AICOMBO
EN
  • CN
  • EN
EN
  • CN
  • EN
EN
  • CN
  • EN
    • User Guide
    • Audio
      • Text to Speech
        POST
      • Gemini Native
        POST
    • Chat
      • Native OpenAI Format
        • ChatCompletions Format
        • Responses Format
        • Responses Async Task Retrieval
        • Responses Async Task Deletion
        • Responses Async Task Retrieval (Streaming)
      • Native Gemini Format
        • Gemini Chat
      • Native Vertex AI Format
        • Gemini Chat
      • Native Anthropic Format
        • Native Claude Format
        • Native Claude Format - Token Count
    • Completions
      • Native OpenAI Format
        POST
    • Images
      • Gemini Format
        • Responses Format
        • OpenAI Compatible
    • Models
      • List Models
        • Native OpenAI Format
        • Native Gemini Format

    User Guide

    AICombo User Guide#

    From sign-up and token management to API integration.
    Replace the platform address with your API Base URL, use a token issued by the platform, and you can start calling models.
    This guide is organized around the actual user journey — ideal for first-time integration, migrating an existing application, and pre-launch checks.

    Table of Contents#

    1.
    Sign Up & Log In
    2.
    Token Management
    3.
    Using the API
    4.
    Personal Settings
    5.
    Usage Logs

    Fastest Path to API Integration#

    Create a token → Verify in the Playground → Copy the API Base URL → Choose a protocol → Make your first call → Configure production

    1. Sign Up & Log In#

    Please register an account before first use. After successful registration, you will be redirected to the home page automatically.

    Log In#

    Open the platform home page and click "Log In" in the top-right corner.
    Enter your username and password, then click "Log In".
    If you forget your password, click "Forgot Password", enter your registered email, and set a new password via the reset link.

    Sign Up#

    Click "Sign Up" on the login page.
    Fill in a username and password.
    Enter your email, click "Send Verification Code", and enter the code you receive.
    Click "Sign Up" to complete account creation.

    2. Token Management#

    Tokens are the credentials for calling the API. Each token can be configured independently with its own scope, quota, and usage limits. Open "Token Management" in the left sidebar to view them.

    Create a Token#

    Click "Add Token".
    Enter a token name — naming by purpose is recommended, e.g. "Production" or "Testing".
    Configure expiration time, unlimited quota, model restrictions, IP allowlist, and group as needed.
    Click "Submit" and copy the full key immediately; it cannot be viewed again after the dialog is closed.
    The token list shows the name, status, total quota, remaining quota, and expiration time. Editing a token does not change its key; deleting a token takes effect immediately and cannot be undone.
    Security tip: store keys in a server-side secret management system; never commit them to a code repository.

    3. Using the API#

    This chapter walks through, from the user's perspective, how to obtain credentials, verify your configuration, and integrate the platform into common development environments. Completing the first two steps is enough to make your first call.

    Before You Start#

    The gateway API Base URL, e.g. https://api.example.com.
    A created and still-valid token key.
    The target model ID — confirm availability in the model list first.

    3.1 Verify Your Token Online#

    The Playground is the quickest way to confirm that your token, model, and gateway all work — without writing any code.
    Click "Playground" in the left navigation.
    Select the model to test on the left.
    Type a message at the bottom and click send.
    View the model's reply in the conversation area.
    If the call fails, check the token status, model restrictions, remaining quota, and model ID.

    3.2 Get the API Base URL#

    Visit the platform home page and click copy in the API Base URL section. Endpoint paths already include the version prefix — do not append /v1 or /v1beta again when concatenating.
    Example addresses:
    https://<YOUR_API_HOST>
https://<YOUR_API_HOST>/v1/responses
https://<YOUR_API_HOST>/v1/models

    3.3 Choose How to Call#

    For new projects, prefer Responses; when migrating an existing OpenAI, Anthropic, or Gemini application, you can keep using the original protocol.
    Your scenarioRecommended entry pointAuthentication
    OpenAI-compatible SDK, text chatPOST /v1/responses or POST /v1/chat/completionsAuthorization: Bearer <API_KEY>
    Anthropic / Claude native callsPOST /v1/messagesBearer Token + anthropic-version
    Vertex / Gemini native callsPOST /v1/publishers/google/models/{model_name}:{action}x-goog-api-key or ?key=
    Query available modelsGET /v1/models or GET /v1beta/modelsPer target protocol

    3.4 Your First API Call#

    Send a non-streaming request with the Responses endpoint:
    A successful request returns a model response object. In production, parse it according to the response data structure of the corresponding endpoint.

