| name | codebase-overview |
| description | Project overview and domain routing table for the UnifAI monorepo. Each domain has its own skill with `paths` scoping for auto-surfacing when editing relevant files. |
UnifAI Codebase Overview
UnifAI is a multi-service platform for building and running AI agent workflows.
The monorepo contains backend services (Python/Flask), a frontend (React/TypeScript),
shared libraries, and infrastructure tooling.
Domain Routing Table
| Path prefix | Domain | Skill |
|---|
multi-agent/ | Multi-Agent System (MAS) | domains/multi-agent/SKILL.md |
rag/ | RAG Pipeline | domains/rag/SKILL.md |
shared-resources/identity/ | Identity & Auth | domains/identity/SKILL.md |
ui/client/src/ | Frontend UI | domains/ui/SKILL.md |
global_utils/ | Global Utilities | domains/global-utils/SKILL.md |
backend/ | Platform Backend | domains/backend/SKILL.md |
rag/infrastructure/celery/ | Celery Workers | domains/celery/SKILL.md |
temporal-worker/ | Temporal Workers | domains/temporal-worker/SKILL.md |
Domain skills are auto-surfaced via paths when editing files in their scope.
Service Architecture Summary
| Service | Role | Tech Stack | Port |
|---|
| multi-agent | AI agent orchestration engine | Python, Flask, MongoDB, Redis, LangGraph | 5001 |
| rag | Document ingestion & retrieval | Python, Flask, MongoDB, Qdrant, Celery | 5002 |
| identity | Auth, teams, SSO | Python, Flask, Keycloak, MongoDB | 5003 |
| backend (platform) | Admin config, platform API | Python, Flask, MongoDB | 5000 |
| ui | Web frontend | React, TypeScript, Vite, TailwindCSS | 3000 |
| global_utils | Shared Python library | Python (package) | — |
| celery | Async task workers for RAG | Python, Celery, Redis | — |
| temporal-worker | Distributed workflow execution | Python, Temporal | — |
Cross-Cutting Concerns
| Concern | Where it lives |
|---|
| Authentication | shared-resources/identity/ + global_utils/identity.py |
| Configuration | Each service's config/ + global_utils/config/ |
| MongoDB access | Each service's adapters/ or infrastructure/ |
| Redis caching | global_utils/redis/ + per-service usage |
| Message queues | RabbitMQ via Celery (global_utils/celery_app/) |
| Workflow orchestration | Temporal (multi-agent/adapters/outbound/temporal/) |
Dev-Guide (optional factual reference)
The .cursor/unifai-dev-guide/ directory contains service documentation with class architecture,
endpoint catalogs, port abstractions, and MongoDB/Qdrant schemas. Domain SKILL.md files
already contain compact versions of the most useful tables. The full dev-guide is available
for deeper reference:
| Resource | Description |
|---|
.cursor/unifai-dev-guide/docs/services/<service>.md | Full endpoint signatures, class graphs, collection schemas |
.cursor/unifai-dev-guide/source-map.yaml | Maps services → composition roots, ports, endpoints, Mongo collections |
.cursor/unifai-dev-guide/topology.yaml | Service graph, tech stacks, runtime edges |
.cursor/unifai-dev-guide/guide-index.yaml | Maps code path globs → doc sections (used by gen-docs.js, CI) |
