Repository navigation
NutriTrace v1.0.4: GPT-5.6 support, OpenAI-compat vision fix, Fitbit deprecation #117
thebigjoe1
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Just released. Signed APK + Docker image are up on the release page: https://github.com/TraceApps/nutritrace/releases/tag/v1.0.4
Patch release. Includes two community-reported AI-provider compatibility fixes and one deprecation gate ahead of Fitbit's Web API wind-down.
Changed
Fixed
OpenAI-compatible endpoints accept vision (image) chat again. Env-locked deployments running the AI proxy against a strict OpenAI-schema endpoint (LiteLLM in front of Bedrock, etc.) were rejecting image messages with
invalid content type=image. The client now emits OpenAI-shape image parts foroai-compat, and the proxy defensively normalises any Anthropic-shape image parts toimage_urlat the boundary before forwarding. Also mirrored to CookTrace and LiftTrace since they share the same proxy code. Thanks to @javydekoning for reporting in #114.GPT-5.6-era chat parameters supported. OpenAI's newer chat models require
max_completion_tokensinstead of the deprecatedmax_tokens, and reject function-tool requests unlessreasoning_effort: 'none'is set. Both the client and server AI paths now branch on the model and base URL to emit the correct parameters. Third-party OpenAI-compatible endpoints (Ollama, LM Studio, LiteLLM proxies) keep receivingmax_tokenssince most have not adopted the new field yet. Backward-compatible with older OpenAI models. Thanks to @librarian for contributing the fix in #115.Signed APK on the release page. Docker image is multi-arch (amd64 + arm64):
docker compose pull && docker compose up -dAll reactions