Version Packages

[neuromaxer]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@appx-org/agent-client@0.2.0

Patch Changes

• @appx-org/agent-protocol@0.2.0

@appx-org/agent-protocol@0.2.0

@appx-org/agent-server@0.2.0

Minor Changes

• 336a997: Breaking (agent-server): the LiteLLM model list now comes from one place — a
  JSON file named by LITELLM_MODELS_PATH.

  Configuring models previously meant choosing between three overlapping tiers
  (LITELLM_MODELS_JSON → LITELLM_MODELS csv → LITELLM_DEFAULT_MODEL alone),
  with four more env vars supplying the per-model values the csv tier could not
  express. The structured tier was the only one a control plane actually used, and
  it required cramming a multi-kilobyte JSON document onto a single env line —
  undiffable, uncommentable, and awkward to generate.

  LITELLM_MODELS_PATH points at a JSON array of the same entries. Everything else
  stays a scalar: LITELLM_BASE_URL, LITELLM_API_KEY, LITELLM_API,
  LITELLM_DEFAULT_MODEL.

  Removed, and now rejected at startup with a message naming the replacement
  rather than silently ignored:

  • LITELLM_MODELS_JSON → move the array into the file
  • LITELLM_MODELS → list the models in the file
  • LITELLM_CONTEXT_WINDOW, LITELLM_MAX_TOKENS, LITELLM_REASONING → per-model
    contextWindow / maxTokens / reasoning
  • LITELLM_COMPAT_JSON → per-model compat
  • LITELLM_DEFAULT_THINKING → defaultThinkingLevel on the default entry

  Rejecting is the point: a stale value would leave the provider unregistered,
  which presents as "the agent has no models" with nothing indicating why. The
  other failure modes are loud for the same reason — an unreadable path, malformed
  JSON, a non-array, or an entry without an id all throw at startup.

  modelPreset() is unchanged and still supplies the dialect for openai/gpt-5.5
  and the DeepSeek models (thinkingLevelMap, defaultThinkingLevel,
  compat.thinkingFormat). File entries override presets field by field, so a bare
  {"id": "openai/gpt-5.5"} still gets the full preset — including DeepSeek's
  null mappings, which mark levels the model cannot do and drive clamping.

  Adds 22 tests covering file loading, default-model selection, preset precedence,
  every failure mode, and each removed variable.

Patch Changes

• @appx-org/agent-protocol@0.2.0