# Production Embedding Semantic QA Receipt Status: `needs_review` Created: `2026-06-15T04:31:21Z` Model target: `gemini-embedding-2` at `768` dimensions ## Summary This receipt spot-checks clean semantic chunks before live Gemini embedding. It checks source references, derived JD context boundaries, proof-hash boundaries, legacy review-gate field absence, and whether the selected rows are more than label-only text. ## Metrics - Sample rows: `50` - Passing rows: `45` - Failing rows: `5` - Raw JD key hits in embedding text: `0` - Legacy review-gate field hits: `0` ## Source Families | Source family | Sample rows | | --- | ---: | | workflow_seed | 50 | ## Checks - FAIL `all_sample_rows_pass` - PASS `no_legacy_review_gate_fields` - PASS `no_raw_jd_key_hits` - PASS `requested_families_covered` - PASS `sample_limit_respected` - PASS `sample_rows_present` --- Ali Mehdi Mukadam - co-authored with Codex