5efb380da5
* feat: automate yiacad pr review lane * fix: stabilize yiacad review and spec gates * fix: scope kicad exports to repo hardware * fix: ignore kicad block library sources in exports * feat: Add infra VPS monitoring runbook and healthcheck scripts - Created INFRA_VPS_RUNBOOK_2026.md detailing operational procedures for monitoring VPS services. - Added infra_vps_healthcheck.sh script for automated health checks on DNS, TLS, TCP, and HTTP for VPS services. - Introduced infra_vps_security_audit.sh for non-intrusive security checks on external VPS services. - Established JSON schema for infra VPS inventory in infra_vps.schema.json. - Developed integration for runtime status reporting in the Next.js API route. - Implemented Playwright tests for smoke testing the application and ensuring core functionalities. - Updated Makefile for development dependencies and testing commands. - Created various test files for unit and end-to-end testing across different components. * feat(agentics): update mesh agents, gates, prompts, and workflows
5.9 KiB
5.9 KiB
Agentic Landscape (Kill_LIFE)
Posture de départ
- Spec-driven backbone:
specs/*etspecs/constraints.yaml - Standards and gates:
compliance/*,bmad/* - Runtime local:
ZeroClaw+ MCP +crazy_lifecomme éditeur - Overlays:
LangGraph,AutoGen,n8nen option contrôlée - Exécution locale de préférence, GitHub pour la traçabilité distante
flowchart TD
Inputs[Issue / PR / demande humaine] --> Labels[Labels ai:* + Scope]
Labels --> Router{Routeur d’exécution}
Router --> SpecPath[specs/ (00->04)]
Router --> GitHubPath[workflow local / GitHub dispatch]
SpecPath --> Validation{specs + compliance}
Validation --> Tools[tools/*]
Tools --> RepoTools[repo_state + compliance + auto_check_ci_cd]
Tools --> MCP[tools/hw/* + MCP launchers]
Tools --> TUI[tools/cockpit/refonte_tui.sh]
Tools --> CADFusion[tools/cad/yiacad_fusion_lot.sh]
CADFusion --> CadEvidence[Evidence CAD (cad-fusion)]
MCP --> Evidence[Evidence packs locaux]
RepoTools --> Evidence
CadEvidence --> Evidence
TUI --> Logs[artifacts/refonte_tui/*.log]
Logs --> Evidence
Evidence --> Audit[docs/evidence/*]
GitHubPath --> CI[.github/workflows/*]
CI --> Audit
Audit --> PM[PM + Plan]
PM --> Handoff[Handoffs / Plans / Labels]
Handoff --> Agents[agents/ + .github/agents/ + prompts]
Modèle de fonctionnement
agents/et.github/agents/définissent les compétences minimales.tools/cockpit/refonte_tui.shest la surface de pilotage texte (status, logs, lot-chain).ZeroClawreste le runtime local d’orchestration, sans remplacer la source de vérité.- Les surfaces MCP sont locales par défaut,
huggingfaceen option distante officielle. - Les preuves et traces sont obligatoires pour les lots de refonte.
Décisions de refonte actives
- pas de dépendance forte externe pour les lanes essentielles,
- tout overlay IA passe par une validation de sécurité (labels, scope, denylist),
- priorité à la reproductibilité documentaire (
specs,plans,diagrams,evidence).
Delta mesh 2026-03-20
Le paysage agentique du projet opere desormais comme un overlay de gouvernance maillée sur trois repos, avec neuf agents principaux.
