Dify

Connect a Dify agent-chat application to AI GO! as a native model, using the typed dify connection. Dify keeps conversation state server-side, scoped to a user; the integration mints a stable user id and round-trips the Dify conversation_id automatically, so multi-turn evaluations preserve state.

Prerequisites

• A published Dify agent-chat app, and the base URL of your Dify deployment.
• A Dify API key for that app.

Configuration

Define the model with connection_type: dify and reference the API key as a secret:

key: "dify-agent"
display_name: "Dify Agent"
description: "Dify agent-chat app, served via the typed `dify` connection config."
task: "chat_completion"
rate_limit: 5
config:
  connection_type: "dify"
  url: $DIFY_URL
  api_key: "<< secrets.DIFY_API_KEY >>"
secrets:
  DIFY_API_KEY: $DIFY_API_KEY

Config fields

Field	Required	Secret	Description
connection_type	yes	no	Must be dify.
url	yes	no	Base URL of the Dify deployment.
api_key	yes	yes	Dify API key (sent as a bearer token).

Credentials

The secrets block uploads values (from your .env or environment) to server-side secret storage, and << secrets.NAME >> references them from the config so no key is stored in plaintext. Provide the values in your .env:

DIFY_URL=https://api.dify.ai
DIFY_API_KEY=app-...

Add and test

lf add model -f dify.yaml
lf test model dify-agent

Use in an evaluation

Reference the model key from a task specification in your run config:

evaluation:
  task_specifications:
    - task_key: my-task
      model_key: "dify-agent"

Conversation state

On the first turn the integration mints a stable user id and starts a Dify conversation; the user id and conversation_id are attached to the assistant message and reused on each subsequent turn, so the agent remembers earlier turns.