Integration of Fine-Tuned GPT Models with Autom Mate

Use Case Hub
, , ,
1

Overview

Autom Mate enables the integration of fine-tuned GPT models hosted on OpenAI into workflows without the need for any coding. With My GPT actions, AI-powered decisions can be made based on user input and integrated into business processes.

This architecture:

  • Uses Mate DB to store session data for maintaining conversation continuity,
  • Seamlessly integrates with external platforms such as Microsoft Teams to deliver GPT responses directly to users,
  • Embeds GPT-driven interactions into automation flows to create dynamic and context-aware systems.

Trigger Type

Initiates the workflow based on an external input using a Webhook trigger model, typically from Microsoft Teams or other integrated platforms.

  • Commonly used to detect when a user submits a question or prompt.
  • Serves as the starting point for invoking GPT-related actions.

Available Actions

1. THREAD-SCOPE

Starts a new conversation or resumes an existing session with GPT.

  • Custom: Uses a predefined thread ID for persistent sessions.
  • Generate: Creates a new thread ID dynamically for each execution.

2. MESSAGE

Sends a user message to the GPT model. Often triggered by external systems.

3. RUN

Processes the message and retrieves a response from the GPT model.

  • Assistant ID : The identifier of the fine-tuned model on OpenAI. It must be configured in the Run action.
  • The response is stored in the Response variable.

Action order must be: THREAD-SCOPE → MESSAGE → RUN . Misconfiguration can result in session context loss.

4. SEND-MESSAGE

Sends the GPT response to the user through platforms like Microsoft Teams.

  • Uses the content of the Response variable.
  • Supports plain text or adaptive card formats.
  • Requires a valid MS Teams connection set up in Vault.

5. Mate DB Actions

Used to store user-specific session data (e.g., thread_id ).

  • EXISTS : Checks if a key exists.
  • GET : Retrieves the value of a key.
  • SET : Stores a new key-value pair.
  • DELETE : Deletes a specified key.

Example Scenario: IT Helpdesk Bot

User: “How do I reset my VPN password?”

Flow:

  1. Check if thread_id exists in Mate DB.
  2. If not, generate a new thread (Generate ).
  3. Send the user message to GPT.
  4. Get the response → Response
  5. If the response includes “Goodbye!”, delete the thread (DELETE ).
  6. Send the response to the user via Microsoft Teams.

Benefits

  • High accuracy with organization-specific AI models
  • Context-aware responses tailored to users
  • Direct integration with platforms like Teams and Whatsapp
  • Fast deployment with no-code implementation

:warning: Integration Guidelines

  • Maintain the correct action sequence.
  • Ensure Assistant ID and OpenAI API Key are properly defined in Vault.
  • Missing or invalid thread_id can break conversation context.
  • Deleting the thread resets the session.

:light_bulb: Have you built a similar helpdesk assistant or a different GPT-powered workflow? Share your use case examples or questions in the comments—let’s collaborate to improve and inspire new solutions across the community!

Autom View
Autom View491Ă—804 30.6 KB

MyGPT_Autom.json (28.8 KB)

1 Like