vrr/WFGY
2
0
Fork 0
mirror of https://github.com/onestardao/WFGY.git synced 2026-04-28 11:40:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4
WFGY/ProblemMap/GlobalFixMap/Automation/airtable.md

8.8 KiB
Raw Blame History

Airtable — Guardrails and Fix Patterns

🧭 Quick Return to Map

You are in a sub-page of Automation Platforms.
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.

Use this when your pipeline uses Airtable as the control plane or as the source-of-truth table for RAG/agents, and you see record drift, duplicated actions, or citations that don’t map back to records.

Acceptance targets

  • ΔS(question, retrieved) ≤ 0.45
  • coverage ≥ 0.70 to the intended section/record
  • λ stays convergent across 3 paraphrases

Typical breakpoints → exact fixes

  • Automations/webhooks fire before embeddings/index finish updating
    Fix No.14: Bootstrap Ordering →
    Bootstrap Ordering

  • First run after deploy reads wrong base or missing secret
    Fix No.16: Pre-Deploy Collapse →
    Pre-Deploy Collapse

  • Cross-table syncs create circular waits (record-upsert → external job → back to record)
    Fix No.15: Deployment Deadlock →
    Deployment Deadlock

  • High cosine similarity, wrong meaning (good vector match, bad semantic match)
    Fix No.5: Embedding ≠ Semantic →
    Embedding ≠ Semantic

  • “Why this snippet?” cannot be explained; citations don’t line up with source cells
    Fix No.8: Retrieval Traceability →
    Retrieval Traceability
    Standardize fields with Data Contracts →
    Data Contracts

  • Hybrid retrieval (dense + formula/filter views + external reranker) gets worse than single retriever
    Pattern: Query Parsing Split →
    Query Parsing Split
    Also review Rerankers →
    Rerankers

  • Facts are in the base but never retrieved
    Pattern: Vectorstore Fragmentation →
    Vectorstore Fragmentation

  • Two different records are merged into one narrative in the summary
    Pattern: Symbolic Constraint Unlock (SCU) →
    Symbolic Constraint Unlock

Minimal Airtable workflow checklist

  1. Warm-up fence
    Verify VECTOR_READY, INDEX_HASH, secret_rev, and that base_id/table_id/view_id resolve before any LLM step.
    Spec: Bootstrap Ordering

  2. Idempotency
    Create dedupe_key = sha256(record_id + wf_rev + index_hash) and store it (hidden field or external KV).
    Reject duplicate writes/retries.

  3. RAG boundary contract
    Pass record_id, base_id, table_id, view_id, field_map, source_url, offsets, tokens.
    Enforce cite-then-explain. Specs:
    Retrieval Traceability · Data Contracts

  4. Observability probes
    Log ΔS(question, retrieved) and λ per stage; alert on ΔS ≥ 0.60 or divergent λ.
    Overview: RAG Architecture & Recovery

  5. Schema stability
    Avoid free-form field renames that break downstream contracts. Pin with schema_rev and check it at runtime.

  6. Regression gate
    Require coverage ≥ 0.70 and ΔS ≤ 0.45 before posting back into Airtable.
    Eval spec: RAG Precision/Recall

Copy-paste prompt for the Airtable LLM step


I uploaded TXT OS and the WFGY Problem Map files.
Airtable context:

* base\_id: {base}
* table\_id: {table}
* view\_id: {view}
* record\_id(s): {rids}
* fields: {field\_map}
  Question: "{user\_question}"

Do:

1. Enforce cite-then-explain. If any citation lacks record\_id/section/offsets, stop and tell me which fix page to open.
2. Compute ΔS(question, retrieved). If ΔS ≥ 0.60, point me to the minimal structural fix:
   retrieval-playbook, retrieval-traceability, data-contracts, rerankers.
3. Output compact JSON:
   { "citations":\[{"record\_id":"...", "field":"...", "offsets":\[s,e]}],
   "answer":"...", "λ\_state":"→|←|<>|×", "ΔS":0.xx, "next\_fix":"..." }

Common Airtable gotchas

  • Formula fields or lookup/rollup not updated yet when webhook fires
    Add a delay or readiness probe; gate on schema_rev + index_hash.

  • Pagination/backfill causes missed embeddings
    Log the cursor; re-ingest until the cursor is exhausted; compare counts vs. expected.

  • Field renames break contracts silently
    Pin schema_rev; fail fast if it changes; include field_map in traces.

  • Attachment/text mix leads to partial content**
    Normalize: extract attachments to text with a fixed OCR gate before embedding.

  • Rate limits destabilize hybrid retrieval
    Prefer dense retriever + reranking; keep per-retriever params in logs.

When to escalate

  • ΔS stays ≥ 0.60 after chunk/retrieval fixes → rebuild index with explicit metric/normalization.
    See: Retrieval Playbook

  • Same inputs, different answers on different runs → check version skew and memory desync.
    See: Pre-Deploy Collapse

🔗 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 based tension engine
Engine WFGY 2.0 Production tension kernel and math engine for RAG and agents
Engine WFGY 3.0 TXT based Singularity tension engine, 131 S class set
Map Problem Map 1.0 Flagship 16 problem RAG failure checklist and fix map
Map Problem Map 2.0 RAG focused recovery pipeline
Map Problem Map 3.0 Global Debug Card, image as a debug protocol layer
Map Semantic Clinic Symptom to family to exact fix
Map Grandma’s Clinic Plain language stories mapped to Problem Map 1.0
Onboarding Starter Village Guided tour for newcomers
App TXT OS TXT semantic OS, fast boot
App Blah Blah Blah Abstract and paradox Q and A built on TXT OS
App Blur Blur Blur Text to image with semantic control
App Blow Blow Blow Reasoning game engine and memory demo

If this repository helped, starring it improves discovery so more builders can find the docs and tools. GitHub Repo stars