Back to browse
GitHub Repository

Keep Postgres entities synced with turbopuffer using logical replication (beta)

13 starsRust

Puffgres logically replicates Postgres entities in turbopuffer

by lucasgelfond·Jul 16, 2026·1 point·0 comments

AI Analysis

●●●BangerBig BrainSolve My Problem

Logical replication beats pgvector drift by deriving vectors from Postgres WAL.

Strengths
  • At-least-once delivery with built-in retries eliminates manual batching and backfill code.
  • Immutable configs with tombstoning prevent schema drift when updating transforms.
  • Co-located state in Postgres means PITR restores keep vectors and source in sync.
Weaknesses
  • Beta status with only 13 GitHub stars means limited community support and examples.
  • Turbopuffer-only; no adapters for Pinecone, Weaviate, or other vector databases.
Target Audience

Backend engineers building RAG pipelines with Postgres + turbopuffer

Similar To

pgvector · Supabase Vector · Neon Vector

Post Description

title instead of two DB calls everything, you write ‘dumb’ JS transforms (i.e. tokenization, embedding) and the core handles logical replication, retries, batching, etc. we actively use on ~200k rows!

Similar Projects