distill-backend-service

Distill Backend Service Framework

Build microservices for the Distill content aggregation platform following standard patterns for service discovery, authentication, and data synchronization.

When to Use This Skill

Use this skill when:

• Creating a new Distill service for a content platform
• Implementing watch/unwatch user lifecycle management
• Building AI-powered content summaries and feeds
• Adding standard API endpoints for service discovery
• Implementing JWT-based authentication with Clerk
• Setting up sync scheduling and rate limiting

Core Architecture

Distill services are independent microservices that:

1. Collect content from external platforms (Twitter, Email, GitHub, YouTube, etc.)
2. Process and store data locally
3. Generate AI summaries and insights
4. Expose LLM-optimized discovery endpoints

Quick Start

Required Endpoints

Every Distill service MUST implement:

Endpoint	Method	Auth	Purpose
/api	GET	None	Service documentation for LLMs
/capabilities	GET	None	Service features and requirements
/health	GET	None	Health check
/watch	POST	JWT	Start monitoring user
/unwatch	POST	JWT	Stop monitoring user
/sync	POST	JWT	Trigger immediate sync
/metadata	GET	JWT	Discover available data

Authentication

All protected endpoints require JWT bearer token:

Authorization: Bearer <jwt_token>

Required JWT claims:

{
  "user_id": "user_123",
  "email": "[email protected]",
  "credits_available": true
}

Implementation Workflow

1. Set up service structure

   • Create standard endpoint handlers
   • Configure JWT validation with Clerk
   • Set up local storage (filesystem, PostgreSQL, or Redis)

2. Implement user lifecycle

   • POST /watch: Register user for monitoring
   • POST /unwatch: Stop monitoring and cleanup
   • Track user status and sync history

3. Build sync logic

   • Scheduler for periodic syncs
   • Rate limiting for platform APIs
   • Incremental data collection
   • Error handling and backoff

4. Add data endpoints

   • /content/recent: Latest content
   • /content/summary/{period}: AI-generated summaries
   • /content/search: Search collected content

5. Generate AI summaries

   • Process collected content
   • Generate daily/weekly digests
   • Tag and categorize content

Standard Response Formats

Service Documentation (GET /api)

{
  "service": "platform-feeder",
  "version": "1.0.0",
  "description": "Collects and processes content from [Platform]",
  "authentication": "Bearer JWT required",
  "endpoints": { ... },
  "usage_example": "...",
  "rate_limits": { ... }
}

Watch Response

{
  "status": "watching",
  "since": "2024-01-10T10:00:00Z",
  "next_sync": "2024-01-10T10:30:00Z",
  "backfill_status": "queued"
}

Metadata Response

{
  "user_status": {
    "watching": true,
    "since": "...",
    "last_sync": "...",
    "next_sync": "..."
  },
  "available_endpoints": [
    {
      "path": "/content/recent",
      "method": "GET",
      "description": "...",
      "item_count": 47
    }
  ],
  "statistics": { ... }
}

Error Response

{
  "error": "error_code",
  "message": "Human-readable message",
  "retry_after": "2024-01-10T11:00:00Z",
  "user_action_required": false
}

Standard Error Codes

Code	HTTP	Description
not_watching	404	User not registered
no_platform_auth	403	Platform credentials missing
token_expired	401	Platform token needs refresh
rate_limited	429	Platform API rate limit
insufficient_credits	402	User out of credits
sync_in_progress	409	Sync already running

Storage Patterns

Choose based on needs:

Filesystem (simple, good for prototyping):

/data/users/{user_id}/
  raw/
    2024-01-10/content.json
  processed/
    summaries/daily_2024-01-10.json
  metadata.json

PostgreSQL (complex queries):

CREATE TABLE user_content (
  user_id VARCHAR(255),
  platform_id VARCHAR(255),
  collected_at TIMESTAMP,
  content JSONB,
  processed BOOLEAN DEFAULT false
);

Redis (rate limiting, caching):

rate_limit:{user_id} → remaining requests
last_sync:{user_id} → timestamp
cache:{user_id}:recent → serialized items

Security Requirements

1. Never log tokens or sensitive data
2. Validate JWT signatures on every request
3. Sanitize data from external platforms
4. Rate limit incoming requests per user
5. Encrypt stored platform credentials
6. Implement request timeouts

Quality Checklist

Before deploying a Distill service:

• All standard endpoints implemented (/api, /capabilities, /health, /watch, /unwatch, /sync, /metadata)
• JWT authentication properly configured
• Service self-documentation is LLM-friendly
• Error responses follow standard format
• Rate limiting implemented for platform APIs
• Sync scheduling handles failures gracefully
• Data retention policies enforced
• Security best practices followed
• Mock mode available for testing
• Health monitoring exposed

Reference Documentation

For complete specification including all endpoint schemas, implementation examples, and advanced patterns, see:

• REFERENCE.md - Full Distill Service Framework Specification