How to build a Custom API Chatbot

Workflow Overview

1. User Question (Input Node)

• Purpose:
  This is where the user enters their portfolio performance question.

  • Example: “How did my tech stocks perform last quarter?”

2. OpenAI (LLM Node)

• Purpose:
  This node takes the user’s question and interprets it using a large language model (OpenAI).

  • What it does:

    • Understands the user’s intent.

    • Translates the question into a format suitable for querying your Airtable database.

    • Builds an Airtable filter formula or query based on the user’s request.

    • The LLM is also configured to reference specific tools for database queries (like Airtable’s database query action).

  • Special configuration:

    • The LLM is set up with a system prompt that makes it act as a data analyst for a multi-strategy fund.

    • It is aware of the available database tools and can use them to fetch or filter data as needed.

3. Output (Output Node)

• Purpose:
  This node displays the final answer to the user.

  • What it shows:

    • A summary and, if relevant, a table with the results of the portfolio performance query.

How the Data Flows

1. User enters a question in the “User Question” input node.

2. The OpenAI node receives the question, interprets it, and formulates a database query.

3. The OpenAI node may use Airtable database tools (as referenced in its configuration) to fetch the relevant data.

4. The resulting answer is sent to the Output node, which presents the summary and any relevant data to the user.

Summary Table