Repository navigation
fix(ai): move the schedule proxy from GitHub Models to Groq - #1
Merged
Merged
Conversation
GitHub Models is fully retired on 2026-07-30, so AI Jadwal would have gone quiet on its own. Worker upstream becomes Groq, model is pinned to openai/gpt-oss-120b, and the secret becomes GROQ_API_KEY. The Firebase ID token check, the CORS allowlist and the server-side model pin are unchanged, since Groq is OpenAI-compatible. Model choice was tested, not assumed: llama-3.3-70b-versatile and qwen/qwen3.6-27b both reject response_format json_schema outright. Of the two that accept it, 120b resolved a relative date correctly where 20b returned a Saturday. Reminder offsets are strings with an enum now, and that is deliberate. Groq's constrained decoder drops the separator between elements of an unconstrained integer array, so asking for two reminders returned [144060] instead of [1440, 60], which normalizeEvent then snapped to a single 1440 and the second reminder disappeared without a trace. Correct on every run with an enum of strings, and correct with no response_format at all, so the fault is the decoder rather than the model. Adds worker/verify-upstream.sh, which hits Groq directly and asserts the schema is accepted, the relative date is right and the reminders are not concatenated. Also banners docs/PLAN.md and docs/DEPLOY.md, which still describe the Next.js on Vercel with Turso architecture that was abandoned.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
One commit. The AI scheduling proxy pointed at GitHub Models, which needs a PAT with
models:read. It now runs on Groq.Includes
worker/verify-upstream.shso the upstream can be probed without deploying, plus the wrangler config, README, and docs updated to match.Verified:
npm run buildin web/ passes (tsc -b + vite build). Fast-forward onto main, no conflicts.Note: the worker still needs
GROQ_API_KEYset as a Wrangler secret before the deployed AI feature works.