OhMyGPT API Docs
    OhMyGPT API Docs
    • About This site
    • About new website
    • Midjourney Open API
      • About Midjourney Service
      • Midjourney Blend Image Generation Task Creation
        POST
      • Midjourney Text-to-Image Task Creation
        POST
      • Midjourney Execute Action Task Creation
        POST
      • Midjourney Query Specific Task ID Execution Status
        POST
      • Midjourney Paginated Bulk Query Tasks
        POST
      • Midjourney Describe Prompt Extraction Task Creation
        POST
      • Retrieve Images Generated by Midjourney from Object Storage
        GET
      • Retrieve Compressed Images Generated by Midjourney from Object Storage + WSRV
        GET
      • Retrieve Images Generated by Midjourney from Object Storage + WSRV
        GET
      • Proxy Discord Images via WSRV
        GET
      • Proxy and Compress Discord Images via WSRV
        GET
    • Management API
      • Create a New API Key
        POST
      • Update API Key Information
        POST
      • Query All of User's API Keys
        POST
      • Delete Specified API Key
        POST
      • Query User Membership Expiration Date
        POST
      • User Balance Query
        POST
      • Query the current model rates
        GET
    • OpenAI
      • About OpenAI Compatible API
      • About Azure OpenAI API
      • Audio
        • Text To Speech API
      • Chat
        • Chat Completion
        • Completion (Legacy)
      • Images
        • Create Image
      • Realtime API
    • Anthropic Claude
      • Deprecated
        • ChatCompletions
        • Claude2 API
      • Anthropic Official Claude API
    • Policies
      • Terms of Service
      • Privacy Policy
    • Other OpenAPIs
      • Search API
        • Google Custom Search API
        • Google Search By Serper
      • Azure API
        • Get Azure TTS voice list
        • Azure TTS

    About This site

    About This Site#

    This site provides various popular and powerful generative AI services.
    Project URL: https://www.ohmygpt.com
    Mirror URLs:
    https://x.dogenet.win
    https://aigptx.top
    https://textai.buzz
    This is the public API documentation of this site, where you can view the API documentation for development and debugging purposes.
    OpenAI's Audio, Chat, Completions, Embeddings, Images, and Moderations interfaces are all supported. The paths and parameters are consistent with OpenAI's. Here we only list some platform-customized and public API documentation. For specific OpenAI-compatible interfaces, please refer to the official documentation: https://platform.openai.com/docs/api-reference
    In actual use, you only need to select one of the API endpoints below as the base URL to replace https://api.openai.com, or append the path in the following documentation.
    API Endpoint List:
    US Main Site Direct Connection Stable, Unavailable in Some Regions: https://api.ohmygpt.com
    Cloudflare CDN Stable, Global Acceleration: https://cfcus02.opapi.win
    Cloudflare Worker Stable, Global Acceleration: https://cfwus02.opapi.win
    High-quality Route Reverse Proxy 1 High-quality Route, Availability Not Guaranteed: https://aigptx.top
    High-quality Route Reverse Proxy 2 High-quality Route, Availability Not Guaranteed: https://cn2us02.opapi.win
    Note:
    Since the servers are located in the United States, network optimization for other regions of the world may not be ideal. Therefore, when your calling location is far from the United States (such as Hong Kong and Singapore in Asia), it is recommended to connect through Cloudflare CDN or reverse proxy API endpoints.
    When using non-streaming request interfaces, it is recommended to use direct connection or reverse proxy endpoints, because the response timeout limit of direct connection and reverse proxy endpoints is relatively relaxed, while Cloudflare's timeout limit is 100s. Exceeding it will cause the connection between you and the server to be disconnected.
    However, when using streaming call API interfaces, you will not be affected by this timeout limit of Cloudflare.

    How to use multimodal models with image input?#

    vision#
    Although the built-in chat on the website does not provide Vision functionality, you can call it through the open API provided by this site in conjunction with third-party open-source/closed-source software.

    LobeChat#

    LobeChat instance hosted by this site: https://lobe.dogenet.win
    Click the link to enter the website, then click on the settings in the lower right corner.
    https://lobe.dogenet.win/settings/llm
    Configure it as shown in the image:
    1710034330061.png
    Among them:
    API Key: Fill in the API Key generated in the "Settings" of this site.
    API Proxy Address: Fill in https://api.ohmygpt.com/v1
    Custom Model Name: You can fill in models from the Claude series, such as claude-3-opus,claude-3-sonnet
    Then you can start using it:
    Multimodal models (such as GPT4V, GeminiProV, Claude3, etc.) support image input, with effects as follows:
    1710034563396.png
    1710034605156.png
    If you encounter a 401 error, it is usually because your account has insufficient balance or the API key does not have permissions for the corresponding model. Please check your account balance or update the permission list of the API key, or create a new API key and try again.

    BotGem(AMA)#

    image.png

    Can I directly use my own OpenAI key on this site?#

    No, uploading your own key is not supported on this site.