    3.5 Integrate in Code#

    Python (OpenAI-compatible SDK)#

    JavaScript (OpenAI-compatible SDK)#

    Environment Variables#

    3.6 Common Capabilities#

    Multi-turn conversations: chain Responses context with previous_response_id, or pass the message history yourself.
    Streaming output: when both the endpoint and the model support it, set stream to true and parse by SSE event boundaries.
    Tool calling: associate tool results with the corresponding call_id to avoid mixing results across turns.
    Multimodal input: use Base64 for small files; prefer controlled URLs or file URIs for large files, and set the correct MIME type.
    Image generation & editing: explicitly specify the model, size, quality, and response format; never write Base64 data directly into logs.
    Text-to-speech: OpenAI and Gemini differ in parameters and audio formats — do not reuse the same post-processing logic for both.

    Main Supported Endpoints#

    CapabilityMethod & PathDescription
    Text-to-speechPOST /v1/audio/speechConvert text to audio
    Model responsesPOST /v1/responsesMulti-turn conversations, tool calling, and reasoning configuration
    Gemini chatPOST /v1/publishers/google/models/{model_name}:{action}Vertex AI native format
    Claude chatPOST /v1/messagesAnthropic Messages native format
    Text completionPOST /v1/completionsLegacy completion endpoint
    EmbeddingsPOST /v1/embeddingsCreate text vector embeddings
    Image generation / editingPOST /v1/images/generations/ or /edits/Generate or edit images
    Model listGET /v1/models or GET /v1beta/modelsQuery available models
    BatchPOST /v1/vertex/batch/jobsCreate and query Vertex batch jobs

    3.7 Troubleshooting Errors#

    SymptomCheck first
    401 / authentication failedWhether the API key is complete and the correct auth header is used
    Model unavailableModel ID, token model restrictions, and the model list result
    429 / too many requestsConcurrency, rate limits, account balance, and token quota
    400 / invalid parametersPath, version prefix, required parameters, and the model's supported range
    Abnormal streaming contentWhether parsing follows SSE event boundaries
    Failures that are generally safe to retry include 408, 429, 500, 502, 503, and 504. Use exponential backoff with random jitter and cap the maximum number of retries. Generative POST requests should only be retried automatically when duplicate results are acceptable to your business or an idempotency mechanism is in place.

    4. Personal Settings#

    After logging in, click your avatar in the top-right corner and open "Personal Settings". Here you can manage your account, security authentication, and notification channels.

    Common Settings#

    Bind an email address or change your password.
    Enable 2FA with an authenticator app for stronger login security.
    Register a Passkey to log in with fingerprint, face recognition, or a security key.
    Configure email, Webhook, Bark, or Gotify notifications.
    Set pricing strategy, IP logging, and sidebar display options as needed.

    5. Usage Logs#

    Open "Usage Logs" in the left sidebar to view the time, model, token consumption, quota, and status of every API call.

    Filtering & Statistics#

    Filter records by time range, model, and token name.
    Click "Query" to refresh the filtered results.
    Open the "Dashboard" to view daily API call volume and quota consumption trends.

    Pre-launch Checklist#

    API keys are injected via environment variables or a server-side secret management system.
    The target model has been verified as available.
    Connection, read, and overall request timeouts are configured.
    Concurrency, rate, and budget limits are in place, with bounded retries configured.
    Logs do not record API keys, full Base64 payloads, or sensitive prompts.
    Timeouts, disconnections, cancellations, and stream termination states are all handled.
    下一页
    Text to Speech
    Built with