grepai-embeddings-lmstudio — скилл для Claude Code: локальные

---
name: grepai-embeddings-lmstudio
description: Configure LM Studio as embedding provider for GrepAI. Use this skill for local embeddings with a GUI interface.
---

# GrepAI Embeddings with LM Studio

This skill covers using LM Studio as the embedding provider for GrepAI, offering a user-friendly GUI for managing local models.

## When to Use This Skill

- Want local embeddings with a graphical interface
- Already using LM Studio for other AI tasks
- Prefer visual model management over CLI
- Need to easily switch between models

## What is LM Studio?

LM Studio is a desktop application for running local LLMs with:
- 🖥️ Graphical user interface
- 📦 Easy model downloading
- 🔌 OpenAI-compatible API
- 🔒 100% private, local processing

## Prerequisites

1. Download LM Studio from [lmstudio.ai](https://lmstudio.ai)
2. Install and launch the application
3. Download an embedding model

## Installation

### Step 1: Download LM Studio

Visit [lmstudio.ai](https://lmstudio.ai) and download for your platform:
- macOS (Intel or Apple Silicon)
- Windows
- Linux

### Step 2: Launch and Download a Model

1. Open LM Studio
2. Go to the **Search** tab
3. Search for an embedding model:
   - `nomic-embed-text-v1.5`
   - `bge-small-en-v1.5`
   - `bge-large-en-v1.5`
4. Click **Download**

### Step 3: Start the Local Server

1. Go to the **Local Server** tab
2. Select your embedding model
3. Click **Start Server**
4. Note the endpoint (default: `http://localhost:1234`)

## Configuration

### Basic Configuration

```yaml
# .grepai/config.yaml
embedder:
  provider: lmstudio
  model: nomic-embed-text-v1.5
  endpoint: http://localhost:1234
```

### With Custom Port

```yaml
embedder:
  provider: lmstudio
  model: nomic-embed-text-v1.5
  endpoint: http://localhost:8080
```

### With Explicit Dimensions

```yaml
embedder:
  provider: lmstudio
  model: nomic-embed-text-v1.5
  endpoint: http://localhost:1234
  dimensions: 768
```

## Available Models

### nomic-embed-text-v1.5 (Recommended)

| Property | Value |
|----------|-------|
| Dimensions | 768 |
| Size | ~260 MB |
| Quality | Excellent |
| Speed | Fast |

```yaml
embedder:
  provider: lmstudio
  model: nomic-embed-text-v1.5
```

### bge-small-en-v1.5

| Property | Value |
|----------|-------|
| Dimensions | 384 |
| Size | ~130 MB |
| Quality | Good |
| Speed | Very fast |

Best for: Smaller codebases, faster indexing.

```yaml
embedder:
  provider: lmstudio
  model: bge-small-en-v1.5
  dimensions: 384
```

### bge-large-en-v1.5

| Property | Value |
|----------|-------|
| Dimensions | 1024 |
| Size | ~1.3 GB |
| Quality | Very high |
| Speed | Slower |

Best for: Maximum accuracy.

```yaml
embedder:
  provider: lmstudio
  model: bge-large-en-v1.5
  dimensions: 1024
```

## Model Comparison

| Model | Dims | Size | Speed | Quality |
|-------|------|------|-------|---------|
| `bge-small-en-v1.5` | 384 | 130MB | ⚡⚡⚡ | ⭐⭐⭐ |
| `nomic-embed-text-v1.5` | 768 | 260MB | ⚡⚡ | ⭐⭐⭐⭐ |
| `bge-large-en-v1.5` | 1024 | 1.3GB | ⚡ | ⭐⭐⭐⭐⭐ |

## LM Studio Server Setup

### Starting the Server

1. Open LM Studio
2. Navigate to **Local Server** tab (left sidebar)
3. Select an embedding model from the dropdown
4. Configure settings:
   - Port: `1234` (default)
   - Enable **Embedding Endpoint**
5. Click **Start Server**

### Server Status

Look for the green indicator showing the server is running.

### Verifying the Server

```bash
# Check server is responding
curl http://localhost:1234/v1/models

# Test embedding
curl http://localhost:1234/v1/embeddings \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nomic-embed-text-v1.5",
    "input": "function authenticate(user)"
  }'
```

## LM Studio Settings

### Recommended Settings

In LM Studio's Local Server tab:

| Setting | Recommended Value |
|---------|-------------------|
| Port | 1234 |
| Enable CORS | Yes |
| Context Length | Auto |
| GPU Layers | Max (for speed) |

### GPU Acceleration

LM Studio automatically uses:
- **macOS:** Metal (Apple Silicon)
- **Windows/Linux:** CUDA (NVIDIA)

Adjust GPU layers in settings for memory/speed balance.

## Running LM Studio Headless

For server environments, LM Studio supports CLI mode:

```bash
# Start server without GUI (check LM Studio docs for exact syntax)
lmstudio server start --model nomic-embed-text-v1.5 --port 1234
```

## Common Issues

❌ **Problem:** Connection refused
✅ **Solution:** Ensure LM Studio server is running:
1. Open LM Studio
2. Go to Local Server tab
3. Click Start Server

❌ **Problem:** Model not found
✅ **Solution:**
1. Download the model in LM Studio's Search tab
2. Select it in the Local Server dropdown

❌ **Problem:** Slow embedding generation
✅ **Solutions:**
- Enable GPU acceleration in LM Studio settings
- Use a smaller model (bge-small-en-v1.5)
- Close other GPU-intensive applications

❌ **Problem:** Port already in use
✅ **Solution:** Change port in LM Studio settings:
```yaml
embedder:
  endpoint: http://localhost:8080  # Different port
```

❌ **Problem:** LM Studio closes and server stops
✅ **Solution:** Keep LM Studio running in the background, or consider using Ollama which runs as a system service

## LM Studio vs Ollama

| Feature | LM Studio | Ollama |
|---------|-----------|--------|
| GUI | ✅ Yes | ❌ CLI only |
| System service | ❌ App must run | ✅ Background service |
| Model management | ✅ Visual | ✅ CLI |
| Ease of use | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| Server reliability | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ |

**Recommendation:** Use LM Studio if you prefer a GUI, Ollama for always-on background service.

## Migrating from LM Studio to Ollama

If you need a more reliable background service:

1. Install Ollama:
```bash
brew install ollama
ollama serve &
ollama pull nomic-embed-text
```

2. Update config:
```yaml
embedder:
  provider: ollama
  model: nomic-embed-text
  endpoint: http://localhost:11434
```

3. Re-index:
```bash
rm .grepai/index.gob
grepai watch
```

## Best Practices

1. **Keep LM Studio running:** Server stops when app closes
2. **Use recommended model:** `nomic-embed-text-v1.5` for best balance
3. **Enable GPU:** Faster embeddings with hardware acceleration
4. **Check server before indexing:** Ensure green status indicator
5. **Consider Ollama for production:** More reliable as background service

## Output Format

Successful LM Studio configuration:

```
✅ LM Studio Embedding Provider Configured

   Provider: LM Studio
   Model: nomic-embed-text-v1.5
   Endpoint: http://localhost:1234
   Dimensions: 768 (auto-detected)
   Status: Connected

   Note: Keep LM Studio running for embeddings to work.
```