    How do I use this site's services in chatGPT-next-web?#

    img

    How do I use this site's services in GPT Academic Optimization?#

    1. Ready-made:
    This instance is hosted by this site. You can directly fill in the API Key generated in the settings of this site to use it:
    https://ga.dogenet.win
    2. Self-deployment:
    Before using, make two modifications to config.py:
    1.
    Modify the API_KEY on line 11 of config.py to your own key from this site.
    2.
    Modify the API_URL_REDIRECT value on line 38 of config.py to {"https://api.openai.com/v1/chat/completions": "https://aigptx.top/v1/chat/completions"}

    How do I use this site's Claude2 service in GPT Academic Optimization?#

    First, please make sure you need to use the Claude2 service. Second, using this site's Claude2 in GPT Academic Optimization requires more configuration and is more complicated.
    Modify as follows:
    1.
    Modify the API_KEY on line 11 of config.py to your own key from this site.
    2.
    Modify the API_URL_REDIRECT value on line 38 of config.py to {"https://api.openai.com/v1/chat/completions": "https://aigptx.top/v1/chat/completions"}
    3.
    Add the model "claude-2-web" to AVAIL_LLM_MODELS on line 73 of config.py.
    4.
    On line 622 of toolbox.py, modify the code to:
    Please pay attention to the indentation.
    5.
    Add a key-value pair to the model_info variable in request_llm/bridge_all.py:
    Then you can use the claude2 model. Select claude-2-web in the web interface to use it.

    About Memberships and GPT4#

    As shown below, this site offers three plans: Free, VIP, and Premium.
    VIP and Premium can unlock API services for all models (including GPT4).

    How do I get access to GPT4?#

    You need to upgrade your plan to VIP to get access to GPT4. This is the lowest-cost way for you on this site and the way we most recommend you use.
    Upgrading to VIP is simple. You only need to top up 20 RMB on this site to get one month of VIP qualification; topping up 40 RMB can get you six months of VIP qualification. As long as you top up the specified amount, you can get VIP qualification without paying for the VIP qualification itself.

    Is the VIP plan a monthly subscription plan?#

    No, after topping up, you will get tokens to offset the consumption of each service use. The VIP qualification is a gift after you top up, and your token balance will not be deducted because you obtained the VIP qualification.

    I topped up 20 RMB and my VIP qualification expired. How long can I get VIP access if I top up another 20 RMB?#

    5 months.
    The image below can help you better understand our billing mechanism.
    pidPffS.png
    In this situation, your multiple payments met the VIP conditions. Starting from the time of your first payment, you will get the corresponding duration of VIP access time.

    I can't access OhMyGPT.com. Is the website service down?#

    The services of this site are very stable. You can check the specific running status at:
    https://status.ohmygpt.com/status/main
    or
    https://status.aigptx.top/status/main
    The main reason you encounter this situation is DNS pollution and SNI blocking from GFW.
    The solution is to use a proxy or use the mirror site of this site: https://aigptx.top
    Join the Telegram group to get the latest notifications in time: https://t.me/+SfPmr1psSNNhMzg1

    About Cost Consumption#

    How to calculate the cost consumption?#

    Except for midjourney, the actual cost deducted by the system is: the cost calculated by the official AI service multiplied by the current rate.
    For example, if you have successfully requested a GPT4 request worth 1 RMB according to the official pricing, it will be multiplied by the corresponding service rate at that time (for example, 0.5), then the actual cost deducted will be 1*0.5=0.5 RMB.

    What does each column in the billing table represent?#

    promptTokens: The context length used by the model for understanding
    completionTokens: The text length generated by the model
    RURT: Request upload time (in milliseconds ms)
    TPS: Request time (in seconds)
    Free: Whether this request consumes free quota

    What are promptTokens and completionTokens?#

    For language models, each vocabulary is converted from human language to a vector that the model can understand in the form of a "vocabulary table", and each mapping corresponds to a token. In other words, token is the smallest unit for language models to understand text.
    At the same time, the interaction of language models has two steps: receiving text and replying with a response. It can be simply understood that you need to ask the model a question before the model will give you a reply. Among them, the number of tokens in the question asked is promptTokens, and the length of the text replied is completionTokens.
    In addition, the pricing for promptTokens and completionTokens varies for different language models, so they need to be distinguished.

    Why are there a lot of promptTokens when using the web version?#

    Models do not have memory capabilities similar to humans. If you need to relate this question to previous questions, you need to send the previous question content and the model's reply content together to the model. This is the main reason for the large promptTokens value.
    If you need to save costs, reducing the context amount in advanced settings can effectively reduce the size of promptTokens for each conversation.

    Why is there a system role in addition to the user and model roles in the web version?#

    The system role corresponds to the system role in OpenAI's API, and it is used to determine the role the model plays and the function it performs in this session.
    If you need to save costs by reducing context, the system is the last context to be removed. For example:
    If the context length is 0, the message sent with the request will only contain your question.
    If the context length is 1, in addition to your question, the message sent with the request will include the system.
    If the context length is greater than 1, in addition to your question and the system, the message sent with the request will include the most recent conversation to this conversation.
    Therefore, you can supplement some important information for this session in the system.

    About API#

    You can integrate various capabilities into your applications through the open API of this site.

    I'm a beginner. Can you provide a code example for calling the service?#

    When calling the OpenAI service, you can choose to use the official OpenAI Python library or use a more general request library + API for integration, either of which works.
    The following sample code uses this site's service to send a request and obtain the result in a non-streaming way, and then print it on the console.
    First, you need to prepare your key.
    Next
    About new website
    Built with