Newsence Core Worker

HTTP endpoints (frontend):
GET  /health
POST /ingest                              - Ingest URL (JSON), image URL (JSON), or user-uploaded blob (multipart)
POST /scrape                              - Sync extraction: {url} JSON or raw bytes -> NormalizedContent {markdown,text,metadata,status}
POST /scrape/jobs                         - Async parse job (non-persisting): {url} or raw bytes -> {jobId}
GET  /scrape/jobs/:id                     - Poll parse job -> {status, result?, error?}
POST /generate-image                      - AI image gen (OpenRouter → R2 → user_files)
POST /embed                               - Generate embeddings
POST /media/delete                        - Batch-delete user-file R2 objects by storage key (#162)
POST /media/gc                            - On-demand reference-nowhere R2 orphan sweep (#162)
GET  /stream/:instanceId                  - Workflow status (SSE)

Signed media:
GET  /media/external/{options}/{mediaUrl} - Upstream image/video passthrough with edge cache
GET  /media/asset/{key}?sig=&exp=         - Authenticated R2 asset