TuskDB
The Postgres client that speaks pgvector.
Inspect embeddings, run similarity search, and tune your index without hand-writing distance SQL every time. No other Mac Postgres client does this.
14-day free trial · no account · no credit card
Nearest neighbors, from the grid.
Right-click any row with a vector column and TuskDB finds its nearest neighbors, ranked by distance. No query to write, no operator to remember.
- Works on any
vector,halfvec, or sparse column - Cosine, inner product, or L2, matched to your index
- Results open in the same grid you already know
Search by meaning, on your terms.
Point TuskDB at your own embedding endpoint: a local Ollama model or any OpenAI-compatible API. Type a phrase, TuskDB embeds it against your endpoint and searches your table. Your text goes to the endpoint you chose and nowhere else.
- Local Ollama or any OpenAI-compatible API
- Your API key lives in the macOS Keychain, like your passwords
- Nothing routes through our servers
Stop guessing at recall.
Did it use the index?
Every vector query in the editor gets a badge showing whether it hit the HNSW or IVFFlat index, or fell back to a scan.
Recall vs latency
Chart recall against latency across ef_search values, so you tune to a number instead of a hunch.
Validated inline
TuskDB reads the column's declared dimensions and flags a mismatch before it becomes a runtime error.
Put your embeddings in a real UI.
Every feature, 14 days, no account.