Route food-label analysis through Cloud Run proxy#21
Open
eman-cickusic wants to merge 1 commit into
Open
Conversation
Analysis now works without a user-supplied LLM key: when no key is saved, requests go to the keyless Cloud Run /analyze proxy (Gemini via Vertex/ADC). A saved key still routes to the existing BYOK providers. - ProxyFoodLabelLlmWorkflow: one /analyze POST per scan, sliced across the 3 pipeline stages via a de-duplicated in-flight Deferred so the pipeline's per-stage timeout+retry joins the same call instead of issuing a second one (one scan = one call, correct token accounting). - SelectingFoodLabelLlmWorkflow: proxy when no key, BYOK when present. - PROXY_BASE_URL BuildConfig field (ZEST_PROXY_BASE_URL override). - Drop the "missing key" scanner banner/badge; scanning is keyless. - Result chat unchanged (still uses a personal key). Tests: parse, single-call/dedup, error mapping, and key-based routing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Routes the app's food-label analysis through the Cloud Run
/analyzeproxy so scanning works with no user-supplied LLM key. When no key is saved, requests go to the keyless proxy (Gemini via Vertex/ADC). A saved key still routes to the existing bring-your-own-key (Gemini/OpenAI/Grok/Groq) providers, so power users keep direct control. Result chat is unchanged (still uses a personal key).Why
The proxy (merged in #20, deployed to
b2-ultra-processed/us-east1) runs Gemini under the runtime service account, so users no longer need to paste an API key to analyze a label.How
ProxyFoodLabelLlmWorkflow— implements the existingFoodLabelLlmWorkflowinterface. The proxy returns nova + ingredients + allergens + usage in one response, so the singlePOST /analyzeis run once as a de-duplicated in-flightDeferred(keyed by the scan's ingredient text) and all three pipeline stages join it. Because the request runs in a workflow-owned scope, the pipeline's per-stagewithTimeout+ retry joins the same call instead of firing a second one — important for Cloud Run cold starts. Net: one scan = one/analyzecall, token usage counted once. Allergen cache-miss fails loudly (rescan) rather than returning a false "no allergens".SelectingFoodLabelLlmWorkflow— picks proxy (no key) vs BYOK (key present); re-read every call so toggling a key in Settings takes effect on the next scan.PROXY_BASE_URLBuildConfig field, default = the Cloud Run URL, overridable viaZEST_PROXY_BASE_URLinlocal.properties.hasApiKeyplumbing — scanning is keyless.Testing
./gradlew :app:testDebugUnitTest— 31 unit tests pass (6 new: response parsing, single-call across 3 stages, concurrent-stage dedup, upstream-error mapping, and key-based routing). androidTest compiles.Follow-up (not in this PR)
backend/main.pyCORS isallow_origins=["*"]— fine for the app, but worth locking down before broad web exposure.🤖 Generated with Claude Code