Skip to content

v0.2.0 — Initial Public Release

Choose a tag to compare

View all tags
@gregpriday gregpriday released this 31 Mar 03:54
· 254 commits to develop since this release
v0.2.0
b05ada0

SearchSocket is now publicly available on NPM.

What's Included

Indexing Pipeline

Incremental updates with content hashing mean only changed pages are re-indexed. Cost tracking gives visibility into embedding API usage across runs.

Source Modes

Four ways to feed content into the indexer: `static-output` (pre-built HTML), `build` (run your build then index), `crawl` (follow a sitemap), and `content-files` (raw Markdown/HTML files).

Semantic Search

Powered by Jina AI embeddings (`jina-embeddings-v3`) with task-specific LoRA adapters — `retrieval.passage` at index time, `retrieval.query` at search time. Results are reranked with Jina AI for improved relevance.

Vector Storage

Uses Turso/libSQL for vector storage, supporting both local SQLite and remote Turso databases.

MCP Server

A Model Context Protocol server with both stdio and HTTP transports, exposing `search` and `get_page` tools for AI assistant integration.

SvelteKit Integration

First-class SvelteKit support via a Vite plugin (triggers indexing on build) and a server hook that mounts the search API endpoint.

Browser Client

`searchsocket/client` entry point for frontend search integration directly in the browser.

CLI

Full-featured CLI with `init`, `index`, `status`, `dev`, `clean`, `prune`, `doctor`, `mcp`, and `search` commands.

Other Highlights

  • Multi-scope support using git branches for environment isolation
  • Page-level score aggregation with score-weighted decay
  • Synthetic page summary chunks with meta extraction
  • Noindex support, gzip sitemap support, Zod request validation, rate limiting, and CORS configuration

Full Changelog

https://github.com/gregpriday/searchsocket/releases/tag/v0.2.0

Assets 2
Loading