Add client tracking and analytics for multiple apps

[ebowwa]

Feature Request: Client Tracking System

Summary

Implement client identification and tracking to monitor usage across different applications using the CaringMind API server.

Known Clients

• CaringMind iOS App - IP: 100.86.17.119 (mobile)
• PrintPeer Web App (Vite) - TBD
• Future clients - Need flexible system

Current State

Server logs show IP addresses but don't identify which client/app is making requests:

INFO:     100.86.17.119:0 - "POST /api/chat/completions HTTP/1.1" 200 OK

Proposed Implementation

Option 1: Server-side tracking (Recommended)

Add client tracking module to CaringMind server:

# backend-python/src/client_tracker.py
class ClientTracker:
    def track_request(self, ip, user_agent, endpoint, model):
        # Identify client from IP, User-Agent, or API key prefix
        # Track: last_seen, request_count, features_used, models_used
        # Store in Redis/SQLite for persistence

Option 2: Extend ai-proxy-core telemetry

• Could add to existing telemetry in ai-proxy-core
• But cleaner to keep deployment-specific tracking separate

Tracking Features Needed

• Client identification (by IP range, User-Agent, API key prefix)
• Request counting per client
• Feature/endpoint usage tracking
• Model usage per client (Gemini 2.0, 1.5, etc.)
• Error rate per client
• Audio vs text request ratio
• Response time analytics per client

Benefits

• Monitor which apps are actively using the API
• Track feature adoption across different clients
• Identify performance issues per client
• Better capacity planning
• Usage-based insights for each app

Quick Implementation for MVP

For now, could just add client identification to logs:

if request.client.host.startswith("100.86.17"):
    client_name = "ios-caringmind"
elif "printpeer" in request.headers.get("origin", "").lower():
    client_name = "vite-printpeer"
logger.info(f"Client: {client_name} - {endpoint}")

Related

• Doppler secrets management (Add OpenAI API support #6)
• Uptime Kuma monitoring (existing)
• Could integrate with existing telemetry infrastructure