Starport documentation
Find the area that answers your question and send a first request to a local gateway.
Starport is a self-hosted LLM inference gateway in one binary. It serves the OpenAI-compatible API at /v1 and the OpenRouter-compatible API at /api/v1, and Starmap supplies its provider and model facts.
Areas of this site
Each area answers one kind of question. Start at the area that matches your task.
| Area | Use it to |
|---|---|
| Start | Run a first gateway, issue keys, and inspect the catalog. |
| Choose an architecture | Select a deployment target, a catalog topology, and storage. |
| Configure | Set values, secret references, and paths. |
| Catalog lifecycle | Learn how a catalog generation becomes routable. |
| Operate Starmap | Run a central Starmap server or an air-gapped mirror. |
| Operate Starport | Manage authentication, identity, fleets, and upgrades. |
| Storage | Select backends, back up, restore, and tune caches. |
| API compatibility | Connect SDKs and read errors and limits. |
| Troubleshoot | Recover a gateway without a console session. |
Version facts in the page header
The page header shows the facts of the build that made this site. Compare them with your gateway before you follow a procedure.
- The Starport release identifies the binary that these pages describe.
- The Starmap version identifies the Starmap module that this release pins. That module reconciles catalog records into one canonical catalog.
Run starport version to see the release of an installed binary. If the release differs from the header, read the pages for your release.
Fastest path to a first request
This path uses a temporary development gateway. It keeps no data after the process stops.
- Install Starport. Refer to Run a persistent local gateway.
- Set a provider inference credential, for example
OPENAI_API_KEY. - Run
starport devand keep the terminal open. - Copy the gateway API key that the command prints.
- In a second terminal, send a request:
export STARPORT_API_KEY="<printed-gateway-api-key>"
curl --fail-with-body \
-H "Authorization: Bearer $STARPORT_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"openai/gpt-4o-mini","messages":[{"role":"user","content":"Hello"}]}' \
http://127.0.0.1:8080/api/v1/chat/completionsA model in the catalog is not always a callable offering. The request succeeds only when the gateway can call a provider with a usable credential. Refer to Inspect the catalog for the difference.
For a gateway that keeps its data, read Run a persistent local gateway. The operator guide holds the full reference.