Claude/multi agent orchestrator sojag#3055
Open
AlexanderSmil wants to merge 8 commits intovinta:masterfrom
Open
Claude/multi agent orchestrator sojag#3055AlexanderSmil wants to merge 8 commits intovinta:masterfrom
AlexanderSmil wants to merge 8 commits intovinta:masterfrom
Conversation
Includes process boundaries (Розділ 0) with integration map for ПЗ.9.10, ПУ.2.10, Prozorro, Мегаполіс, АСКОД, and full RACI matrix (15 steps × 5 roles). https://claude.ai/code/session_01C85MmENY41FpGmV9ZprDnV
…endations-9hMj4 docs: add normalized regulation ПЗ.9.4 for TMC nomenclature management
Full multi-school psychological self-exploration platform combining the Leela board game with a five-school analytical engine (Jungian, Freudian, CBT, Gestalt, Transactional Analysis). Components delivered: - core/security.py: HMAC-SHA256 pseudonymisation + AES-GCM (Fernet) encryption - core/kms_lite.py: KMS abstraction stub (AWS KMS / Vault ready) - storage/vector_storage.py: encrypted JSONL store; _SAFE_FIELDS whitelist enforces privacy-by-design (session_id NEVER persisted in plain form) - storage/faiss_adapter.py: FAISS kNN index with graceful degradation - ai/embedding_client.py: async embed client with disk cache - ai/synthesizer.py: ConflictSynthesizer — dialectical polarity detection - agents/qa/*: five psychological analyzers (async, pluggable) - agents/feedback/user_feedback_agent.py: rating 1-5 validation + storage - agents/learning/self_learning_agent.py: feedback aggregation + metrics - models/unified_insight.py: dataclasses for the full insight graph - game/leela_psychological.py: main facade — asyncio.gather over all layers - exporters/clinical_exporter.py: Markdown / JSON / PDF clinical reports - web/feedback_api.py: Flask Blueprint + create_app factory - web/templates/feedback.html: Ukrainian-language feedback form - tests/test_anonymization.py: 13 privacy / cryptography unit tests - tests/test_integration.py: 9 end-to-end integration tests - docs/clinical_guide.md: clinical usage guide for therapists and clients - requirements.txt, Dockerfile, docker-compose.yml, .env.example - scripts/init_secrets.sh, scripts/build_release.py - .github/workflows/ci.yml: matrix test + lint + pip-audit pipeline Test results: 22/22 passed (pytest 9.0.3, Python 3.11) https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
Compressed archive of the complete Leela Quest+Psychology AI v1.1.0 project (45 source files, 74 entries, 30 KB) — excludes __pycache__, runtime data directories, and secrets per .gitignore rules. https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
…alidation
Addresses all Medium/High gaps from the Staff Engineer Architectural Audit:
1. tests/test_web_api.py (NEW — 14 tests)
- POST /api/feedback: valid, missing fields, rating bounds (0,6,1,5),
oversized comment, malformed JSON, empty body
- GET /api/feedback/stats: empty queue + after submissions
- GET /api/feedback/<id>: unknown id + per-insight filtering
Fixture pattern: isolated VectorStorage per test, cleared _feedback_queue
2. web/feedback_api.py
- Add MAX_COMMENT_LENGTH = 2000 guard (400 if exceeded)
- Treat get_json() returning None as invalid JSON (400, not silent empty)
3. .github/workflows/ci.yml
- Remove "|| true" from pip-audit — build now fails on real CVEs
- Replace with documented allowlist comment (empty by default)
Test results: 36/36 passed (22 existing + 14 new)
https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
