JupyterLab and Notebooks

JupyterLab and Notebooks

ADEPT integrates JupyterLab as a first-class scientific computing environment, providing authenticated access to the full platform API directly from notebooks.

Overview

JupyterLab runs as part of the ADEPT stack with:

• Single sign-on via Keycloak (OAuth2/OIDC)
• Direct access to the ADEPT API for agent interactions
• Pre-configured scientific Python environment
• Session-scoped file access for uploaded data

Accessing JupyterLab

JupyterLab is available through the ADEPT web interface after authentication. Access is controlled by Keycloak group membership -- users in the notebook-users group are granted access.

Authentication

JupyterLab authenticates through the same OAuth2 flow as all other ADEPT services:

1. User navigates to the JupyterLab URL
2. OAuth2 proxy redirects to Keycloak login
3. After authentication, a session cookie grants access
4. API calls from notebooks use the same user identity

Using ADEPT from Notebooks

Chat with the Agent

import httpx

BASE_URL = "https://your-adept-server.example.com/v1"
TOKEN = "your-jwt-token"
headers = {"Authorization": f"Bearer {TOKEN}"}

# Create a thread
resp = httpx.post(f"{BASE_URL}/threads", headers=headers)
thread_id = resp.json()["id"]

# Send a message via the Responses API
resp = httpx.post(f"{BASE_URL}/responses/chat/completions", headers=headers, json={
    "model": "default",
    "messages": [{"role": "user", "content": "Analyze my dataset"}],
    "thread_id": thread_id,
    "stream": False,
})
print(resp.json()["choices"][0]["message"]["content"])

Upload Files

with open("experiment_data.csv", "rb") as f:
    resp = httpx.post(f"{BASE_URL}/files", headers=headers,
        files={"file": ("experiment_data.csv", f, "text/csv")},
        data={"purpose": "assistants"})
file_id = resp.json()["id"]

Tips

• Session Continuity: Use the same thread_id across multiple cells to maintain conversation context.
• Large files: For files over 100 MB, consider splitting them or using batch processing tools.
• Streaming: Set "stream": True and iterate over the SSE response for real-time output.
• Multi-agent teams: Create specialized teams via the CreateMultiAgentSession tool.
• Kernel restarts: Thread state persists server-side, so restarting your kernel does not lose conversation history.