---
title: AI QA Test Case Generator
emoji: 🧪
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
---

# AI QA — RAG Test Case Generator (fully offline)

FastAPI server that generates QA test cases from Jira context using
**local GGUF models** (llama-cpp-python, CPU-only) + Pinecone vector search.
No HuggingFace Inference API credits required.

## Models baked into the image

| Role | Model | Size |
|------|-------|------|
| Embedding | Qwen3-Embedding-8B-Q4_K_M | ~4.7 GB |
| Generation | StableHorde API (Qwen3-14B) | cloud |

## Endpoints

| Method | Path | Description |
|--------|------|-------------|
| GET | `/health` | Liveness + model info |
| POST | `/generate` | JSON test cases |
| POST | `/stream` | SSE streaming output |
| POST | `/gentest` | Slack slash command |

## `/generate` — request body

```json
{
  "feature": "user login with email and password",
  "team": "casting"
}
```

`team` is optional. Valid values: `casting`, `talent`, `agency`, `be`, `mobile`

## Space secrets (Settings → Repository secrets)

| Secret | Purpose |
|--------|---------|
| `STABLEHORDE_API_KEY` | StableHorde API key (required) |
| `SLACK_BOT_TOKEN` | Slack bot token (optional) |
| `SLACK_SIGNING_SECRET` | Slack signing secret (optional) |
| `STABLEHORDE_MODEL` | Override model name (default: `Qwen3-14B`) |