…metrics Closes all remaining architectural audit recommendations. 1. ai/embedding_synthesizer.py (NEW — v2.0 audit item vinta#6) - Dynamic polarity detection via cosine similarity between concept embeddings - Replaces static POLARITY_MAP with semantic distance (threshold 0.55) - Intensity labels: profound/strong/noticeable/subtle based on distance - Concept embedding cache to avoid re-embedding within a session - Works with any school — no manual polarity pairs needed 2. agents/devops/health_agent.py (NEW) - HealthAgent: synchronous self-diagnostic of security/storage/ai/synthesis - Thread-isolated asyncio.run() to avoid event loop conflicts - Returns {"status": "healthy"|"degraded", "checks": {...}, "version": "1.2.0"} - Implements async execute() for orchestrator compatibility 3. monitoring/metrics_collector.py (NEW) - Thread-safe rolling-window MetricsCollector - timer() context manager, record_ms(), increment() counter - snapshot() with avg/min/max/p95 per metric name - Module-level singleton: from monitoring.metrics_collector import metrics 4. game/leela_psychological.py - Instrument analyze_turn: records total latency, per-school call counts, consent_granted/consent_denied counters 5. web/feedback_api.py - GET /api/health — liveness/readiness probe (200 healthy, 503 degraded) - GET /api/metrics — performance metrics snapshot 6. tests/test_embedding_synthesizer.py (NEW — 17 tests) - 6 EmbeddingSynthesizer tests (polarity detection, cache, alignment) - 6 MetricsCollector tests (counters, timers, rolling window, p95, reset) - 5 HealthAgent tests (all-ok, latency, async execute, unknown action) Test results: 53/53 passed https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
…ring, pytest.ini
1. journals/journal_manager.py (NEW)
- JournalEntry dataclass: turn, cell_id, cell_name, move_type, schools,
synthesis_snippet (≤200 chars), archetypes, timestamp
- JournalManager: record(), load_session(), export_narrative(), delete_session()
- Privacy: only synthesis snippet stored — never raw user text
- Path traversal sanitisation on session_id → filename mapping
- GDPR erasure via delete_session()
2. game/leela_psychological.py
- Wire EmbeddingSynthesizer as primary (2+ layers) / ConflictSynthesizer fallback
- Optional JournalManager param: engine = LeelaPsychologicalEngine(journal=jm)
- Journal.record() called after each turn when journal is set
3. pytest.ini (NEW)
- asyncio_mode = strict, testpaths = tests, pythonpath = .
- Enables: cd leela_psychology && python -m pytest
4. tests/test_integration.py
- test_multi_layer_uses_embedding_synthesizer: verifies v2.0 synthesis path
- test_journal_records_turns: verifies turn counter and entry count
5. tests/test_journal_manager.py (NEW — 13 tests)
- JSONL file creation, turn counter, load/delete, snippet truncation,
privacy (no raw text), session isolation, narrative export,
path traversal protection
6. leela_psychology_v1.2.0.tar.gz (replaces v1.1.0)
- 91 entries, 43 KB — includes all v1.2.0 files
Test results: 69/69 passed
https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
…docs, CSS
1. leela_psychology/__init__.py (NEW)
- Package root with __version__ = "1.2.0" and version history
2. monitoring/__init__.py (NEW)
- Explicit package init (was namespace package — now proper package)
3. docs/architecture.md (NEW)
- Full system architecture: ASCII diagram, package structure,
data flow / privacy chain, v1.x vs v2.0 synthesis comparison,
HTTP API reference, test suite breakdown
4. web/static/leela.css (NEW)
- Extract inline styles from feedback.html into proper CSS file
- CSS custom properties (design tokens) for consistent theming
- Added :focus, :active, :hover states and .health-ok/.health-degraded
5. web/templates/feedback.html
- Replace <style> block with <link rel="stylesheet" href="/static/leela.css">
6. scripts/build_release.py
- Add check_package_init_files(): 15 packages verified
- Add check_v120_modules(): 6 v1.2.0 files + content markers
- Add check_version(): __version__ == "1.2.0" assertion
- All 5 non-test checks pass: ./scripts/build_release.py --skip-tests
7. leela_psychology_v1.2.0.tar.gz
- Rebuilt clean (no .pytest_cache): 87 entries, 45 KB
https://claude.ai/code/session_01Nmtk7oaHGcmm56rjXeMa3Z
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Project
Project Name
Checklist
Add project-name- [project-name](url) - Description ending with period.Why This Project Is Awesome
Which criterion does it meet? (pick one)
Explain:
How It Differs
If similar entries exist, what makes this one unique?