OpsRadar is a production-oriented inspection platform built on FastAPI, PostgreSQL, Redis and Celery. The current production scope supports real Linux/Unix host inspection through SSH, encrypted resource credentials, RBAC, audit logs, report export, and a task-oriented AI assistant that drives inspection workflows.

The AI assistant is a workflow orchestrator, not a free-form chat bot. It recognizes OpsRadar actions, persists workflow state, and waits for user confirmation when a business modal or a mutating action is required.

flowchart TD
  U["User message"] --> I["Intent detection"]
  I --> W["Create workflow"]
  W --> E["Check application/environment"]
  E --> A["Check assets / scope"]
  A --> C["Test connectivity"]
  C --> S["Service discovery"]
  S --> R["Match rule sets"]
  R --> T["Create / confirm inspection task"]
  T --> X["Run inspection"]
  X --> P["Generate report"]
  P --> Q["Create issues"]
  Q --> N["AI summary"]
  N --> F["Next actions"]

Workflow callbacks keep the flow alive after the user completes a modal:

• environment_created
• asset_created
• asset_selected
• connection_tested
• services_discovered
• rules_confirmed
• task_created
• task_finished

The assistant only returns actions and explanations that are backed by platform data. It does not invent assets, issues, reports, or repair results.

Prepare .env and TLS certificates:

cp .env.example .env
mkdir -p deploy/certs
# put deploy/certs/tls.crt and deploy/certs/tls.key in place

Start the stack:

docker compose up -d --build
docker compose exec opsradar-api python3 -m backend.app.cli init-admin
docker compose exec opsradar-api python3 -m backend.app.cli check

Open https://<server>/.

The production stack contains:

• nginx: HTTPS entrypoint and reverse proxy.
• opsradar-api: FastAPI + Gunicorn/Uvicorn, serving API and frontend.
• opsradar-worker: Celery worker executing inspection tasks.
• opsradar-beat: CronPlan scanner and task dispatcher.
• postgres: authoritative data store.
• redis: Celery broker and result backend.

Start local PostgreSQL and Redis first, then run:

python3 -m pip install -r requirements.txt
python3 -m backend.app.cli migrate
python3 -m backend.app.cli seed-builtin
python3 -m backend.app.cli init-admin
bash scripts/run_dev.sh

Run the worker in another terminal:

bash scripts/run_worker.sh

Open http://127.0.0.1:4173/.

python3 -m backend.app.cli migrate
python3 -m backend.app.cli seed-builtin
python3 -m backend.app.cli init-admin
python3 -m backend.app.cli check

• OPSRADAR_ENV: development or production.
• OPSRADAR_SECRET_KEY: JWT signing secret.
• OPSRADAR_ENCRYPTION_KEY: AES-GCM credential encryption key. Required in production.
• OPSRADAR_DATABASE_URL: PostgreSQL SQLAlchemy URL.
• OPSRADAR_REDIS_URL: Redis URL used by service checks.
• OPSRADAR_CELERY_BROKER_URL: Celery broker URL.
• OPSRADAR_CELERY_RESULT_BACKEND: Celery result backend URL.
• OPSRADAR_API_WORKERS: Gunicorn worker count.
• OPSRADAR_WORKER_CONCURRENCY: Celery worker concurrency.
• OPSRADAR_SSH_CONNECT_TIMEOUT: SSH connection timeout in seconds.
• OPSRADAR_SSH_COMMAND_TIMEOUT: SSH command timeout in seconds.
• OPSRADAR_MAX_TASK_SECONDS: Celery hard task time limit.
• OPSRADAR_PUBLIC_BASE_URL: public URL used by deployment metadata.
• OPSRADAR_CORS_ORIGINS: comma-separated allowed browser origins.
• OPSRADAR_REPORT_DIR: report output directory.
• OPSRADAR_CHROME_PATH: Chrome/Chromium executable path for PDF export.

• PostgreSQL-only persistence with Alembic migrations.
• Redis + Celery task dispatch and worker execution.
• Real SSH shell inspection through AsyncSSH for host resources.
• AES-GCM encrypted resource credentials and masked inspection output.
• RBAC-protected API routes and audit logs for write/security events.
• Docker Compose production stack with Nginx TLS proxy.
• HTML, DOCX and PDF report export.
• Persistent AI workflow state machine with callback-driven task execution.

With PostgreSQL, Redis, API and worker running:

python3 scripts/check_services.py
python3 -m compileall backend
node --check assets/app.js
python3 scripts/smoke_test.py http://127.0.0.1:4173

Expected smoke output:

smoke ok

See docs/deployment/architecture.md.

For the full delivery-level technical design, see docs/TECHNICAL_DESIGN.md.