/
Enable and setup the AI backend
  • Verified

    • Enable and setup the AI backend

    Owned by PIPEFORCE
    Last updated: Jan 12, 2025
    3 min read

     

    By default PIPEFORCE comes with it’s own built-in AI backend hosted in Germany which complies with the GDPR rules and uses the default secret ai-default-secret.

    PIPEFORCE can also combine and pipeline multiple AI backends and models in order to use exactly the combination and security level you need to deliver your business solution.

    Enable the default AI backend

    In case you would like to enable all the AI features and access the trained models in order to integrate them into your business solutions, you have to make sure you have enabled the PIPEFORCE AI backend by adding the secret ai-default-secret in the secrets section. Once this secret was created, you can access the AI backend and all provided LLM models without any further setup required.

    If not already exists, you can obtain the ai-default-secret from the PIPEFORCE support. But usually this is already done for you in case you have PIPEFORCE AI included in your contract.

    Change the AI backend

    If you do not specify otherwise, the default PIPEFORCE AI backend as configured in the secret ai-default-secret is used in any AI conversation.

    But you can add additional AI backends and use them in parallel. To do so, you have to add a secret for each AI backend you would like to additionally use.

    This secret must be of type secret-text and must contain a JSON value which has a structure like this:

    { "base_url": "string", "model": "string", "api_token": "string", "max_token": "integer", "custom_headers": { "name": "value" } }

    Whereas:

    • base_url: The base url of the API (requried).

    • model: The AI model to be used (required).

    • api_token: The secure API token to be used.

    • max_token: The max prompt tokens allowed to be sent (defaults to 800 if missing)

    • custom_headers: Optional key-value pairs to be passed along as HTTP headers on any request. This is handy for example in case basic authentication or any other additional header setting is required.

    Connect to OpenAI (ChatGPT)

    In case you would like to integrate OpenAI as additional AI backend for example, you could use these settings in your secret:

    { "base_url": "https://api.openai.com/v1", "model": "gpt-3.5-turbo", "api_token": "your_token", "max_token": 800 }

    Replace your_token by the API token you can generate on the OpenAI website.

    Use the AI backend secret in your AI commands

    In any of the AI commands, if no secret is specified, the default AI backend is used. You can refer to a different AI backend, by using its secret in the AI command.

    Lets assume the OpenAI secret from above was added under a secret with name openai-secret, then to prompt this OpenAI backend from inside your pipeline you can do something like this:

    pipeline: - ai.prompt.send: secret: openai-secret prompt: "Tell me a joke"

    This works the same way with any other AI command.

    Chaining multiple AI backends in a pipeline

    You can combine multiple AI backends in a single pipeline. Lets assume this example: There is one AI backend which is optimized to extract fields from an invoice and a second one to remove all privacy data, then you can combine them easily using a pipeline like this:

    What happens here is:

    1. The invoice PDF is loaded and automatically converted to an AI compatible format.

    2. The prompt is sent with the information extracted from the PDF to the first ai-invoice-backend.

    3. The result of the first prompt is then passed automatically to the next ai-privacy-backend.

    This way you can combine the power of multiple AI backends to leverage your result.