Skip to main content

​
Background and Value

  • The platform uses API Key as the sole call credential. All requests first validate quota, expiration, and status to prevent overdrafts or abuse.
  • Compatible with OpenAI / Claude / Gemini call formats, allowing you to reuse existing clients by only replacing the Base URL and token.

​
Use Cases

  • First-time integration, wanting to confirm how to carry tokens and call paths.
  • Teams or multiple clients sharing a unified entry point, requiring controlled quotas.
  • Want clear feedback when quota is insufficient or routes are unavailable, facilitating self-diagnosis.

​
Core Capabilities

  • Token Management: Generate API Keys in the console, where you can set quota, expiration time, and enabled status. Requests first validate these conditions; unmet conditions result in direct rejection.
  • Quota and Pre-deduction: Each request pre-deducts quota; successful returns confirm deduction, failures roll back pre-deduction, ensuring quota accuracy.
  • Multi-model Compatibility: Supports path and parameter formats for OpenAI /v1, Claude /claude/v1, and Gemini /gemini/:version.
  • Common Header Rules: Carry tokens according to compatible endpoints:
    • OpenAI compatible: Authorization: Bearer sk-xxx
    • Claude compatible: x-api-key: sk-xxx
    • Gemini compatible: x-goog-api-key: sk-xxx
  • Log Visible Fields: Calls record models, token names, channels, request latency (request_time), prompt/completion tokens, etc., facilitating troubleshooting and reconciliation.

​
Key Usage Points

  1. Obtain Token: Create an API Key in the console, set quota and expiration; ensure status is enabled.
  2. Choose Endpoint and Base: Replace the client’s API Base with the platform address, e.g., https://app.memorylake.ai, following the corresponding compatible endpoint paths.
  3. Include Header: Fill in the token header according to the common rules above; keep other parameters in official format.
  4. Send Request: You can first call GET /v1/models or corresponding model list endpoints to confirm currently available models.
  5. View Results: After successful requests, view models, channels, latency, and tokens in the logs; aggregated statistics summarize usage and quota consumption by date.

​
Common Authentication and Quota Notes

  • Insufficient Quota: Token quota insufficiency results in direct rejection; check quota or adjust multiplier and retry.
  • Token Invalid: Expired or disabled tokens reject requests; update token status or regenerate.
  • No Available Route: When target models are unavailable in the current route, a prompt is shown; switch models or adjust route configuration.
  • Specify Route: If a fixed route is needed, specify via token suffix or specific header convention (if configured); if not specified, default routing applies.

​
Limitations and Notes

  • Only API Key is supported as call credential; does not involve account login methods.
  • Quota validation covers quota, expiration, and status; pre-deduction rolls back on failure.