06. API and Infrastructure
This document outlines the authentication strategy, rate limiting mechanisms, and the zero-cost architecture powering Omni-Post AI.
Free-Tier API Strategy
How Costs are Kept at $0/Month:
| Service | Free Tier | Expected Usage | Cost |
|---|---|---|---|
| Gemini (Flash) | 1500 requests/day | High volume nodes | $0 |
| Gemini (Pro) | ~100 requests/day | ~20-30/day | $0 |
| Tavily API | 1000 credits/month | ~40-60 credits/mo | $0 |
| Twitter API | 450 posts/month | ~20-30/month | $0 |
| LinkedIn API | ~100 req/24hrs | ~20-30/month | $0 |
| Threads API | Rate limited | ~20-30/month | $0 |
| Dev.to | ~30 req/30sec | ~20-30/month | $0 |
| Hashnode | 500 req/min | ~20-30/month | $0 |
| Google Drive | 1TB storage | <1GB | $0 |
| Notion API | 3 requests/sec | ~100 requests/day | $0 |
| Sanity CMS | 100K requests/month | ~30/month | $0 |
| Cloudflare Tunnel | Unlimited | 24/7 | $0 |
| Obsidian MCP | Local compute | Unlimited | $0 |
Scalability: This setup handles 100+ content pieces/month comfortably before hitting any free-tier limits.
OAuth2 Implementation
Omni-Post relies on n8n's built-in OAuth2 handling for all connected platforms.
Twitter/X API:
- Flow: 3-Legged OAuth2
- Tier: Free tier (450 posts/month, 50 requests/15 minutes)
- Scopes:
tweet.write,tweet.read
LinkedIn API:
- Flow: 3-Legged OAuth2
- Tier: Free tier (organic posts only, 100 req/24 hours)
- Scopes:
w_member_social - Note: Enforces a strict 1-image maximum (API fails silently on >1).
Reliability Strategy:
Critical posting nodes use retryOnFail: true. n8n handles token refreshes automatically upon encountering a 401 Unauthorized response. Manual re-authentication is available via the n8n UI if the refresh token expires.
Rate Limiting & Quotas
Current Approach
Rate limits are currently handled via n8n's built-in retry mechanisms:
javascript
Future Enhancement: Proactive Token Refresh & Exponential Backoff
For scaled production systems, future developments will include:
- Scheduled Preemptive Refresh: A cron-triggered workflow running every 4 hours to preemptively refresh tokens before expiry.
- Exponential Backoff Custom Functions: Instead of linear retries, implement
delay = Math.pow(2, attempt) * 1000(e.g., 1s, 2s, 4s, 8s, 16s) to handle HTTP 429 errors from high-traffic platforms.
Internal Network Connectivity
- n8n Hosting: Running via a self-hosted Docker container, exposed securely via Cloudflare Tunnel.
- Obsidian MCP Bridge: Connects over
http://host.docker.internal:3010/mcpto the locally running MCP server. The Obsidian application must be open for the MCP context overlay to successfully return data; otherwise, the workflow gracefully degrades to the static Portfolio API logic.