CloudPilot AI API Reference

CloudPilot AI provides an ultra-low latency inference proxy and autonomous cluster orchestration API. It is 100% wire-compatible with standard OpenAI and Anthropic SDK payload specifications, allowing zero-code refactoring drop-in integration.

1. Quickstart Guide

Install the official Python package or use standard HTTP requests with your API key:

pip install cloudpilot-sdk

2. Authentication

All API requests must contain your secret API key passed in the Authorization HTTP header as a Bearer token:

Authorization: Bearer cp_live_sec_your_api_key_here

POST

/v1/chat/completions

Routes prompt to the target model cluster with automated sub-15ms semantic edge caching and intelligent token arbitrage.

cURL Request Example

curl -X POST https://api.cloudpilotai.org/v1/chat/completions \
  -H "Authorization: Bearer cp_live_sec_99a8x" \
  -H "Content-Type: application/json" \
  -H "X-CloudPilot-Cache: enabled" \
  -H "X-CloudPilot-Fallback: aws-us-east-1" \
  -d '{
    "model": "claude-3-5-sonnet-20241022",
    "messages": [
      {"role": "system", "content": "You are CloudPilot AI DevOps assistant."},
      {"role": "user", "content": "Optimize GPU spot cluster allocation."}
    ],
    "temperature": 0.2,
    "stream": false
  }'

Response Payload (Cache Hit in 11.2ms)

{
  "id": "cp_chat_89af1002x",
  "object": "chat.completion",
  "created": 1739884800,
  "model": "claude-3-5-sonnet-20241022",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "GPU allocation strategy optimized for AWS spot instances."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 28,
    "completion_tokens": 14,
    "total_tokens": 42
  },
  "cloudpilot_telemetry": {
    "cache_status": "HIT",
    "cache_latency_ms": 3.8,
    "total_latency_ms": 11.2,
    "token_reduction_pct": 68.4,
    "edge_node": "aws-use1-pop-04"
  }
}

POST

/v1/cluster/rebalance

Triggers autonomous GPU node reallocation, spot replacement, or multi-cloud hot standby failover.

3. Python SDK Usage

from cloudpilot import CloudPilotClient

client = CloudPilotClient(api_key="cp_live_sec_99a8x")

# Autonomous inference query
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Cluster memory diagnosis"}]
)

print(response.choices[0].message.content)