Skip to main content

​
What You’ll Get

  • Reuse existing OpenAI / Claude / Gemini clients; you only need to change the Base URL and token to make calls.
  • The platform validates quotas, pre-deducts, and directly rejects when insufficient, preventing overdrafts.
  • Call logs record fields such as models, token names, channels, request latency, and tokens, facilitating troubleshooting and reconciliation.

​
Use Cases

  • Need to quickly verify token availability and quota settings.
  • Want to switch to a unified entry point without modifying existing client code.
  • Need to view request latency, tokens, and consumption to support subsequent reconciliation or troubleshooting.

​
Quick Path: Get It Working

  1. Prepare: Copy your token and note the platform API Base (example: https://app.memorylake.ai).
  2. Choose a Model: If unsure, first retrieve the model list: GET /v1/models (OpenAI-compatible endpoint).
  3. Send a Minimal Call: Follow the official format, only replace the Base URL and token:
curl --request POST \
  --url https://app.memorylake.ai/v1/chat/completions \
  --header 'Authorization: Bearer sk-your-token' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "hi~"}]
  }'
Path conventions: OpenAI endpoint at /v1, Claude at /claude/v1, Gemini at /gemini/:version. All call examples use actual available platform paths.

​
Key Usage Points

  • Quota Rules: Each call validates and pre-deducts quota; requests are rejected when quota is insufficient. Successful returns confirm deduction; failures roll back pre-deduction.
  • Model Selection: Supports OpenAI / Claude / Gemini compatible formats. View currently available models via GET /v1/models or corresponding model list endpoints.
  • Channels and Proxies: Channels can be configured with proxies or multipliers as needed. Final routing distributes based on model, channel configuration, and optional specified channel headers.
  • Special Features: Supports capabilities like o1/o3 ReasoningEffort selection, Claude thinking, Gemini search and code execution, passed through via official parameters.

​
What You’ll See After a Successful Call

  • Log Records: Fields such as model name, token name, channel, request latency (request_time), prompt and completion tokens are recorded for tracking and troubleshooting.
  • Aggregated Statistics: Request counts, consumed quotas, tokens, and request latency are aggregated by date, facilitating dashboard display and quota reconciliation.

​
Common Troubleshooting

  • Insufficient Quota: Verify that token or account quota is sufficient; insufficient quota directly rejects requests.
  • No Available Route: Confirm that the target model is enabled in the current route, or adjust model selection.
  • Specify Route: When a fixed route is needed, specify via token suffix or specific header; if not specified, default automatic routing applies.

​
You Might Also Want to Know

  • Want to see more examples or special parameter usage: Check the Protocols and Entry Points documentation.
  • Want to verify billing: Reconcile using log fields and aggregated statistics; use a test token first to verify multipliers and model wildcards if needed.