Live: DSI Platform — New Features

Session: "DSI Platform — New Features" interrupted

Pipeline complet pour les nouvelles fonctionnalités de la plateforme Macaron. Discovery réseau → Comité stratégique GO/NOGO → Architecture → Sprint Dev (6 devs spécialisés) → CI/CD → QA parallèle → Deploy staging/prod → Rétrospective.

Start the conversation:

Chat with:

Inter-Agent Feed

Patterns

100%

idle

Thinking...

PRs & Deliverables

No PRs yet. Agents will produce them.

Artifacts & Decisions

No artifacts yet

Session Memory

No session memory yet

Project Memory

Claude Code Instructions

# MACARON AGENT PLATFORM ## WHAT Web multi-agent platform SAFe-aligned. Agents collaborate (debate/veto/delegate) autonomously. FastAPI + HTMX + SSE + SQLite. Dark purple theme. Port 8099. ## RUN ``

auto-seed · convention

GitHub Copilot Instructions

# Software Factory — Copilot Instructions ## Build & Run ```bash # Install dependencies pip install -r requirements.txt # Run locally (from the parent _SOFTWARE_FACTORY/ directory) python -m uvicor

auto-seed · convention

Project Specifications

# MACARON AGENT PLATFORM — SPECIFICATIONS v1.0 > **Vision** : Une plateforme web locale permettant de créer, orchestrer et piloter des équipes d'agents IA collaboratifs sur des projets logiciels, ave

auto-seed · architecture

Project README

# 🍪 Software Factory **Multi-agent collaborative platform web** — agents spécialisés (Métier, Lead Dev, Testeur, Sécurité, DevOps...) qui dialoguent entre eux via le protocole A2A, orchestrés selon 8

auto-seed · architecture

Project Vision

Software Factory — Real Agentic ≠ Workflow Automation. Agents collaborent (débat/véto/délégation) pour produire du code. 128 agents SAFe, 23 patterns, 12 workflows, 1222 skills. Stack: FastAPI + Jinja

auto-seed · vision

Shared Memory

agent-count

45+ agents across 5 SAFe levels: Strategic (DSI, CTO, CPO), Portfolio (managers), Program (RTE, SM), Team (dev, QA, DevOps, Security)

team

adversarial-review

Multi-vendor cognitive diversity: Code critic (MiniMax), Security critic (GLM), Architecture critic (Opus). Same LLM cannot evaluate its own output.

team

platform-purpose

Software Factory: emulate a full DSI (Direction des Systèmes d'Information) with autonomous AI agents collaborating on real software projects

vision

differentiation

Real agentic orchestration ≠ workflow automation. Agents debate, veto, negotiate, delegate — not just if/then boxes with LLM wrappers

vision

zero-skip

NEVER skip tests, checks, or quality gates. FIX > SKIP. Always.

convention

tdd-first

Red-Green-Refactor. Tests before code. Coverage 80%+. Complexity < 15 cyclomatic.

convention

fractal-decomposition

L1: Split into 3 concerns (FEATURE → GUARDS → FAILURES). L2: KISS atomic (IMPL → TEST → VERIFY)

convention

lean-values

Quality > Speed, Feedback rapide, Éliminer waste, Respect des personnes, Amélioration continue, Flux continu

convention

stack

FastAPI + HTMX + SSE + SQLite (WAL) + Jinja2 templates

architecture

patterns

8 workflow patterns: Parallel, Sequential, Loop, Router, Aggregator, Hierarchical, Network, Human-in-loop

architecture

Supprimer