Skip to content

Full Pipeline Demo

End-to-end walkthrough of the Orion content pipeline: detect a trend, generate content, review it, publish, and monitor the results. This guide covers both the CLI and the Dashboard so you can use whichever interface you prefer.

Prerequisites

Start the full Orion stack:

cp .env.example .env
docker compose -f deploy/docker-compose.yml up -d

Wait for all services to become healthy:

docker compose -f deploy/docker-compose.yml ps

Demo mode

If you don't have the backend running, you can use demo mode instead. See the Demo Mode guide for setup instructions.

Trigger the Scout service to scan external sources for trending topics.

Scan Frequency

In production, trend scans run on a configurable schedule. For this demo, we trigger a scan manually. You can adjust the automatic scan interval via the SCOUT_SCAN_INTERVAL environment variable.

# Authenticate first
orion auth login
# Email: admin@orion.local
# Password: orion_dev

# Trigger a trend scan
orion scout trigger

Expected output:

Scan triggered successfully.
Sources: google, rss
Region:  US

View detected trends:

orion scout list-trends

Expected output:

┌────────┬──────────────────────────────────────────────┬───────┬────────────────┬────────┐
│ ID     │ Topic                                        │ Score │ Source         │ Status │
├────────┼──────────────────────────────────────────────┼───────┼────────────────┼────────┤
│ t-001  │ AI Agents Replace Junior Devs — Hype or ...  │  0.94 │ google_trends  │ NEW    │
│ t-003  │ Apple Vision Pro 2 Leak Sparks AR/VR Debate  │  0.91 │ twitter        │ NEW    │
│ t-002  │ Rust Adoption Surges in Enterprise Backend   │  0.87 │ rss            │ NEW    │
│ t-004  │ Open-Source LLMs Close the Gap on GPT-5      │  0.82 │ google_trends  │ NEW    │
│ t-006  │ WebAssembly Enters Server-Side Mainstream     │  0.73 │ twitter        │ NEW    │
└────────┴──────────────────────────────────────────────┴───────┴────────────────┴────────┘
  1. Open http://localhost:3001 and log in with admin / orion_dev
  2. Navigate to the Trends page from the sidebar
  3. Trends are displayed with their virality score, source, and status
  4. New trends detected by Scout appear here automatically via WebSocket updates

Step 2: Review Content

Once Scout detects a trend, the Director service automatically picks it up via Redis pub/sub and begins generating content through the LangGraph pipeline (strategist, creator, media, editor stages).

# List all content items
orion content list

# List content in JSON format
orion content list --format json

Expected output:

┌────────────┬────────────────────────────────────────────┬────────────┬───────────────────┐
│ ID         │ Title                                      │ Status     │ Created           │
├────────────┼────────────────────────────────────────────┼────────────┼───────────────────┤
│ c-a1b2c3d4 │ AI Agents: Hype vs Reality in 2026        │ review     │ 2026-03-18 09:15  │
│ c-e5f6a7b8 │ Why Rust Is Taking Over Enterprise        │ generating │ 2026-03-18 09:20  │
│ c-c9d0e1f2 │ Vision Pro 2: What We Know So Far         │ generating │ 2026-03-18 09:22  │
└────────────┴────────────────────────────────────────────┴────────────┴───────────────────┘

View full details of a content item:

orion content view c-a1b2c3d4
  1. Navigate to the Content Queue page from the sidebar
  2. Content items are listed with their status (generating, review, approved, published)
  3. Click on any item to see its full details: script, media assets, and metadata
  4. Items in review status are ready for your approval

Step 3: Approve Content

Content goes through a human-in-the-loop review stage before publishing.

# Approve content for immediate publishing
orion content approve c-a1b2c3d4

Expected output:

Content c-a1b2c3d4 approved.
Status: approved -> publishing

You can also reject content:

orion content reject c-a1b2c3d4
  1. Open a content item from the Content Queue
  2. Review the generated script, images, and video preview
  3. Click Approve to send the content for publishing
  4. Or click Reject and provide feedback for regeneration

Step 4: Publish

Once approved, the Publisher service handles distribution to configured social media accounts.

# List connected social accounts
orion publish accounts

# Publish to a specific platform
orion publish send c-a1b2c3d4 --platform twitter

# View publishing history
orion publish history

Expected output:

Publishing c-a1b2c3d4 to twitter...
Published successfully.
Post URL: https://twitter.com/orion_demo/status/1234567890
  1. Navigate to the Publishing page from the sidebar
  2. Approved content appears in the publishing queue
  3. Select the target platform and click Publish
  4. Published items show their external URL and engagement metrics

Step 5: Monitor

Track system health, pipeline performance, and content analytics.

# Overall system status
orion system status

# Detailed health check of all services
orion system health --format json

Expected output for orion system status:

Orion System Status
───────────────────
Mode:       LOCAL
GPU:        Available (NVIDIA RTX 4090, 24GB)
Services:   6/6 healthy
Queue:      3 items pending
Uptime:     2h 15m
  1. Analytics page — content performance metrics, engagement charts, trend correlation
  2. System Health page — service status cards, GPU utilization gauge, queue depth
  3. Real-time updates via WebSocket connections

For deeper observability, see the Monitoring guide for Grafana dashboards and Prometheus metrics.

What Happens Behind the Scenes

The complete pipeline flow:

sequenceDiagram
    participant User
    participant Gateway
    participant Scout
    participant Redis
    participant Director
    participant Media
    participant Editor
    participant Publisher

    User->>Gateway: POST /api/v1/scout/api/v1/trends/scan
    Gateway->>Scout: Proxy request
    Scout->>Scout: Poll external sources
    Scout->>Redis: Publish orion.trend.detected
    Redis->>Director: Deliver event
    Director->>Director: LangGraph pipeline (strategist -> creator)
    Director->>Redis: Publish orion.content.created
    Redis->>Media: Deliver event
    Media->>Media: Generate images (ComfyUI/Fal.ai)
    Media->>Redis: Publish orion.media.generated
    Redis->>Editor: Deliver event
    Editor->>Editor: Render video (TTS + stitch + subtitles)
    Editor->>Redis: Publish orion.content.ready
    Note over User: Human reviews and approves
    User->>Gateway: POST /approve
    Gateway->>Director: Update status
    Director->>Redis: Publish orion.content.approved
    Redis->>Publisher: Deliver event
    Publisher->>Publisher: Post to social platforms

Each service communicates exclusively through Redis pub/sub -- there are no direct HTTP calls between Python services.

Next Steps