Skip to content

Latest commit

 

History

History
176 lines (129 loc) · 10.3 KB

File metadata and controls

176 lines (129 loc) · 10.3 KB

DeepSeek — Guardrails and Fix Patterns

🧭 Quick Return to Map

You are in a sub-page of LLM_Providers.
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 page when failures look provider specific on DeepSeek models. Examples include JSON tool-call drift, unexpected safety blocks, long reasoning preambles that leak into the final channel, or unstable answers across seeds. Each fix maps to WFGY pages so you can verify with measurable targets.

Core acceptance

  • ΔS(question, retrieved) ≤ 0.45
  • Coverage ≥ 0.70 for the target section
  • λ remains convergent across 3 paraphrases

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.
  1. Probe with λ_observe
  • Vary k ∈ {5, 10, 20}. Flat high curve suggests index or metric mismatch.
  • Reorder prompt headers. If ΔS spikes, lock the schema.
  1. Apply the module
  • Retrieval drift → BBMC + Data Contracts.
  • Reasoning collapse → BBCR bridge + BBAM variance clamp.
  • Dead ends in long runs → BBPF alternate path.

Typical breakpoints and the right fix

1) Tool call JSON drifts or fields missing

Symptoms: function arguments renamed, nulls where objects expected, tool order wrong.
Why: provider side decoding or safety rewrite, schema not anchored.
Do this:

2) “Reasoning prelude” leaks into final answer

Symptoms: long internal thoughts appear as user-visible output or consume budget.
Why: channel separation not fixed in the contract; model routes text to a single stream.
Do this:

3) High similarity yet wrong meaning

Symptoms: top-k looks relevant but answer cites the wrong slice.
Why: embedding metric vs semantics, or chunk boundary bleed.
Do this:

4) Answers flip across seeds or sessions

Symptoms: small paraphrases flip the conclusion.
Why: uncontrolled variance and unstable memory joins.
Do this:

5) Hybrid retrieval worse than single retriever

Symptoms: HyDE + BM25 underperforms.
Why: query parsing split or ranker saturation.
Do this:

6) Very long tasks collapse near the end

Symptoms: truncation, repetition, or reset at the tail.
Why: entropy collapse in extended chains.
Do this:


Escalation criteria

Open provider tickets only after these pass:

  • ΔS ≤ 0.45 across 3 paraphrases with fixed schema
  • Coverage ≥ 0.70 on the target section
  • Live traces show correct tool ordering and bounded variance
    See Debug Playbook

If the very first call fails after a new deploy, check boot order and fences:
Bootstrap Ordering, Deployment Deadlock, Pre-Deploy Collapse


Copy-paste prompt for a safe triage

I uploaded TXT OS and the WFGY Problem Map.

My DeepSeek bug:
- symptom: [brief]
- traces: [ΔS(question,retrieved)=…, ΔS(retrieved,anchor)=…, λ states, tool logs]

Tell me:
1) which layer is failing and why,
2) which exact WFGY page to open from this repo,
3) the 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. Do not change infra.

🔗 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