mirror of
https://github.com/proffesor-for-testing/agentic-qe.git
synced 2026-09-19 08:45:47 +08:00
Dragan Spiridonov
af4140efb9
Apply Phoenix Architecture principles to AQE — regenerate internals, conserve the interface. ADR-113 (evals are oracles, durable-first tests, mutation-score gate): - src/validation/oracle-eval.ts grades a generated test by running it against a reference impl + operator mutants (reuses qe-arena, ADR-104); rejects assertion-less output - oracle mode wired into scripts/run-skill-eval.ts (skips under the simulating runner; runs live) - qe-test-generation rewritten durable-first; its eval flipped to oracle - src/feedback/regenerability-gate.ts: mutation-score + regenerability gate (warn-default, opt-in block) - live-validated: Claude sonnet-4-6 clears the oracle 5/5; weak local-model output correctly fails ADR-114 (conservation layer, Phoenix essay 14): - src/validation/conservation-guard.ts + scripts/conservation-guard.ts guard CLI commands, output schemas, and dashboard API against breaking removals (deprecation registry exempts) - policy doc + baselines in verification/conservation/; wired into invariant-check.yml non-blocking LLM provider fixes: - OpenRouterProvider.type was 'openai' but registered under 'openrouter' -> ProviderManager failover never found it; fixed to 'openrouter' end-to-end - Gemini default model gemini-1.5-pro was retired -> gemini-2.5-flash; models/pricing refreshed from the live Gemini ListModels API - scripts/provider-health.ts (npm run providers:health) Cheaper-model eval lanes: - scripts/oracle-model-bench.ts + openrouter-models.ts; openai/gpt-oss-120b clears the oracle ~50-70x cheaper than Claude Sonnet; guide in docs/guides/cheaper-model-eval-lanes.md Verification: npm run build ok; 115 new+touched tests pass; 724 LLM tests pass; lint + parity clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implementation Plans
This directory contains GOAP-style implementation plans for major AQE features.
Active Plans
Skill Validation: Trust But Verify
Status: Planning Priority: High Estimated Effort: 214 hours across 5 phases
Implements a 4-layer validation system for AQE skills:
Level 3: Evaluation Suite (measure behavior)
Level 2: Executable Validator (verify output)
Level 1: JSON Schema (structure output)
Level 0: SKILL.md Instructions (guide intent)
Key Deliverables:
- Trust tier frontmatter extension for SKILL.md
- JSON output schemas for all P0 skills
- Executable validators with graceful degradation
- YAML-based evaluation test suites
- CI/CD workflow for multi-model validation
- ReasoningBank integration for learning
Related Resources:
- Schemas:
/docs/schemas/ - Templates:
/docs/templates/ - Exemplars:
.claude/skills/testability-scoring/,.claude/skills/brutal-honesty-review/
Plan Structure
Each plan follows the SPARC-GOAP methodology:
- Specification: Define goal state and success criteria
- Pseudocode: Plan action sequences and state transitions
- Architecture: Structure solution components
- Refinement: TDD implementation with iteration
- Completion: Integration, validation, deployment
Milestone Template
Milestone X.Y: Name
Description: What this milestone achieves
Complexity: Low/Medium/High
Estimated Hours: N
Agent Assignment: qe-agent-name
Dependencies: [milestone IDs]
Parallel Opportunity: Yes/No
Deliverables:
- path/to/deliverable
Success Criteria:
- [ ] Measurable outcome
Memory Namespaces
Plans define memory namespaces for pattern learning:
aqe/{feature}/
├── patterns/ - Learned patterns
├── outcomes/ - Execution outcomes
├── metrics/ - Performance metrics
└── trends/ - Historical trends
Creating New Plans
- Copy the structure from an existing plan
- Define clear goal state vs current state
- Break into SPARC phases with milestones
- Assign agents and estimate complexity
- Map dependencies and parallel opportunities
- Define success criteria for each milestone
- Add to this README