104 lines
2.6 KiB
Markdown
104 lines
2.6 KiB
Markdown
# Animatrix
|
|
|
|
Animatrix is a constrained full-stack Wan 2.2 generation surface built on top of the live Desineuron ComfyUI ingress.
|
|
|
|
It is designed to be a production-facing prompt console, not a node editor.
|
|
|
|
## Current Product Scope
|
|
|
|
Current live scope:
|
|
|
|
- local operator auth
|
|
- asset upload and library
|
|
- grounded start-frame generation
|
|
- animate mode
|
|
- audio performance surface
|
|
- aspect ratio selection
|
|
- duration selection
|
|
- batch submission from `x1` to `x4`
|
|
- result viewing and download
|
|
|
|
Current non-live scope:
|
|
|
|
- true end-frame conditioning
|
|
- exact loop closure workflows
|
|
- additional model families beyond Wan 2.2 A14B
|
|
|
|
If a control is not wired to the live Comfy workflow, it should not be treated as production-capable.
|
|
|
|
## Repository Structure
|
|
|
|
- `frontend/`: Next.js operator surface
|
|
- `backend/`: FastAPI API, auth, storage, orchestration
|
|
- `workflows/`: versioned ComfyUI workflow templates
|
|
- `infra/`: deployment and service scaffolding
|
|
- `scripts/`: utility scripts
|
|
- `Animatrix Project Truth.md`: project truth artifact
|
|
|
|
## Runtime Topology
|
|
|
|
Production topology today:
|
|
|
|
- frontend: Linux box
|
|
- backend: Linux box
|
|
- public host: `animatrix.desineuron.in`
|
|
- Comfy target: `https://comfy.desineuron.in`
|
|
- GPU execution: AWS GPU box using local Wan models on NVMe
|
|
|
|
The app should never assume direct browser access to the GPU box.
|
|
|
|
## Local Development
|
|
|
|
Backend:
|
|
|
|
```bash
|
|
cd backend
|
|
python3.12 -m venv .venv
|
|
source .venv/bin/activate
|
|
pip install -r requirements.txt
|
|
cp .env.example .env
|
|
python run.py
|
|
```
|
|
|
|
Frontend:
|
|
|
|
```bash
|
|
cd frontend
|
|
npm install
|
|
cp .env.local.example .env.local
|
|
npm run dev
|
|
```
|
|
|
|
Local production-style frontend build:
|
|
|
|
```bash
|
|
./scripts/build_frontend_standalone.sh
|
|
```
|
|
|
|
## Production Configuration Notes
|
|
|
|
Expected production assumptions:
|
|
|
|
- frontend browser calls use same-origin `/api`
|
|
- backend `COMFYUI_BASE_URL` points to `https://comfy.desineuron.in`
|
|
- backend `BACKEND_BASE_URL` points to `https://animatrix.desineuron.in`
|
|
- Linux box stores app DB, uploaded assets, and persisted outputs
|
|
- GPU box stores ComfyUI runtime and Wan model files on NVMe
|
|
|
|
## Operational Notes
|
|
|
|
Two truths matter in production:
|
|
|
|
1. A valid workflow file does not guarantee a valid runtime.
|
|
2. Comfy loader availability must match the model names referenced in the workflow.
|
|
|
|
Animatrix now depends on live runtime model preflight to avoid silent drift.
|
|
|
|
## Documentation
|
|
|
|
Start with:
|
|
|
|
- [Animatrix Project Truth.md](F:/Workin%20In%20Progress/DESINEURON/GITLAB/Animatrix/Animatrix%20Project%20Truth.md)
|
|
|
|
This README is the first stronger pass and should continue to evolve with deployment truth, workflow inventory, and operator runbooks.
|