37 lines
1.2 KiB
Markdown
37 lines
1.2 KiB
Markdown
# Sprint 62 Plan: Adapter Quality Telemetry and Failure Taxonomy
|
|
|
|
## Context
|
|
|
|
To improve adapters systematically, failures must be categorized, measured, and
|
|
trended. Sprint 62 introduces a unified transpilation failure taxonomy.
|
|
|
|
---
|
|
|
|
## Goals
|
|
|
|
1. Define canonical failure taxonomy for lowering/raising/equivalence/gates
|
|
2. Attach taxonomy tags to all failure packets
|
|
3. Add trend analysis for repeated failure modes
|
|
4. Prioritize fixes by impact and frequency
|
|
|
|
---
|
|
|
|
## Steps
|
|
|
|
### Step 849: Transpilation failure taxonomy schema (12 tests)
|
|
### Step 850: Packet tagging integration across pipelines (10 tests)
|
|
### Step 851: Failure aggregation indexer (10 tests)
|
|
### Step 852: Trend and recurrence detector (10 tests)
|
|
### Step 853: Prioritization scoring model (impact x frequency x tier) (10 tests)
|
|
### Step 854: Recommended remediation generator (8 tests)
|
|
### Step 855: `whetstone_get_failure_trends` MCP tool (8 tests)
|
|
### Step 856: `whetstone_get_top_adapter_gaps` MCP tool (8 tests)
|
|
### Step 857: Weekly quality report artifact generator (8 tests)
|
|
### Step 858: Sprint 62 integration summary + regression (8 tests)
|
|
|
|
---
|
|
|
|
## Data Rule
|
|
|
|
- Every failed gate/equivalence check must emit one primary taxonomy code.
|