Dify Integration¶

Pisama integrates with Dify to monitor AI application workflows and detect failure modes in Dify-built agents.

Setup¶

Register a Dify Instance¶

Connect your Dify instance to Pisama:

curl -X POST http://localhost:8000/api/v1/dify/instances \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "production-dify",
    "url": "https://my-dify-instance.com",
    "api_key": "your_dify_api_key"
  }'

Register Apps for Monitoring¶

curl -X POST http://localhost:8000/api/v1/dify/apps \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "instance_id": "<instance_id>",
    "app_id": "my-chatbot-app",
    "name": "Customer Support Bot"
  }'

Configure Webhook¶

In your Dify instance, set up a webhook to send workflow execution data to Pisama:

Webhook URL: https://your-pisama.com/api/v1/dify/webhook
Method: POST
Headers:
  X-MAO-API-Key: <your_api_key>

Detection Capabilities¶

Pisama detects the following failure modes in Dify workflows:

• Hallucination: LLM nodes generating unsupported claims
• Loop detection: Workflow loops and retry storms
• Context overflow: Token accumulation across workflow nodes
• Task derailment: Agents going off-topic
• State corruption: Data transformation errors between nodes
• Cost tracking: Token usage and cost per workflow execution

API Endpoints¶

Real-Time Monitoring¶

Connect to the SSE stream for live Dify execution updates:

curl -N http://localhost:8000/api/v1/dify/stream \
  -H "Authorization: Bearer $TOKEN"