| name | kg-webui-extraction |
| description | The agent-webui document / entity extraction surface — submit text or a URL for GPU-slot fact extraction, watch jobs stream facts live, pause/resume, and export results. Use when you want to extract entities/facts from a document in the UI, monitor an extraction job, or download extracted facts as JSONL. |
| license | MIT |
| tags | ["graph-os","webui","extraction","facts","documents","entities"] |
| tier | surface |
| metadata | {"author":"Genius","version":"0.1.0"} |
kg-webui-extraction (surface)
Documents the document→knowledge-graph fact-extraction UI (AU-ECO.connector.git-task-resolver;
KG-2.64/2.65/2.66): submit a document, watch a GPU-slot job emit facts round by
round over SSE, pause/resume it, and export the deduplicated facts.
UI route (agent-webui/src/App.tsx)
/extraction — ExtractionView (submit form, live job list, streaming fact
log, JSONL download).
Backing endpoints (from src/lib/api.ts, under /api/enhanced/extract/)
POST /submit {text|url, rounds?, dedup?, dedup_field?, dedup_threshold?} —
enqueue a job (submitExtraction), returns {status, job_id}.
GET /jobs — job rows (listExtractionJobs), each an ExtractionJob
(state, total_facts, unique_facts, duplicate_facts, preempted…).
GET /status/{job_id} — single job status (getExtractionStatus).
POST /pause/{job_id} / POST /resume/{job_id} (pauseExtraction /
resumeExtraction).
GET /stream/{job_id} — SSE of round_start|fact|…|job_done
(extractionStreamUrl, consumed via EventSource).
GET /jsonl/{job_id} — download facts as JSONL (extractionJsonlUrl).
How an agent drives it
Navigate to /extraction, or call directly: POST /api/enhanced/extract/submit
with text or a URL, poll GET /api/enhanced/extract/status/{job_id} (or attach
an EventSource to /stream/{job_id}), and pull results from /jsonl/{job_id}.
For headless document ingestion prefer the kg-ingest verb skill
(document_process / source_sync); use this surface for interactive,
observable extraction runs.