Video Intelligence Pipeline
Every minute of every camera → stored chunk + metadata + VLM summary + embeddings in Qdrant.
Demo mode — the live AI pipeline is paused; showing archived data.
Chunks (24h)
0
0 embedded
Backlog
0
chunks awaiting metadata
Errors (24h)
0
flagged chunk_metadata rows
Embedding worker
…
0 video_pipeline workers
Chunk throughput (chunks / min, last 60m)
No chunks recorded yet.
Embeddings in Qdrant
—
frame_emb
SigLIP image
—
scene_emb
VLM summary
—
object_emb
object text
Per-camera pipeline
Loading…