all-mpnet-base-v2 · 768d · pgvector
5 pruning strategies

Adaptive Vector Retrieval

Ask anything.

The document answers.

Upload a research paper, report, or manual. Choose how your vectors get pruned. Get back precise answers with page-level attribution.

01 · PDF Ingestion
02 · Sentence Chunking
03 · Embedding (768d)
04 · Vector Pruning
05 · pgvector Store
06 · ANN Retrieval
07 · LLM Generation

Adaptive Pruning

Five strategies from cosine to MMR - each with a different retention philosophy.

Dual Retrieval Mode

Full context for short docs. pgvector ANN + optional MaxSim for long ones.

Pipeline Inspector

Dev mode exposes every step - embeddings, scores, thresholds, and pruning decisions.

Interface

Document intelligence

Upload

Drop your PDF here

or click to browse

Query