Agent Chat

Agent Chat provides an interactive assistant that helps you design extraction configurations and answer document-related questions. It runs on a dedicated queue so chat requests stay responsive while document processing workloads are running.

How It Works

1. Send a chat message with optional sessionId and documentGuid.
2. The API queues the request and returns a job GUID.
3. Poll the status endpoint until the response is complete.

API Usage

Start a chat

curl -X POST https://api.docudevs.ai/agent/chat \
  -H "Authorization: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role": "user", "content": "Help me build an invoice extraction schema"}
    ],
    "sessionId": "session-123"
  }'

Poll for status

curl -X GET https://api.docudevs.ai/agent/status/<JOB_GUID> \
  -H "Authorization: $API_KEY"

Python SDK

import asyncio
from docudevs.docudevs_client import DocuDevsClient

async def chat_with_agent():
    client = DocuDevsClient(token="your-api-key")

    response = await client.agent_chat_and_wait(
        messages=[{"role": "user", "content": "Extract line items from invoices"}],
        session_id="session-123",
    )

    print(response["response"]["message"])

asyncio.run(chat_with_agent())

Case-bound chat

Pass caseId to work on one case. The agent then follows the case's active memo instructions, works through case-bound tools and proposes changes that a person confirms. playbook and context select an investigation. See Case agent and Agent actions.

UI

The web app exposes Agent Chat at /app/agent, including optional document upload for context. With ?caseId= in the URL the page opens the case agent.