The agent is a container you run on the machine with the GPU. It authenticates with the token below, leases jobs from this server, computes on the GPU, and posts results back — all over HTTP. Start it when you want a burst; stop it to reclaim the card.
{{ masked ? maskedToken : tokenValue }}
Point the agent at {{ baseUrl }} with this token. Rotating
invalidates the old token — update the agent after you rotate.
Errored jobs park after 3 failed attempts. This requeues just those (their errors cleared, attempts reset) — use after updating the agent, without re-running the whole done library.
Enqueues every image that doesn't have a CCIP embedding yet. Nothing processes until the agent is running.
Enqueues every image that doesn't have concept-crop embeddings yet — the localized vectors that help small/local tags (glasses, etc.) surface. New images get these automatically; this catches the back-catalogue.
Re-runs the FULL pipeline (figure detection + CCIP + concept/panel crops) on every image — use after changing crop detectors so the back-catalogue gets re-cropped, not just new images. Heavy: re-processes the whole library.
How close a figure must be (CCIP cosine) to suggest a character. Higher = stricter — fewer but more confident matches. 0.85 recommended; below ~0.80 a heavily-tagged character starts matching everything.
When on, a very-confident character match tags the image on its own (daily, reversible) — so identity tags keep flowing without review. Stricter than the suggest cut; 0.92 recommended.
Switching the model is a DIFFERENT embedding space. After Save model, run Re-embed library (the GPU agent re-embeds whole images + concept crops), then Retrain heads — suggestions degrade until both finish. SigLIP 2 (512px) is a 1152-d drop-in over SigLIP 1; new installs default to it. Your existing library stays on its current model until you re-embed.