Web guide · derived from pinned Core sources
On this page
Serve a model locally
Guild includes a native OpenAI-compatible HTTP server. The Python layer in serve/ provides a web app and an Anthropic-shaped API in front of the engine. These are related but distinct entry points.
Native server
./build/guild serve <model>The typical local API base is:
http://127.0.0.1:11434/v1Use your binary’s guild serve --help for current bind, model, and request options. Compatibility is evolving; check the exact endpoint and streaming behavior required by your client.
Python server and web app
The one-click installer writes a run configuration used by serve/server.py. The Python layer also integrates MCP tools through serve/mcp.py and supports an --api-key option.
python serve/server.py --helpKeep the service bound to 127.0.0.1 during local use. Do not expose it beyond localhost without the Python server’s --api-key. Follow the repository’s deployment instructions for the version you use.
Cloud status
Illatio Cloud is a separate, planned hosted product. The local server command does not create a Cloud account or provision a hosted endpoint.
Sources: repository README, architecture, and correctness recovery.