# Hello Ollama
## Prerequisites
- **Ollama** - a local AI model server, which is used to handle embeddings and generate responses.
### Quick Start
The simplest way to run this sample is using the included `run.sh` script, which handles installation, server startup, and model pulling automatically:
```bash
./run.sh
```
### Monitoring and Running
For an enhanced development experience, use the provided `run.sh` script to start the sample with automatic reloading:
```bash
./run.sh
```
This script uses `watchmedo` to monitor changes in:
- `src/` (Python logic)
- `../../packages` (Genkit core)
- `../../plugins` (Genkit plugins)
- File patterns: `*.py`, `*.prompt`, `*.json`
Changes will automatically trigger a restart of the sample. You can also pass command-line arguments directly to the script, e.g., `./run.sh --some-flag`.
### Manual Setup
If you prefer to set up manually:
1. **Install Ollama**: Download from [ollama.com](https://ollama.com/download).
2. **Start the server**:
```bash
ollama serve
```
3. **Pull models**:
```bash
ollama pull mistral-nemo:latest
ollama pull gemma3:latest
```
4. **Run the sample**:
```bash
genkit start -- uv run src/main.py
```
## Testing This Demo
1. **Open DevUI** at http://localhost:4000
2. **Test basic flows**:
- [ ] `say_hi` - Simple generation with gemma3
- [ ] `say_hi_stream` - Streaming response
- [ ] `say_hi_constrained` - Constrained output
3. **Test tools** (requires mistral-nemo):
- [ ] `calculate_gablorken` - Tool calling demo
4. **Notes**:
- gemma2:latest does NOT support tool calling
- Use mistral-nemo for tool-based flows
- First run may be slow (model loading)
5. **Expected behavior**:
- Responses generated locally (no API calls)
- Streaming shows incremental output
- Tools work with compatible models only
