# AirMicroDrip Deployment Guide

## Prerequisites

- Hugging Face account
- HF_TOKEN environment variable set

## Quick Deploy

```bash
cd airmicrodrip
export HF_TOKEN=your_hf_token_here
bash deploy.sh
```

## Manual Deploy (Git Push)

```bash
cd airmicrodrip
export HF_TOKEN=your_hf_token_here
export HF_SPACE_ID=josephrw/membra-airmicrodrip

# Initialize git
git init
git config user.email "deploy@membra.ai"
git config user.name "MEMBRA Deploy"
git add -A
git commit -m "Deploy AirMicroDrip"

# Push to Hugging Face
git push "https://$HF_TOKEN@huggingface.co/spaces/$HF_SPACE_ID" main --force
```

## Runtime Configuration

The app does not require API keys or Space secrets at runtime. Optional overrides can be added in HF Space Settings:

| Variable | Description | Example |
|----------|-------------|---------|
| `TOKEN_MINT` | Override the default public token mint | `EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v` |
| `INFERENCE_API_URL` | Optional no-key local/Ollama-compatible LLM endpoint for live benchmarking | `http://localhost:11434` |
| `SOLANA_RPC_URL` | Override the public Solana RPC endpoint | `https://api.mainnet-beta.solana.com` |

## Data Sources

All data is fetched from real APIs — no mocks, no simulations:

- **Trading**: Gate.io Futures API (real-time prices, funding rates)
- **Slippage**: DexScreener API (real DEX volume/liquidity data)
- **Holders**: Solana RPC `getTokenLargestAccounts` (real on-chain holder data)
- **LLM Liquidity**: Real HTTP inference benchmark (Ollama or OpenAI-compatible)
- **Liquidation**: Gate.io prices + local position DB

## Architecture

```
HF Space (Docker)
  └── Flask API
      ├── /api/slippage/stats    → DexScreener API
      ├── /api/holders/stats     → Solana RPC
      ├── /api/liquidity/stats   → Inference benchmark
      ├── /api/trading/stats     → Gate.io API
      ├── /api/funding/stats     → Gate.io API
      ├── /api/liquidation/stats → Gate.io + local DB
      ├── /api/token-launch/status  → Merkle launch manifest
      ├── /api/token-launch/prepare → Rebuild unsigned launch tree
      ├── /api/token-launch/pool-setup → Unsigned mint/pool setup plan
      ├── /api/collateral/scan   → Background HF account file/LOC collateral scan
      ├── /api/collateral/status → Latest collateral root and evidence
      └── /                      → Dashboard UI
```

## Verification

After deployment:

1. Visit `https://huggingface.co/spaces/YOUR_SPACE`
2. Check System Status shows real backend states for trading, funding, liquidation, holders, and slippage
3. Confirm the header says "No-key backend"
4. Confirm "One Merkle Tree Token Launch" shows `unsigned_ready`
5. LLM liquidity should show `local_only` until a real local endpoint is connected

## Troubleshooting

- **Holder/slippage waiting**: Public Solana or DexScreener data has not returned yet; no key is required
- **LLM liquidity local-only**: This is expected without a real local inference endpoint
- **Merkle launch unsigned_ready**: The root and proofs are prepared; SPL mint/pool creation still requires a real wallet signature
- **"Gate.io API unreachable"**: Check network connectivity; app uses fallback only if API fails