denpex-mcp 0.2.8 → 0.2.10

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/README.md +14 -13
  2. package/dist/index.js +7404 -871
  3. package/package.json +9 -6
package/README.md CHANGED
@@ -7,7 +7,7 @@ straight off disk and hands back the root cause and the fix.
7
7
 
8
8
  - 🩺 **Root cause + escalating fix** (primary → fallback → emergency), verification command, and prevention.
9
9
  - 🧭 **Find the culprit rank** in a distributed/NCCL hang, and route GPU Xid faults to the right node action — offline, instantly.
10
- - 📚 **Bundled failure encyclopedia** (300+ classes offline; the full corpus runs online) searchable by error code or symptom.
10
+ - 📚 **Bundled failure encyclopedia** (500+ documented classes) searchable by error code or symptom.
11
11
  - 🔒 **Zero-egress mode** (`DENPEX_LOCAL=1`): diagnose entirely on-host, nothing leaves the machine. Plus client-side PII scrubbing even in cloud mode.
12
12
 
13
13
  ---
@@ -95,10 +95,11 @@ secure data centers, and privacy-sensitive labs. Add `DENPEX_LOCAL=1`:
95
95
  } } }
96
96
  ```
97
97
 
98
- The bundled engine (fast patterns + a 300+-class encyclopedia snapshot + Xid routing + cascade
99
- localization) covers the common failure classes; the full, continuously-grown corpus and the
100
- deep-analysis tiers run online. The LLM fallback for novel
101
- errors is cloud-only and simply unavailable offline.
98
+ The package bundles the same canonical deterministic diagnosis pipeline used by the hosted
99
+ wrapper: 10,700+ patterns, causal detectors and contradiction gates, Xid routing, cross-rank
100
+ cascade localization, and the 500+ class encyclopedia. It returns the initiating rank/node
101
+ when the evidence identifies one. Only the hosted LLM fallback and live-research enrichment
102
+ are unavailable offline.
102
103
 
103
104
  ---
104
105
 
@@ -107,7 +108,7 @@ errors is cloud-only and simply unavailable offline.
107
108
  | Tool | What it does | Needs |
108
109
  |------|--------------|-------|
109
110
  | `diagnose_crash` | Root cause + escalating fix from `logs` / `file` / `dir`. Also accepts `telemetry` (dmesg / nvidia-smi / ibstat), `codeContext`, `topologyMap`, typed `operatorAnswers`, and (after an anonymous MCP restart) the prior `refinementToken`. | canonical cloud by default; explicit offline mode |
110
- | `lookup_failure` | Search the bundled encyclopedia (300+ classes) by code/keyword/symptom. | works offline |
111
+ | `lookup_failure` | Search the bundled encyclopedia (500+ classes) by code/keyword/symptom. | works offline |
111
112
  | `analyze_cascade` | Find the culprit rank in a distributed hang; route Xid faults. | works offline |
112
113
  | `submit_fix_outcome` | Report whether a fix worked. **This is what makes the next diagnosis better.** | API key |
113
114
  | `check_quota` | Your plan and remaining diagnoses this month. | API key |
@@ -134,10 +135,9 @@ Two fields deserve attention when you write the agent loop:
134
135
  - **`confidence: null`** does not mean zero. It means the engine measured this band's
135
136
  historical accuracy and declined to state a number. Treat the verdict as a lead. Do
136
137
  not substitute the raw score.
137
- - **`unfalsified: true`** means no falsification contract ran on the verdict. It is
138
- always true offline (the offline engine has no contract layer), so an offline
139
- `GPU_XID_ECC` has *not* been checked against telemetry that would contradict it.
140
- Confirm before you drain a node or file an RMA.
138
+ - **`unfalsified: true`** identifies only the explicitly enabled reduced compatibility
139
+ fallback, where the canonical contradiction graph did not run. Normal cloud and
140
+ `DENPEX_LOCAL=1` diagnosis both use the canonical engine and set it to false.
141
141
 
142
142
  ---
143
143
 
@@ -156,10 +156,11 @@ Two fields deserve attention when you write the agent loop:
156
156
 
157
157
  ```bash
158
158
  npm install
159
- npm run build # regenerates the offline corpus, then bundles dist/index.js
159
+ npm run build # regenerates the canonical engine + corpus, then bundles dist/index.js
160
160
  npm test # 111 unit + protocol-integration tests
161
161
  npm run inspector # interactive protocol inspector (MCP Inspector)
162
162
  ```
163
163
 
164
- The offline corpus (`src/data/denpex_patterns.json`) is generated from the repo's single
165
- source of truth via `npm run bundle:patterns` — never hand-edit it.
164
+ The offline corpus (`src/data/denpex_patterns.json`) and canonical engine
165
+ (`src/offline/canonical-engine.mjs`) are generated from the repo's single source of truth
166
+ during `npm run build` — never hand-edit either artifact.