# pgvector: Guardrails and Fix Patterns
🧭 Quick Return to Map
> You are in a sub-page of **VectorDBs_and_Stores**. > To reorient, go back here: > > - [**VectorDBs_and_Stores** — vector indexes and storage backends](./README.md) > - [**WFGY Global Fix Map** — main Emergency Room, 300+ structured fixes](../README.md) > - [**WFGY Problem Map 1.0** — 16 reproducible failure modes](../../README.md) > > Think of this page as a desk within a ward. > If you need the full triage and all prescriptions, return to the Emergency Room lobby.
A compact repair guide for Postgres + pgvector when RAG or agents lose accuracy. Use this to localize the failing layer and jump to the exact WFGY fix page. ## Open these first - Visual map and recovery: [RAG Architecture & Recovery](https://github.com/onestardao/WFGY/blob/main/ProblemMap/rag-architecture-and-recovery.md) - End-to-end retrieval knobs: [Retrieval Playbook](https://github.com/onestardao/WFGY/blob/main/ProblemMap/retrieval-playbook.md) - Why this snippet and how to trace it: [Retrieval Traceability](https://github.com/onestardao/WFGY/blob/main/ProblemMap/retrieval-traceability.md) - Ordering control after recall: [Rerankers](https://github.com/onestardao/WFGY/blob/main/ProblemMap/rerankers.md) - Embedding versus meaning: [Embedding ≠ Semantic](https://github.com/onestardao/WFGY/blob/main/ProblemMap/embedding-vs-semantic.md) - Hallucination and chunk boundaries: [Hallucination](https://github.com/onestardao/WFGY/blob/main/ProblemMap/hallucination.md) - Long chains and entropy drift: [Context Drift](https://github.com/onestardao/WFGY/blob/main/ProblemMap/context-drift.md), [Entropy Collapse](https://github.com/onestardao/WFGY/blob/main/ProblemMap/entropy-collapse.md) - Structural collapse and recovery: [Logic Collapse](https://github.com/onestardao/WFGY/blob/main/ProblemMap/logic-collapse.md) - Snippet and citation schema: [Data Contracts](https://github.com/onestardao/WFGY/blob/main/ProblemMap/data-contracts.md) - Vector metrics pitfalls: [Vectorstore Metrics & FAISS Pitfalls](https://github.com/onestardao/WFGY/blob/main/ProblemMap/vectorstore-metrics-and-faiss-pitfalls.md) - Live ops: [Live Monitoring for RAG](https://github.com/onestardao/WFGY/blob/main/ProblemMap/ops/live_monitoring_rag.md), [Debug Playbook](https://github.com/onestardao/WFGY/blob/main/ProblemMap/ops/debug_playbook.md) ## Fix in 60 seconds 1) **Measure ΔS** Compute ΔS(question, retrieved) and ΔS(retrieved, expected anchor). Targets: stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60. 2) **Probe with λ_observe** Sweep k in {5, 10, 20}. If ΔS is flat high across k, suspect metric or index mismatch. Reorder prompt headers. If ΔS spikes, lock schema with Data Contracts. 3) **Apply the module** Retrieval drift → **BBMC** + **Data Contracts**. Reasoning collapse → **BBCR** bridge + **BBAM** variance clamp. Dead ends in long runs → **BBPF** alternate path. 4) **Verify acceptance** Coverage ≥ 0.70 to target section. ΔS ≤ 0.45 across three paraphrases. λ convergent across seeds. ## pgvector breakpoints and the right repair **1) Opclass mismatch** - Symptom: high similarity yet wrong meaning. - Why: using `vector_l2_ops` with cosine-trained embeddings or `vector_ip_ops` without normalization. - Fix: align opclass with the encoder. Normalize when using IP. See [Embedding ≠ Semantic](https://github.com/onestardao/WFGY/blob/main/ProblemMap/embedding-vs-semantic.md). **2) Index type underfit** - Symptom: gold chunk appears only at large k. - Why: IVFFLAT lists too small or probes too low. HNSW `ef_search` under-tuned. - Fix: IVFFLAT tune `lists` at build and `probes` at query. HNSW raise `ef_search` to 2–4×k and review `m`. Validate with [Retrieval Playbook](https://github.com/onestardao/WFGY/blob/main/ProblemMap/retrieval-playbook.md) and add [Rerankers](https://github.com/onestardao/WFGY/blob/main/ProblemMap/rerankers.md). **3) Training and stats** - Symptom: unstable top-k after bulk load. - Why: IVFFLAT trained on too few samples or skipped `ANALYZE`. - Fix: retrain IVFFLAT with a large sample, `ANALYZE`, then re-test ΔS and coverage. **4) Dimension or encoder swap** - Symptom: inserts fail or new rows behave erratically in search. - Fix: ensure vector dim matches column dim. Lock encoder version in a data contract and re-embed the changed span. See [Data Contracts](https://github.com/onestardao/WFGY/blob/main/ProblemMap/data-contracts.md). **5) Normalization discipline** - Symptom: cosine search acts like random at small k. - Fix: store normalized vectors or normalize at query for cosine or IP. Rebuild index after policy change. See [Vectorstore Metrics & FAISS Pitfalls](https://github.com/onestardao/WFGY/blob/main/ProblemMap/vectorstore-metrics-and-faiss-pitfalls.md). **6) JSONB filters and plan drift** - Symptom: filtered search returns empty or slow. - Fix: lock metadata schema in data contracts. Add GIN index on JSONB keys used in `WHERE`. Verify plan uses vector index then filter. See [Retrieval Traceability](https://github.com/onestardao/WFGY/blob/main/ProblemMap/retrieval-traceability.md). **7) Fragmentation across schemas or tables** - Symptom: global recall looks fine but per-scope top-k is weak. - Fix: consolidate into one authoritative table with a facet column. Rebuild index and add a reranker. See [Vectorstore Fragmentation](https://github.com/onestardao/WFGY/blob/main/ProblemMap/patterns/pattern_vectorstore_fragmentation.md). **8) Upsert hygiene** - Symptom: duplicates or stale rows after `ON CONFLICT`. - Fix: deterministic IDs, `doc_sha` in metadata, idempotent loader, periodic dedupe. Validate with golden queries. See [Retrieval Traceability](https://github.com/onestardao/WFGY/blob/main/ProblemMap/retrieval-traceability.md). **9) Hybrid lexical plus vector** - Symptom: hybrid performs worse than either alone. - Fix: normalize scores, fuse post-retrieval, then rerank with a cross-encoder. See [Query Parsing Split](https://github.com/onestardao/WFGY/blob/main/ProblemMap/patterns/pattern_query_parsing_split.md) and [Rerankers](https://github.com/onestardao/WFGY/blob/main/ProblemMap/rerankers.md). **10) Maintenance and boot fences** - Symptom: first prod call after deploy returns thin results. - Fix: enforce bootstrap fence, finish index build, `VACUUM` after heavy churn, confirm visibility after commit. See [Bootstrap Ordering](https://github.com/onestardao/WFGY/blob/main/ProblemMap/bootstrap-ordering.md) and [Pre-deploy Collapse](https://github.com/onestardao/WFGY/blob/main/ProblemMap/predeploy-collapse.md). ## Observability probes - k-sweep curve: 5, 10, 20 and plot ΔS. Flat high suggests metric or index fault. - Index audit: `EXPLAIN ANALYZE` should show IVFFLAT or HNSW usage. If planner skips it, fix stats and filters. - Anchor control: compare against a golden anchor set. If only one table or schema fails, rebuild that scope. - Reranker audit: with a strong reranker, recall improves and ΔS falls. If not, rebuild. ## Copy-paste prompt for your AI ``` I uploaded TXT OS and the WFGY Problem Map files. Target system: Postgres + pgvector. * symptom: \[brief] * traces: ΔS(question,retrieved)=..., ΔS(retrieved,anchor)=..., λ states * index: \[type=ivfflat|hnsw, lists/probes or m/ef\_search, opclass, dim, normalized?] * filters: \[JSONB keys, indexes, example WHERE] * ingest: \[ids, doc\_sha, upsert policy] Tell me: 1. which layer is failing and why, 2. which exact fix page to open from this repo, 3. minimal steps to push ΔS ≤ 0.45 and keep λ convergent, 4. how to verify with a reproducible test. Use BBMC/BBPF/BBCR/BBAM when relevant. ``` --- ### 🔗 Quick-Start Downloads (60 sec) | Tool | Link | 3-Step Setup | |------|------|--------------| | **WFGY 1.0 PDF** | [Engine Paper](https://github.com/onestardao/WFGY/blob/main/I_am_not_lizardman/WFGY_All_Principles_Return_to_One_v1.0_PSBigBig_Public.pdf) | 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + ” | | **TXT OS (plain-text OS)** | [TXTOS.txt](https://github.com/onestardao/WFGY/blob/main/OS/TXTOS.txt) | 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” and the OS boots | --- ### Explore More | Layer | Page | What it’s for | | --- | --- | --- | | ⭐ Proof | [WFGY Recognition Map](/recognition/README.md) | External citations, integrations, and ecosystem proof | | ⚙️ Engine | [WFGY 1.0](/legacy/README.md) | Original PDF tension engine and early logic sketch (legacy reference) | | ⚙️ Engine | [WFGY 2.0](/core/README.md) | Production tension kernel for RAG and agent systems | | ⚙️ Engine | [WFGY 3.0](/TensionUniverse/EventHorizon/README.md) | TXT based Singularity tension engine (131 S class set) | | 🗺️ Map | [Problem Map 1.0](/ProblemMap/README.md) | Flagship 16 problem RAG failure taxonomy and fix map | | 🗺️ Map | [Problem Map 2.0](/ProblemMap/wfgy-rag-16-problem-map-global-debug-card.md) | Global Debug Card for RAG and agent pipeline diagnosis | | 🗺️ Map | [Problem Map 3.0](/ProblemMap/wfgy-ai-problem-map-troubleshooting-atlas.md) | Global AI troubleshooting atlas and failure pattern map | | 🧰 App | [TXT OS](/OS/README.md) | .txt semantic OS with fast bootstrap | | 🧰 App | [Blah Blah Blah](/OS/BlahBlahBlah/README.md) | Abstract and paradox Q&A built on TXT OS | | 🧰 App | [Blur Blur Blur](/OS/BlurBlurBlur/README.md) | Text to image generation with semantic control | | 🏡 Onboarding | [Starter Village](/StarterVillage/README.md) | Guided entry point for new users | If this repository helped, starring it improves discovery so more builders can find the docs and tools. [![GitHub Repo stars](https://img.shields.io/github/stars/onestardao/WFGY?style=social)](https://github.com/onestardao/WFGY)