Sentient-Field/sgp-tribe3

1---2title: SGP-Tribe33emoji: 🧠4colorFrom: green5colorTo: yellow6sdk: docker7pinned: false8license: cc-by-nc-4.09---10 11# SGP-Tribe312 13**Sentient Generative Principal — Brain Encoding Calibration System**14 15A fork of [TRIBE v2](https://huggingface.co/facebook/tribev2) (Meta AI) extended with:16 17- Full multimodal inference: video + audio + text simultaneously18- SGP 9-node parcellation using the Schaefer-200 cortical atlas19- Dual-stream (dorsal/ventral) activation separation20- White matter tract co-activation edge weight output21- Stimulus calibration matrix accumulation across sessions22 23## API Endpoints24 25### `POST /predict`26Upload a video file (mp4, max 120s). Returns SGP node activation profile.27 28**Request:** `multipart/form-data` with `video` field29 30**Response:**31```json32{33  "stimulus_id": "uuid",34  "sgp_nodes": {35    "G1_broca": 0.73,36    "G2_wernicke": 0.61,37    "G3_tpj": 0.55,38    "G4_pfc": 0.48,39    "G5_dmn": 0.32,40    "G6_limbic": 0.67,41    "G7_sensory": 0.81,42    "G8_atl": 0.59,43    "G9_premotor": 0.4444  },45  "streams": {46    "ventral_mean": 0.64,47    "dorsal_mean": 0.5848  },49  "dominant_hemisphere": "left",50  "n_timesteps": 42,51  "n_vertices": 2048452}53```54 55### `GET /health`56Returns model load status.57 58### `POST /warmup`59Triggers model loading (call once after cold start).60 61## Architecture62 63Built on the SGP-LLM framework. Nodes map to:64- **G1 Broca** — phonological production, syntactic processing65- **G2 Wernicke** — auditory comprehension, lexical-semantic decoding  66- **G3 TPJ** — stream convergence, sensorimotor interface67- **G4 PFC** — executive control, coherence/veto68- **G5 DMN** — generativity, self-referential processing69- **G6 Limbic** — emotional weighting, memory consolidation70- **G7 Sensory** — primary perceptual input encoding71- **G8 ATL** — cross-modal semantic integration hub72- **G9 Premotor** — action planning, motor speech preparation73 74## Citation75 76Built on TRIBE v2:77```78@article{dAscoli2026TribeV2,79  title={A foundation model of vision, audition, and language for in-silico neuroscience},80  author={d'Ascoli, Stéphane and Rapin, Jérémy and Benchetrit, Yohann and others},81  year={2026}82}83```84