v0.4.18 - Ollama Local Embeddings
What's New
Ollama as Default Embedding Provider - CV-Git now uses local embeddings by default, eliminating the need for cloud API keys for semantic search.
Features
- 🦙 Ollama Integration: Local embedding generation using
nomic-embed-textmodel (768 dimensions) - 🐳 Docker Support: Auto-starts Ollama container on
cv initandcv sync - 🔄 Automatic Fallback: Falls back to OpenRouter/OpenAI if Ollama unavailable
- 🩺 Health Checks:
cv doctornow shows Ollama status and model availability
Changes
- Default embedding provider changed from
openroutertoollama - Added
ensureOllama()infrastructure for Docker container management - Updated
cv init,cv sync, andcv findto use Ollama-first approach - Fixed VectorManager to properly respect Ollama configuration
- Added text truncation (500 chars) to prevent Ollama batch size issues
Benefits
- Zero API costs for embeddings
- Offline capable semantic search
- Faster local embedding generation
- Privacy - code never leaves your machine
Installation
Quick Install (recommended):
curl -fsSL https://raw.githubusercontent.com/controlVector/cv-git/main/install.sh | bashnpm:
npm install -g @controlvector/cv-gitDebian/Ubuntu:
wget https://github.com/controlVector/cv-git/releases/download/v0.4.18/cv-git_0.4.18_amd64.deb
sudo dpkg -i cv-git_*.debGetting Started
cv doctor # Check system status (shows Ollama health)
cv init # Initialize repo (auto-starts Ollama)
cv sync # Build knowledge graph with local embeddings
cv find "authentication logic" # Search locallyRequires Docker for Ollama container, or configure cloud providers as fallback.