Skip to main content

Use Flow Knowledge Base in Custom LLMs

LLMs can help with tasks such as writing, coding, and answering questions. It adapts to context and user input to provide relevant, conversational responses. They can be integrated into developer tools or workflows to assist with documentation, debugging, and productivity.

This guide walks you through creating a Custom LLM, such as ChatGPT's GPTs or Gemini's Gems, that can reference the Flow Data Sources file to answer questions.

warning

You'll need a ChatGPT Plus subscription to use the Custom GPT feature or a Gemini Pro subscription to use the Custom Gem feature.

📍 Step 1: Open the "Custom LLM" Section​

  1. Log in to ChatGPT or Gemini
  2. In the sidebar on the left
  • If using ChatGPT, click on GPTs
  • If using Gemini, click on Explore Gems
Screenshot 2025-06-12 at 15 12 17 Screenshot 2025-06-12 at 15 11 56

📍 Step 2: Click "Create a GPT"​

  • In the GPTs screen, click the "Create" button in the top-right corner.
  • In the Explore Gems screen, click the New Gem button in the center-right of the screen.
Screenshot 2025-06-12 at 15 15 18 Screenshot 2025-06-12 at 15 15 33

📍 Step 3: Walk Through the Agent Coniguration​

  • If using ChatGPT, click on the Configure button before proceeding.
  1. Fill the Name and description of your custom LLM.
    • You can name the agent FlowDocs and set the description as: Provides knowledge about building on Flow.
  2. Instructions: Tell it how to behave and what to prioritize (e.g., always reference the uploaded document)
  3. Capabilities: Enable file browsing, code interpreter, or DALL·E if needed

Suggested Prompt​


_10
I want to make an agent called FlowDocs that uses the linked file as its primary source.  This file changes, so it should reference the live file at least once a day:  https://github.com/onflow/Flow-Data-Sources/blob/main/merged_docs/all_merged.md
_10
_10
Please imagine you are a fast and smart junior developer who is eager to help and has memorized all the information in the linked file

Please let us know if you find any other useful customization prompts!

  • You can further customize your personalized agent by providing more files and determining the actions it can do.

📍 Step 5: Test Your GPT​

Once the personalized agent is built, you'll be taken to a preview chat window. Test it by asking a few questions based on your uploaded document.

📍 Step 6: Save and Publish (Optional)​

When you're ready:

  • Click "Update & Save" to finalize
  • You can choose to keep it private or make it public

✅ That's it!​

You've now created a custom GPT that references your uploaded file as a primary source. You can update the file or instructions later if needed.

Rate this page