Skip to content

providers: verify LM Studio and llama-server end to end, and write down the exact lines #5

Description

@morningstarnasser

The catalogue carries lm-studio and a custom OpenAI-compatible entry, and the code path is the same one Ollama and llmman use — but nobody has put a real LM Studio or a bare llama-server behind Talos and written down what it took.

What this is. Run Talos against each: TALOS_MODEL_PROVIDER=lm-studio (default http://localhost:1234/v1) and custom with TALOS_BASE_URL_CUSTOM=… for llama-server. Note what worked, what needed a line in talos.env, whether the readiness probe (GET /models) found the model under the name you configured, and whether a thinking model answers in time (a small reasoning model can spend the whole 180 s budget thinking — say so if it did). Then a short section in README.md under "Install" with the exact lines, and a doctor hint if something was non-obvious.

Boundaries. Docs, doctor.py, at most a catalogue field (catalog.py). If a code path turns out to be wrong, open a separate issue with the measurement rather than fixing it in the docs PR.

Proof. The transcript of one real talos ask per provider, in the PR.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentationgood first issueGood for newcomershelp wantedExtra attention is needed

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions