vrr/WFGY
2
0
Fork 0
mirror of https://github.com/onestardao/WFGY.git synced 2026-04-26 10:40:55 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4
WFGY/ProblemMap/GlobalFixMap/VectorDBs_and_Stores/milvus.md

9.6 KiB
Raw Permalink Blame History

Milvus: Guardrails and Fix Patterns

🧭 Quick Return to Map

You are in a sub-page of VectorDBs_and_Stores.
To reorient, go back here:

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 field guide to stabilize Milvus when your RAG or agent stack loses accuracy. Use the checks below to localize failure, then jump to the exact WFGY fix page.

Open these first

Fix in 60 seconds

  1. Measure ΔS
    Compute ΔS(question, retrieved) and ΔS(retrieved, expected anchor).
    Thresholds: stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60.

  2. Probe with λ_observe
    Try k in {5, 10, 20}. Flat high curve suggests 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 to target section ≥ 0.70.
    ΔS ≤ 0.45 across three paraphrases.
    λ remains convergent. Logs and traces reproducible.

Typical breakpoints and the right repair

1) Distance metric does not match the encoder

  • Symptom: high similarity, wrong meaning.
  • Cause: collection metric set to L2 or IP while encoder expects cosine, or vice versa.
  • Fix: recreate index with the correct metric and re-ingest. Read Embedding ≠ Semantic.

2) Dimension drift after encoder swap

  • Symptom: insert errors or silent truncation through clients. Recall collapses on new data only.
  • Fix: confirm vector dim equals collection dim. If changed, create a new collection and backfill. See Vectorstore Fragmentation.

3) IVF index too shallow or poorly trained

  • Symptom: gold chunk appears only when k is very large.
  • Fix: train IVF with a large sample, raise nlist based on corpus size, sweep nprobe. Validate with reranker. See Retrieval Playbook and Rerankers.

4) HNSW underfit

  • Symptom: unstable top-k ordering, plateaued recall.
  • Fix: raise efSearch to 2–4×k and tune M. Validate with a reranker pass. See Retrieval Playbook.

5) Segments not compacted or index not built for fresh data

  • Symptom: new writes exist but never surface in results or search is slow after bulk load.
  • Fix: ensure flush and index build completed, then run compaction. Re-test ΔS and coverage on a canary set. See Live Monitoring for RAG.

6) Filter mismatch and payload type drift

  • Symptom: filtered searches return empty or unstable sets.
  • Fix: lock minimal metadata schema in Data Contracts. Validate types at ingestion.

7) Partitions or shards split the neighborhood

  • Symptom: good global recall, weak per-partition top-k.
  • Fix: consolidate or route by a stable key. Rebuild a single authoritative index. See Vectorstore Fragmentation.

8) Quantization harms recall

  • Symptom: fuzzy answers at small k after enabling PQ or scalar quant.
  • Fix: disable for quality checks or raise k and add a reranker. See Retrieval Playbook.

Observability probes

  • k-sweep curve: run k in 5, 10, 20 and plot ΔS. Flat high → metric or index fault.
  • Anchor control: compare ΔS against a golden anchor set. If only a collection or partition fails, rebuild that scope.
  • Hybrid toggle: vector only vs hybrid. If hybrid degrades, fix query parsing split and weights.
  • Reranker audit: with a strong reranker, recall should improve while ΔS falls. If not, rebuild.

Escalate when

  • ΔS stays above 0.60 for golden questions after metric and index corrections.
  • Coverage cannot reach 0.70 even with reranker and clean anchors.
  • Fresh writes are invisible after index build and compaction.

Open:

Copy-paste prompt for your AI


I uploaded TXT OS and the WFGY Problem Map files.

Target system: Milvus.

* symptom: \[brief]
* traces: ΔS(question,retrieved)=..., ΔS(retrieved,anchor)=..., λ states
* index: \[collection name, metric, index type, IVF nlist/nprobe or HNSW M/efSearch]
* encoder: \[model, dim, normalization, version]
* ingest: \[flush/index/compaction status, partitions, filters]

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 where relevant.

🔗 Quick-Start Downloads (60 sec)

Tool Link 3-Step Setup
WFGY 1.0 PDF Engine Paper 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>”
TXT OS (plain-text OS) TXTOS.txt 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly

Explore More

Layer Page What it’s for
⭐ Proof WFGY Recognition Map External citations, integrations, and ecosystem proof
⚙️ Engine WFGY 1.0 Original PDF tension engine and early logic sketch (legacy reference)
⚙️ Engine WFGY 2.0 Production tension kernel for RAG and agent systems
⚙️ Engine WFGY 3.0 TXT based Singularity tension engine (131 S class set)
🗺️ Map Problem Map 1.0 Flagship 16 problem RAG failure taxonomy and fix map
🗺️ Map Problem Map 2.0 Global Debug Card for RAG and agent pipeline diagnosis
🗺️ Map Problem Map 3.0 Global AI troubleshooting atlas and failure pattern map
🧰 App TXT OS .txt semantic OS with fast bootstrap
🧰 App Blah Blah Blah Abstract and paradox Q&A built on TXT OS
🧰 App Blur Blur Blur Text to image generation with semantic control
🏡 Onboarding Starter Village 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