User documentation. If you are running Tempo9 rather than building it, this is the directory you want.
| Page | Read it when |
|---|---|
| Getting started | You have a model file and want a server |
| Models and quantizations | You want to know if YOUR file will run |
| Connect an app | You have an app and want it to use Tempo9 |
| Ollama models | You already pulled models with Ollama |
| CLI reference | You want every flag |
| API reference | You are writing a client |
| Limits | Before you design around something we do not do |
Engineering notes — investigations, benchmark method, internal design — live
in ../docs/ and are not user documentation. The split is
deliberate: a page that explains how to connect an app and a page that
records why a crash happened have different readers, and mixing them makes
both harder to find.
- Ollama models
2026-08-31 · Run the models Ollama already pulled, with no download and no copy.
- Models and quantizations
2026-08-31 · Which GGUF quantizations and architectures run, verified by serving them.
- Limits
2026-08-31 · What Tempo9 does not do, before you design around it.
- Getting started
2026-08-31 · Run a GGUF model as a local server in one command.
- Connect an app
2026-08-31 · Point any OpenAI-compatible or Ollama-compatible app at Tempo9.
- CLI reference
2026-08-31 · Every tempo9 flag and what it defaults to.
- API reference
2026-08-31 · Every route: OpenAI, Anthropic, Responses, and Ollama native.