| Agent | Repo owner prefere | Responsabilite |
|---|---|---|
PM-Mesh |
Kill_LIFE |
backlog tri-repo, priorisation, ordre de propagation |
Arch-Mesh |
Kill_LIFE |
contrats inter-repos, ADR, compatibilite |
Embedded-CAD |
Kill_LIFE |
firmware, hardware, evidence embarquee |
Runtime-Companion |
mascarade |
runtime IA, providers, adaptateurs MCP |
CAD-Fusion |
Kill_LIFE |
pilote YiACAD, health CAD, preuve et rollback |
Web-Cockpit |
crazy_life |
UI cockpit, workflow editor, supervision |
QA-Compliance |
transversal | matrices, smokes, preuves |
Docs-Research |
transversal | README, specs, Mermaid, benchmark OSS |
SyncOps |
Kill_LIFE |
SSH, repo-state, TUI, logs, sync continue |
Sous-agents standardises:
Lot-Planner,Risk-TriagerSchema-Guard,Compat-AuditorFirmware-Lane,CAD-Lane,CAD-FusionProvider-Bridge,Runtime-SmokeUI-Flow,Schema-ConsumerSpec-Mirror,Evidence-PackMermaid-Map,OSS-BenchmarkSSH-Health,Log-Ops
Skills opératoires actives:
bash-cli-tui:SyncOps,CAD-Fusion,Doc-Runbookplaywright:Web-Cockpitquand une validation UI réelle devient nécessaireOSS-Benchmark:Docs-Research,CAD-FusionEvidence-Pack:QA-Compliance,Docs-Research
Routing prioritaire des lots ouverts:
PM-MeshgardeT-RE-204comme lot exécutable bloquant.CAD-Fusiontient la lane parallèleT-RE-209/T-RE-210.SyncOpspublie la synthèse hebdomadaire et la checklist de sortie (T-RE-301/T-RE-304).
Regle de coexistence: nous ne sommes pas seuls dans le codebase; toute intervention doit rester dans un write_set declare et observable.
Matrice detaillee publiee: docs/AGENT_SPEC_MODULE_MATRIX_2026-03-20.md
Delta d'execution machine-readable:
Schema-Guardpublie des schemas JSON versionnes pouragent_handoff,repo_snapshot,workflow_handshake.Runtime-SmokeetSchema-Consumerconsomment ces contrats depuis la lane soeurKill_LIFE-mainau moment des preflights compagnons.- La validation CI visible du handshake passe desormais par
.github/workflows/mesh_contracts.yml.
Delta agentics 2026-03-29
Mise a jour de la lane PR review automation (branche feat/yiacad-pr-review-automation, PR #18).
Nouveautes
- Gate S2 (bmad/gates/gate_s2.md) -- couvre la lane Web & PR review : Playwright, BullMQ worker, evidence intelligence_program/, validation summary_short.schema.json.
- Subagent PR-Review-Orchestrator -- ajoute a Web-CAD-Platform, write set : web/project/.ci/.
- Schema-Guard -- reference specs/contracts/yiacad_uiux_output.schema.json comme contrat evidence de la lane review.
- QA-Compliance -- gate S2 ajoutee aux gates de validation release.
- bmad/templates/handoff.md -- enrichi : sections Gate passed + Consumer contracts.
- bmad/rituals/kickoff.md -- identification dynamique des gates (S0/S1/S2/...) selon le domaine du lot.
Table agents canoniques (2026-03-29)
| Agent | owner_agent | Handoff contract | Gates |
|---|---|---|---|
| PM-Mesh | Kill_LIFE | agent_handoff.schema.json | S0 |
| Arch-Mesh | Kill_LIFE | agent_handoff.schema.json | S0 |
| Schema-Guard | Kill_LIFE | workflow_handshake.schema.json | S0 |
| QA-Compliance | transversal | operator_lane_evidence.schema.json | S0, S1, S2 |
| KillLife-Bridge | Kill_LIFE | agent_handoff.schema.json | S0 |
| Runtime-Companion | Kill_LIFE | agent_handoff.schema.json | S0 |
| SyncOps | Kill_LIFE | operator_lane_evidence.schema.json | S0 |
| Web-CAD-Platform | Kill_LIFE | summary_short.schema.json | S0, S2 |
| Docs-Research | transversal | agent_handoff.schema.json | S0 |
| Embedded-CAD | Kill_LIFE | agent_handoff.schema.json | S0, S1 |
| Firmware | Kill_LIFE | operator_lane_evidence.schema.json | S1 |
| UX-Lead | Kill_LIFE | agent_handoff.schema.json | S0 |