📑 Data Contracts — Stable Interfaces for RAG & Agents

Everything WFGY touches is JSON-first and versioned. These “contracts” make pipelines observable, reproducible, and easy to debug.

Quick Nav
Retrieval Playbook · Traceability · Eval · Ops · Patterns: SCU · Memory Desync

0) Envelope (required for all records)

{
  "schema_version": "1.0.0",
  "event": "ingest.write | retrieve.run | rerank.run | answer.decide",
  "ts": "2025-08-13T10:22:59Z",
  "trace_id": "uuid",
  "agent_id": "scout|medic|engineer|retriever|system",
  "mem_rev": "r42", 
  "mem_hash": "sha256:..."
}

mem_rev/mem_hash prevent memory overwrite and desync.
Use the same envelope for logs and datasets.

1) Chunk record

Purpose: atomic, traceable text unit for retrieval.

{
  "$schema": "https://wfgy.dev/schemas/chunk-1.0.json",
  "chunk_id": "c_00123",
  "doc_id": "d_wfgy_paper",
  "section_id": "s_intro",
  "span": {"line_start": 120, "line_end": 154},
  "lang": "en",
  "text": "Delta-S measures semantic stress ...",
  "hash": "sha256:...",
  "embedding": {
    "model": "sentence-transformers/all-MiniLM-L6-v2",
    "dim": 384,
    "vector": [0.012, -0.044, ...],
    "normalized": true,
    "metric": "cosine"
  },
  "anchors": ["ΔS", "semantic stress"]
}

Rules

Keep original text + normalized text (case/punctuation) if you apply normalization downstream.
Always store metric and normalized.

2) Query record

{
  "$schema": "https://wfgy.dev/schemas/query-1.0.json",
  "q_id": "q_2025_08_13_0001",
  "text": "How does ΔS detect retrieval failure?",
  "lang": "en",
  "hyde": "Generate a canonical query about ... (optional)",
  "tokens": {"count": 12, "analyzer": "icu"},
  "hints": {"doc_id": ["d_wfgy_paper"], "section_id": []}
}

3) Retrieval result (candidate)

{
  "$schema": "https://wfgy.dev/schemas/retrieved-1.0.json",
  "q_id": "q_2025_08_13_0001",
  "ranker": "dense|bm25|hybrid",
  "k": 50,
  "items": [
    {
      "chunk_id": "c_00123",
      "doc_id": "d_wfgy_paper",
      "score": 0.812,           // retriever-native score
      "cosine": 0.91,           // optional explicit cosine
      "ΔS_q_ctx": 0.36,         // optional, if ground anchor available
      "source": "dense",
      "features": {"bm25": 8.3, "dense": 0.91}
    }
  ]
}

4) Rerank result

{
  "$schema": "https://wfgy.dev/schemas/rerank-1.0.json",
  "q_id": "q_2025_08_13_0001",
  "model": "BAAI/bge-reranker-base",
  "k_in": 60,
  "k_out": 8,
  "items": [
    {
      "chunk_id": "c_00123",
      "pre_score": {"dense": 0.91, "bm25": 8.3},
      "post_score": {"ce": 0.82},
      "reason": "mentions ΔS definition and failure threshold",
      "selected": true
    }
  ]
}

5) Prompt frame (schema-locked)

{
  "$schema": "https://wfgy.dev/schemas/prompt-frame-1.0.json",
  "system": "You are a grounded assistant. Cite before you explain.",
  "task": "Answer the user's question using ONLY the cited snippets.",
  "constraints": ["No cross-source merging", "Cite line spans"],
  "citations": [
    {"id": "c_00123", "doc_id": "d_wfgy_paper", "section_id": "s_intro", "span": [120,154]}
  ],
  "question": "How does ΔS detect retrieval failure?"
}

6) Answer + trace

{
  "$schema": "https://wfgy.dev/schemas/answer-1.0.json",
  "q_id": "q_2025_08_13_0001",
  "answer_text": "ΔS measures the semantic gap ...",
  "cited_chunks": ["c_00123", "c_00987"],
  "λ_observe": "→",
  "metrics": {"ΔS_q_ctx": 0.38, "latency_ms": 922}
}

7) Metrics pack (for CI)

{
  "$schema": "https://wfgy.dev/schemas/metrics-1.0.json",
  "dataset": "goldset_v1",
  "recall@50": 0.91,
  "nDCG@10": 0.62,
  "ΔS_mean": 0.41,
  "ΔS_p95": 0.58,
  "λ_convergent_rate": 0.82
}

Acceptance checklist

✅ All records include envelope (schema_version, event, ts, trace_id, mem_rev/hash).
✅ Chunks persist metric and normalized flags.
✅ Prompts are schema-locked (cite → explain).
✅ Answers store cited chunk IDs and λ state.
✅ Metrics committed per PR (goldset.jsonl).

🔗 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.

6.6 KiB Raw Blame History Unescape Escape