You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Comprehensive Vector Data Tooling. The universal interface for all vector database, datasets and RAG platforms. Easily export, import, backup, re-embed (using any model) or access your vector data from any vector databases or repository.
A comprehensive RAG FastAPI service that handles document uploads and retrievals, built with Python. Uses PyMuPDF for document processing, turbopuffer for vector storage, OpenAI for models, and cohere for reranking.
A retrieval gateway for your vector store. Speaks the Turbopuffer wire protocol and adds caching, embedding, transforms and scans without changing client code.
Query turbopuffer namespaces from DuckDB. Results come back as Arrow tables, and SQL functions run BM25 and vector search from inside a query. Unofficial.
Real-time AI agent for querying live courtroom video with sub-500ms latency. Multimodal search combining video intelligence, speech-to-text, and hybrid search. Built with Stream, Twelve Labs, Deepgram, and Gemini Live API.
Search evaluation with Jev as the judge. turbopuffer does the retrieval and hosted reranking, and DuckDB keeps every run and computes the metrics in SQL.
Powerpipe mod: 15 security controls (tenant isolation, data residency, CMEK, schema hygiene, ops) and branded dashboards for turbopuffer namespaces. Runs over the somoore/turbopuffer Steampipe plugin.
Mirror of raw api specs for turbopuffer designed for distilled SDK use (@distilled.cloud/turbopuffer). ⚠️ these are unpatched and are not an accurate representation of actual api behavior; https://github.com/alchemy-run/distilled contains patched smithy specs for this api
Real-time AI agent for querying live courtroom video with sub-500ms latency. Multimodal search combining video intelligence, speech-to-text, and hybrid search. Built with Stream, Twelve Labs, Deepgram, and Gemini Live API.
Candidate search over a Turbopuffer database for ten hiring roles. Attribute-filtered scans, keyword prescreens and a GPT-4o-mini judge gave the pipeline as committed an 87.7 average eval score, and grader-guided resubmission took the final slates to 90.3. Vector search in Runs 2 and 3 reached 66.7.
People search over 100K+ LinkedIn profiles: GPT-4o query rewriting, Voyage-3 embeddings, Turbopuffer filtered ANN retrieval, and structured re-ranking to return the top 10 candidates.