Skip to content

data-contracts.md

Latest commit

 

History

History
147 lines (109 loc) · 6.18 KB

data-contracts.md

File metadata and controls

147 lines (109 loc) · 6.18 KB

Data Contracts — Enforcing Snippet Schema & Payload Integrity

🧭 Quick Return to Map

You are in a sub-page of MemoryLongContext.
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.

RAG and long-context reasoning collapse quickly if payloads drift.
Data Contracts define the minimal schema that every retrieval, citation, and reasoning step must satisfy to remain auditable and reproducible.

When to use

  • Citations appear but fields like snippet_id or section_id are missing.
  • JSON payloads change shape across runs.
  • Downstream agents receive free-text with no schema lock.
  • Long-context sessions silently lose attribution or overwrite fields.
  • Multi-agent handoffs mutate keys or flatten nested fields.

Root causes

  • Loose schemas: free-form JSON without validation.
  • Field drift: different casing, missing offsets, or swapped names.
  • Silent truncation: long answers cut JSON blocks.
  • Inconsistent contracts: each component defines its own schema.
  • Opaque citations: only plain text without structured trace.

Core acceptance targets

  • Every snippet payload must include:
    {snippet_id, section_id, start_line, end_line, source_url, offsets, tokens}
  • Coverage ≥ 0.70 for the target section.
  • ΔS(question, retrieved) ≤ 0.45.
  • λ convergent across 3 paraphrases.
  • Contracts must validate under the same schema across sessions.

Structural fixes

  • Schema lock
    Define a JSON schema for citations and enforce validation at ingestion.

  • Contract inheritance
    Pass the same schema downstream to every agent and reasoning step.

  • Casing & normalization
    Enforce consistent field names and Unicode normalization.

  • Fail fast
    If schema validation fails, block reasoning and return fix instructions.

Fix in 60 seconds

  1. Define contract: 
    {
  "snippet_id": "string",
  "section_id": "string",
  "start_line": "int",
  "end_line": "int",
  "source_url": "string",
  "offsets": [int],
  "tokens": [string]
}

2. Validate every retrieval step against the contract.
3. Store the validated payload in the trace log.
4. Require cite-then-answer. Reject orphan claims.
5. Report ΔS and λ for each reasoning step.

---

## Copy-paste prompt

```
You have TXT OS and the WFGY Problem Map.

Task: enforce Data Contracts for retrieval and citation.

Protocol:
1. Validate that each snippet includes {snippet_id, section_id, start_line, end_line, source_url}.  
2. Reject orphans: if missing fields, stop and return fix tip.  
3. Require cite-then-answer.  
4. Log {ΔS(question,retrieved), λ_state, mem_rev, mem_hash}.  
5. Answer only with citations that pass contract validation.
```

---

## Common failure signals

* Citations alternate across runs → contract not enforced.
* JSON mode fails in provider → schema too loose.
* Free-text answers with no snippet\_id → orphan claims.
* Multi-agent pipelines mutate payloads → inconsistent contracts.

---

### 🔗 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 + \<your question>”   |
| **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” — OS boots instantly |

---

<!-- WFGY_FOOTER_START -->

### 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)

<!-- WFGY_FOOTER_END -->