CodeWeave-SENG456

# CodeWeave - SENG 456 Capstone **Capstone option:** 6 - Autonomous Software Development Team **Version:** 8.0 - evaluation-oriented capstone package CodeWeave is a bounded, stateful multi-agent software engineering system with one Chief Orchestrator and five specialists: Product Owner, System Architect, Coder, QA/Tester, and independent Code Reviewer. The Chief selects **one next specialist per turn** from current state, artifact freshness, critique evidence, and root-cause ownership. It does not use a fixed rejection-to-Coder pipeline. ## Start here ```bash python scripts/run_capstone.py --verify ``` For a full offline evidence regeneration: ```bash python scripts/run_capstone.py --reproduce ``` Then read: - `GRADER_QUICKSTART.md` - `assessment/ASSESSMENT_MATRIX.md` - `orchestration/orchestration_spec.json` ## Core course requirements ### State Management - shared blackboard: `workspace/project_state.json` - schema: `workspace/state_schema.json` - inspectable evolution: `evidence/offline_reference_run/state/` - causal freshness: `freshness_matrix.*` ### Reflection & Feedback Loops The reproducible reference case intentionally exposes a race condition. QA fails it; Reviewer classifies the root cause as architectural; Chief routes to System Architect; architecture and implementation are revised; QA re-runs; Reviewer approves. ### Dynamic Routing The key non-trivial transition is: ```text code-reviewer -> system-architect ``` See sequence 5 in `evidence/offline_reference_run/routing_history.json`. ## Concrete implementation `demo/equipment_loan_api/` contains a working Python/SQLite University Equipment Loan API with nine automated tests, including concurrency and borrower-ownership cases. ## Machine-readable architecture `orchestration/orchestration_spec.json` declares the full orchestration contract independently of prompt prose: agents, triggers, state ownership, routes, freshness dependencies, quality gates, loop bounds, terminal states, safety policy, and evidence paths. ## GitHub import bridge Myskillos currently supports importing from GitHub. Build a minimal GitHub-ready source repository with: ```bash python scripts/build_github_import.py ``` Then use `CodeWeave_GitHub_Import_V8.zip` as the repository contents. See `github_import/README.md`. This bridge is source material only and must not be mislabeled as the genuine Myskillos export. ## Myskillos Use `MYSKILLOS_SETUP.md` and `myskillos/` to build/import the real platform project. Genuine Myskillos screenshots/export belong only under `evidence/myskillos/`. Offline evidence is explicitly separated and **must never be presented as Myskillos telemetry**. ## Live experimental evaluation After the genuine Myskillos run is recorded, compute platform metrics with: ```bash python scripts/evaluate_live_run.py --require-pass ``` The evaluator measures six routing probes, the critique-revision-retest-approval loop, clean termination, the multimodal RT-06 route, optional freshness violations, and specialist coverage when a route sequence is captured. Results are written to `evidence/myskillos/live_run/EXPERIMENTAL_RESULTS.{json,md}`. See `evaluation/LIVE_EVALUATION_PROTOCOL.md`. ## Report - Complete standalone review copy: `report/standalone/codeweave_report.pdf` - Report body for instructor template: `report/report_content.tex` The instructor explicitly requested the official `.tex` template. Therefore the standalone PDF is a review artifact, not the final LMS report until its content is transferred into the official template. ## Final readiness ```bash python scripts/submission_status.py ``` When genuine external artifacts are available, place them in `external_dropbox/` and run: ```bash python scripts/run_capstone.py --finalize ``` This performs ingestion, provenance hashing, evidence indexing, pre-submission audit, and final LMS ZIP construction. The package intentionally refuses to claim final LMS readiness until all three external artifacts exist: 1. official instructor-template report PDF, 2. genuine Myskillos export ZIP, 3. genuine structured live-run record + screenshots.