Revornix Architecture Overview

This document is organized based on the current repository code and docker-compose.yaml plus each service’s .env.example files.

1. System Components and Responsibilities

• Web frontend (Next.js): UI and interaction layer; entry at web, build-time NEXT_PUBLIC_* injection.
• API backend (FastAPI): core business and data orchestration; entry at api/main.py, provides docs, columns, AI, graph, notifications, and more.
• Celery worker: async and heavy tasks; entry at celery-worker/common/celery/app.py, handles embedding, summaries, graphs, podcasts, notifications.
• Hot News service (Node): aggregation service; entry at hot-news/src/index.ts, provides ranking APIs.
• Nginx (optional): reverse proxy and unified entry; config at nginx/nginx.conf.
• MCP (built-in): MCP client/server capability inside the API, routes at api/mcp_router/*.

2. Data and Infrastructure

• PostgreSQL: primary business data store; ORM models in api/models and celery-worker/models.
• Redis: cache and task queue (Celery broker/backend), config in api/config/redis.py and celery-worker/config/redis.py.
• Milvus: vector search and similarity, config in api/config/milvus.py.
• Neo4j: knowledge graph storage/query, config in api/config/neo4j.py.
• MinIO (file-backend): default file system and object storage, controlled by FILE_SYSTEM_*.
• Optional external object storage: AWS S3 / Aliyun OSS / generic S3; logic in api/file/* and celery-worker/file/*.

3. Core Requests and Data Flow

1. User request from Web -> API handles business logic -> reads/writes PostgreSQL/Redis/Milvus/Neo4j -> returns response.
2. Document/column processing -> API enqueues Celery task -> worker parses/summarizes/embeds/builds graph/podcasts -> writes results and notifies.
3. File upload/access -> API generates file URL -> MinIO or external storage read/write -> frontend accesses via public URL.
4. Hot News ranking -> Web calls Hot News service -> service aggregates data and returns list.

4. Deployment Topology (Default Docker Compose)

• web: frontend container, depends on api and file-backend.
• api: main backend container, depends on postgres, redis, neo4j, milvus.
• celery-worker: task worker, depends on postgres, redis, neo4j, milvus.
• hot-news: aggregation service.
• file-backend: MinIO file system.
• postgres, redis, neo4j, milvus, etcd, minio: infrastructure components.
• nginx: unified entry and routing.
• rsshub + browserless: RSSHub crawler and browser dependencies.

5. Observability and Extension Points

• Sentry: enable per API and worker via API_SENTRY_* and WORKER_SENTRY_*.
• Langfuse: trace requests via LANGFUSE_*.
• Model providers: multi-model/provider support; built-in initialization in api/data/sql/create.py.
• Third-party login and SMS/email: Google/GitHub/WeChat/Tencent SMS/SMTP integrated at the API layer.

6. Initialization and Startup Flow (Brief)

1. Prepare .env and each service .env.
2. Start base services (Postgres/Redis/Neo4j/Milvus/MinIO).
3. Run api/data/milvus/create and api/data/sql/create to initialize data.
4. Start API, worker, web, and Hot News.