QuackIR: Jimmy Lin's EMNLP Paper Shows RDBMSes Match Vector DBs for RAG Retrieval
lintool · x · 2026-09-26
QuackIR (EMNLP 2025 Industry Track, Jimmy Lin et al.) argues enterprises can skip dedicated vector databases for RAG by using relational databases they already run. The toolkit implements retrieval in DuckDB, SQLite, and PostgreSQL; benchmarks show sparse and dense retrieval comparable to established IR toolkits. Open-sourced at quackir.io.
More from coding & agent
- Vision MCP Server routes image analysis through OpenRouter to add vision to any model — modelcontextprotocol · 2026-09-26
- Jevless: Jev-style typed decisions from any model with logprobs, plus a local server — seraphius · 2026-09-26
- The real local agent harness stack: Aider, Cline, OpenHands and the multi-agent gap — Prestigious_Style267 · 2026-09-26
- Slack AI agents get centralized logs of every tool call for team governance — adamse · 2026-09-26
- PP Embeds Source Maps Into Component Pixels So Agents Can Locate Code From Any Screenshot — jasonkneen · 2026-09-26
- Claude Code worktrees boosted productivity, but testing changes is now the bottleneck — CuriousityIgnited · 2026-09-26