root@thinkprompt-proxy:~#

AI-ThinkPrompt Centralized LLM Router Service (v1.0.0)

▶ System Overview

This service acts as an intelligent, stateless router sitting between your downstream client apps and upstream LLM providers (e.g., OpenRouter, Google GCP). It automatically handles API key rotation, Rate Limit 429 cooling, and multi-tier cascading failovers (Free → Paid → Emergency).

▶ Integration Protocol

To integrate, point your standard OpenAI-compatible SDKs or `fetch` calls to this exact endpoint:

POST https://ai.thinkprompt.com/ai/api/llm

▶ Authorized App Clients ("app_id")

To prevent unauthorized external abuse and track internal usage, the proxy requires an Authorized App Client. In your Admin settings, you can generate App Clients. Each gives you a unique identifier.

You must pass this identifier in your requests using the app_id HTTP header. This logs latency and errors back to your specific internal application.

▶ cURL Implementation Example

curl "https://ai.thinkprompt.com/ai/api/llm" \
  -X POST \
  -H "Content-Type: application/json" \
  # Pass your designated Internal App ID here (Authorized App Clients)
  -H "app_id: router-client-1" \
  -d '{
    "messages": [
      { "role": "system", "content": "You are a helpful proxy." },
      { "role": "user", "content": "Hello World!" }
    ],
    # Stream is natively piped back to bypass Vercel serverless timeouts
    "stream": true 
  }'

_ waiting for incoming requests...