@leumas/flow-recorder

Record real runtime hook flows into replayable timelines, and expose them over the API. Part of the Leviathan runtime — captured flows feed the perception→rules→action loop (P5), so apps stop being black boxes and become observable, replayable, trainable organisms.

Works with any hook emitter that supports onAny/offAny or on('')/off(''), plus emit().

Record & replay

import { createFlowRecorder } from '@leumas/flow-recorder';

const recorder = createFlowRecorder({ hooks }); // hooks = your live-hooks emitter
recorder.start();
hooks.emit('server:start');
hooks.emit('music:loaded', { track: 'dreamscape.mp3' });
recorder.stop();

const flow = recorder.export();           // { meta:{source,started,ended}, events:[{name,ts,payload}] }
await recorder.replay({ hooks, speed: 1 }); // re-emit in order; delay = original delta * speed

Pass filter as an array of event names or ({ name, payload }) => boolean to limit capture. ts is relative to started in ms.

API (mounted in leumas-api at /api/flows)

import { createFlowRouter } from '@leumas/flow-recorder';
app.use('/api/flows', createFlowRouter({ store, gate: requirePassNode(), getHooks }));

• GET /api/flows → list saved flows (summaries)
• GET /api/flows/:id → one full flow (meta + events)
• POST /api/flows → save a flow { meta, events } (gate meters writes)
• POST /api/flows/:id/replay → replay onto a host-provided emitter (getHooks(id, body))
• DELETE /api/flows/:id → delete

store defaults to an in-memory store (createMemoryFlowStore). Back it with the leumas.db flows collection by passing a { list, get, save, remove } adapter.