diff --git a/TODO.md b/TODO.md index d06cd5f..263c6ef 100644 --- a/TODO.md +++ b/TODO.md @@ -233,6 +233,7 @@ - The result modal should now stay within viewport height, but this still needs real browser confirmation on multiple short-height displays because CSS-only constraints were the source of the latest user-visible regression. - Artgrid preview playback now has a server-side ffmpeg transcode path for `.m3u8` style preview URLs, but this trades storage savings for runtime CPU cost. - The reverted `5ca7aef` experiment showed that simply widening collector caps and Gemini candidate count can backfire when the added candidates are weak: final visible count fell sharply even though backend raw candidate count increased. +- Gemini Vision batch execution can still see mixed failure modes: true model/output-format failures and ignorable low-value/no-visual candidate skips. These should not be conflated in user-facing warnings. - The local self-test script is better than before, but it is still a smoke test, not full integration coverage. ## Current Risks Around Search Quality @@ -554,6 +555,7 @@ ## Highest-Value Next Steps - [ ] Reduce `/api/search` latency further without collapsing result count - [ ] Rebuild the reverted search-expansion work from the previous stable baseline, but only after measuring where candidate quality collapses between ranked pool and final merge +- [ ] Continue hardening Gemini response parsing so truncated JSON or mixed-quality candidate batches degrade more gracefully without surfacing alarming warnings - [ ] Build a repeatable repo-local bootstrap script or documented setup command set for non-root machines so fresh PC setup does not depend on shell history - [ ] Improve Envato / Artgrid preview acquisition reliability so Gemini Vision sees real frames more often - [ ] Browser-verify the new result modal at multiple viewport heights and confirm translated Source Summary readability on real long descriptions @@ -624,6 +626,25 @@ - If behavior in the browser does not match the latest backend/frontend code, the first assumption should be stale frontend assets until proven otherwise ## Recent Change Log +- Date: `2026-03-17` +- What changed: + - Reduced Gemini Vision batch size from `8` to `6` candidates to lower the chance of oversized/truncated JSON responses. + - Added an explicit Gemini output token cap for the vision JSON response path. + - Changed Gemini batch failure accounting so ignorable per-candidate skips such as low-value thumbnails or missing visuals no longer count as a user-facing partial batch failure when useful recommendations were still recovered. + - Added unit coverage for the hard-error filtering behavior. +- Why it changed: + - The user reported the warning `gemini vision partially failed on 2 of 2 batches`. + - The provided log `ai-media-hub-2026-03-17T04-35-33-028Z.log` showed two distinct failure classes during the second repeated search: + - a real batch-level JSON extraction failure caused by truncated Gemini output + - multiple ignorable sequential-retry skips caused by low-value thumbnails or missing visuals + - Those were being combined into an alarming partial-failure warning even though useful recommendations were still recovered. +- How it was verified: + - `go test ./...` + - `bash scripts/selftest.sh` +- What is still risky or incomplete: + - This reduces false-positive partial-failure warnings, but true Gemini output truncation can still happen occasionally if model behavior changes again. + - Further hardening may still be needed if Gemini starts returning different malformed JSON shapes. + - Date: `2026-03-17` - What changed: - Reverted commit `5ca7aef` (`Strengthen search breadth and modal fitting`) to restore the previous stable search/modal baseline. diff --git a/backend/services/gemini.go b/backend/services/gemini.go index c61894b..1d79df9 100644 --- a/backend/services/gemini.go +++ b/backend/services/gemini.go @@ -306,6 +306,7 @@ User query: ` + query, }, "generationConfig": map[string]any{ "responseMimeType": "application/json", + "maxOutputTokens": 1400, }, } diff --git a/backend/services/gemini_test.go b/backend/services/gemini_test.go index 0004f73..2f26575 100644 --- a/backend/services/gemini_test.go +++ b/backend/services/gemini_test.go @@ -213,3 +213,18 @@ func TestMergeRecommendationsExcludesIrrelevantAndPendingFiller(t *testing.T) { t.Fatalf("unexpected merged result: %#v", merged) } } + +func TestFilterHardGeminiErrorsIgnoresLowValueVisualFailures(t *testing.T) { + errs := []string{ + "candidate thumbnail is low value", + "no candidate thumbnails or preview frames could be fetched for gemini vision", + "gemini vision JSON extraction failed: no complete JSON object found", + } + filtered := filterHardGeminiErrors(errs) + if len(filtered) != 1 { + t.Fatalf("expected only hard errors to remain, got %#v", filtered) + } + if !strings.Contains(filtered[0], "JSON extraction failed") { + t.Fatalf("unexpected filtered errors: %#v", filtered) + } +} diff --git a/backend/services/ranker.go b/backend/services/ranker.go index 919d0bd..12b2470 100644 --- a/backend/services/ranker.go +++ b/backend/services/ranker.go @@ -105,7 +105,7 @@ func EvaluateAllCandidatesWithGemini(service *GeminiService, query string, ranke } func EvaluateAllCandidatesWithGeminiWithDeadline(service *GeminiService, query string, ranked []SearchResult, deadline time.Time) ([]AIRecommendation, GeminiBatchStats, error) { - const chunkSize = 8 + const chunkSize = 6 const maxConcurrentBatches = 2 if service == nil { return nil, GeminiBatchStats{}, fmt.Errorf("gemini service is not configured") @@ -187,6 +187,7 @@ func EvaluateAllCandidatesWithGeminiWithDeadline(service *GeminiService, query s }) } recovered, recoveredErrs := recoverGeminiBatchSequentially(service, query, ranked, batch.index*chunkSize) + hardErrs := filterHardGeminiErrors(recoveredErrs) if len(recovered) > 0 { stats.SequentialRetried++ stats.Succeeded++ @@ -197,9 +198,9 @@ func EvaluateAllCandidatesWithGeminiWithDeadline(service *GeminiService, query s seen[item.Link] = true merged = append(merged, item) } - if len(recoveredErrs) > 0 { + if len(hardErrs) > 0 { stats.Failed++ - for _, recoveredErr := range recoveredErrs { + for _, recoveredErr := range hardErrs { if len(stats.Errors) < 5 { stats.Errors = append(stats.Errors, recoveredErr) } @@ -207,6 +208,9 @@ func EvaluateAllCandidatesWithGeminiWithDeadline(service *GeminiService, query s } continue } + if len(hardErrs) == 0 { + continue + } stats.Failed++ if len(stats.Errors) < 5 { stats.Errors = append(stats.Errors, batch.err.Error()) @@ -329,6 +333,35 @@ func MergeUniqueRecommendations(base, extra []AIRecommendation) []AIRecommendati return merged } +func filterHardGeminiErrors(errs []string) []string { + filtered := make([]string, 0, len(errs)) + for _, item := range errs { + if isIgnorableGeminiError(item) { + continue + } + filtered = append(filtered, item) + } + return filtered +} + +func isIgnorableGeminiError(message string) bool { + lower := strings.ToLower(strings.TrimSpace(message)) + if lower == "" { + return false + } + for _, token := range []string{ + "no candidate thumbnails or preview frames could be fetched for gemini vision", + "candidate thumbnail is low value", + "candidate has no thumbnail or preview video", + "image url is empty", + } { + if strings.Contains(lower, token) { + return true + } + } + return false +} + func MergeGeminiBatchStats(base, extra GeminiBatchStats) GeminiBatchStats { merged := base merged.CandidateCap += extra.CandidateCap