Architecture#

The hosted deployment keeps interpretation and storage in separate layers:
An agent such as ChatGPT or Claude interprets the user’s request and any observations.
OAuth authorizes the agent for a workspace. Consent happens in the browser and is separate from inventory traffic.
The agent calls Quilombo through the MCP Streamable HTTP endpoint (
/mcp).Django validates workspace boundaries and writes inventory facts in transactions.
Neon PostgreSQL stores the tenant-scoped data.
Quilombo does not upload or interpret source media. The client can include provenance metadata when it writes facts, while semantic reasoning remains with the agent or client.
Regenerate#
From the repository root, install Graphviz and run:
uv run --no-project --python 3.14 --with diagrams python docs/architecture_diagram.py
The generated image is written to docs/_static/quilombo-architecture.png.