembic env directory).

Initialize once in the target project (creates .dragula/, a template config.ini, SQLite tables, and the Chroma data directory):

dragula init .

Edit .dragula/config.ini for your LLM providers.

Minimal Gemini config

[app]
top_k = 6

[chat]
provider = gemini
model = gemini-2.5-flash
api_key = ${GEMINI_API_KEY}
base_url =
timeout_seconds = 60

[embedding]
provider = gemini
model = gemini-embedding-001
api_key = ${GEMINI_API_KEY}
base_url =
timeout_seconds = 60

Mixed providers (chat remote, embedding local)

[app]
top_k = 6

[chat]
provider = gemini
model = gemini-2.5-flash
api_key = ${GEMINI_API_KEY}
base_url =
timeout_seconds = 60

[embedding]
provider = openai_compatible
model = text-embedding-3-small
base_url = http://localhost:11434/v1
api_key =
timeout_seconds = 60

Usage

Initialize (first time only):

dragula init .

Build index:

dragula index .

Run web UI (default http://127.0.0.1:8000):

dragula serve .

Optional server binding:

dragula serve . --host 0.0.0.0 --port 8080

HTTP API

Data layout

Under the project root:

• .dragula/config.ini — configuration
• .dragula/symbols.sqlite3 — structured symbol/chunk/description tables
• .dragula/chroma/ — Chroma persistent vector store