sgrep v2.0.0

sgrep v2.0.0

🚀 What's New

Modal GPU Offload

Offload embeddings to Modal.dev A10G GPUs for faster indexing and search on large codebases:

• sgrep index --offload - Index using cloud GPUs
• sgrep search --offload - Search using cloud embeddings
• Concurrent embedding requests with keep-alive connections
• Configurable batch size (default: 128) and concurrency
• Auto-deploy Modal service on first use

⚡ Performance Improvements

• Parallel embedding batches with progress-safe ordering
• Keep-alive tuned HTTP agent for Modal requests
• Optional gzip compression for payloads
• Improved embedding progress tracking

🔧 Configuration

• New [modal] config section for token and endpoint settings
• Batch size and concurrency configurable via config or env vars
• Simplified GPU config: A10G used for all Modal workloads
• Enhanced logging with consistent formatting

🐛 Bug Fixes & Improvements

• Handle missing namespaces on delete operations
• Truncate Modal embeddings to model limits
• Increase Modal startup timeout to prevent initialization failures
• Use Jina embeddings on Modal for index/search compatibility

📦 Build & Release

• Switch from bun publish to npm publish
• Fix npm publish workflow authentication
• Add manual workflow trigger for publishing

---

What's Changed

• Fix bun publish npm authentication by @dallenpyrah in #58
• Switch from bun publish to npm publish by @dallenpyrah in #59
• Fix npm publish workflow and add manual trigger by @dallenpyrah in #60

Full Changelog: v1.5.5...v2.0.0