When an assistant calls a tools, give the option to pass the thread_id so that the tool can add files to the code interpreter

[owengo]

An issue with custom gpts is that tools do not allow data transfer.
Let's say you want to create a tool which will generate a big .csv file you want then to process with the code interpreter, you have to give it in the tool response so it is extremly slow, inefficient and limited by the size of context.

A solution for this would be to provide the tool ( the function described in the openapi spec ) with the thread_id of the discussion, for example using a custom header. This way the remote function would be able to:

• generate an output file
• upload it using the file api
• update the thread's tool_resoures so that the code interpreter can access the generated file

This feature would effectively allow actions to pass efficiently big data files to the assistant's code interpreter without depending on the model's context size. It would be unique to Librechat since as of today, AFAIK, OpenAI does not support it, even for enterprise customers.

Exemple:

Imagine the following api with the get_monthly_data action. The action takes a query parameter "date" and requires the header "x-openai-thread-id".
The implementation would generate the csv data, upload it as "monthly_data_at_{date}.csv", update the thread passed as x-openai-thread-id and make a response like this: "data for {date} is in /mnt/data/monthly_data_at_{date}.csv".
Librechat would just have to detect the action requires the header x-openai-thread-id and instanciate it when executing the action.

openapi: 3.1.0
info:
  title: Monthly Data API
  version: 1.0.0
  description: API to retrieve monthly data based on a specified date.
paths:
  /get_monthly_data:
    get:
      summary: Get Monthly Data
      description: Retrieve data for the specified month.
      parameters:
        - name: date
          in: query
          required: true
          description: The date string representing the month for the data (e.g., YYYY-MM format).
          schema:
            type: string
            format: date
        - name: x-openai-thread-id
          in: header
          required: true
          description: The thread ID of the current run.
          schema:
            type: string
      responses:
        '200':
          description: Successful response with the path to the monthly data.
          content:
            application/json:
              schema:
                type: object
                properties:
                  data:
                    type: string
                    description: The path to the monthly data file corresponding to the requested date.
        '400':
          description: Invalid date format or missing parameters.
        '401':
          description: Unauthorized request due to missing or invalid thread ID.
        '500':
          description: Internal server error.