docs/DEPLOY.md

# Deployment & Submission Guide

**Deadline: June 15, 2026** (tomorrow!)

## What You Need to Submit

1. Gradio Space hosted under the `build-small-hackathon` HuggingFace organization
2. Short demo video
3. Social media post

---

## How to Deploy

### 1. Create the Space

Go to https://huggingface.co/organizations/build-small-hackathon and create a new Space:
- Framework: **Gradio**
- Hardware: **T4 GPU**
- Visibility: **Public**

### 2. Push Your Code

Option A — clone the empty Space and copy files in:
```bash
git clone https://huggingface.co/spaces/build-small-hackathon/SPACE_NAME
cd SPACE_NAME
# copy project files in
git add .
git commit -m "Initial submission"
git push
```

Option B — add as a remote to this repo:
```bash
git remote add space https://huggingface.co/spaces/build-small-hackathon/SPACE_NAME
git push space master:main
```

### 3. Check `requirements.txt`

The Space build system runs `pip install -r requirements.txt`. Verify it includes:
- `gradio`
- `transformers`
- `accelerate`
- `torch`

### 4. Model Weights

Make sure `src/llm.py` loads Qwen2.5-3B-Instruct from the Hub (not a local path):
```python
AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-3B-Instruct")
```

### 5. character.glb (IMPORTANT)

The GLB model is not in the repo. Add it via git-lfs before pushing or the 3D avatar will be broken:
```bash
git lfs install
git lfs track "*.glb"
git add static/models/character.glb
git commit -m "add character model"
```

---

## Badges You're Targeting

- **Off the Grid** — local model, no cloud APIs
- **Off-Brand** — custom HTML/JS frontend injected into Gradio
- **Tiny Titan** — Qwen2.5-3B (≤4B params) *(verify this is an official badge)*

---

## Hackathon Page

https://huggingface.co/build-small-hackathon