What ViewBrain is, who uses it, key features, how to run an analysis, and current limitations.

fMRI, BOLD, HRF, fsaverage5, brain networks, and how each of the 5 scores is computed from the ROI timeseries — including the exact brain regions and formulas.

Getting started guide, prerequisites, local setup, connecting to servers, and deploying changes.

Server inventory, SSH tunnel, public request path, A100 lifecycle management, and security model.

GPU specs, the golden snapshot, auto-start and auto-stop behaviour, SSH key management, cost control, and recovery.

Step-by-step journey from video upload to rendered results, including timing and error handling.

Architecture, inputs (Whisper, LLaMA 3.2-3B, ViJePA2), predictions.npy, tribev2_meta.json field-by-field reference, modality attribution, and attention matrix.

Objectives, prompt construction, Claude API config, timestamp accuracy, and how to extend objectives.

All pages — Analyser, Details, Analysis, Servers, Users, Settings — plus auth flow, roles, URL routing, and deployment.

All endpoints: upload, jobs, fetch_url, channel/profile scan, tribe-score, dashboard library/benchmark, video_meta, A100 control, recommendations.

Internal endpoints via CX33 SSH tunnel: POST /process (blocking), GET /health (model_loaded, GPU stats), worker setup.

Login, JWT, media tokens, user management, settings endpoints. Rate limiting, roles, and how auth.js works.

Cost breakdown, A100 auto-stop mechanism, monthly estimates, ThunderCompute credits management.

Normal operations, disaster recovery scenarios, credentials reference, and deployment procedures.