Web Server¶

basemode-loom includes a FastAPI server for headless use or building custom frontends.

Starting the server¶

basemode-loom serve --host 127.0.0.1 --port 8000

Or from Python:

import uvicorn
from basemode_loom import GenerationStore
from basemode_loom.api import create_app

store = GenerationStore()
app = create_app(store)
uvicorn.run(app, host="127.0.0.1", port=8000)

REST API¶

List trees¶

GET /api/roots

Returns a list of all root nodes as JSON objects.

Create a root¶

POST /api/roots
Content-Type: application/json

{"text": "Your prompt here"}

Delete a tree¶

DELETE /api/roots/{root_id}

Get full tree¶

GET /api/roots/{root_id}/tree

Returns {"nodes": [...]} with all nodes in the tree.

Get tree stats¶

GET /api/roots/{root_id}/stats

Returns the same stats as basemode-loom stats.

WebSocket API¶

For live interactive sessions, connect to the WebSocket endpoint:

ws://localhost:8000/ws/session

The WebSocket handler manages a full LoomSession. On connection, send an init message to specify which tree to open:

{"type": "init", "root_id": "<root_id>"}

The server then streams state updates and generation events back to the client as JSON. Navigation and generation commands are sent as messages to the server.

This is the same session interface used internally by the TUI, so any command that works in the TUI is available over WebSocket.