RAG infrastructure for developers

From documents to answers. One API.

Create an index, push your documents, and get grounded answers with citations — over one clean API. Retrieval, chunking, hybrid search, and reranking included. Built on Postgres. Fully managed, or deployed in your own cloud on Enterprise.

No credit card · cancel anytime

How it works

Three commands. From signup to a grounded answer with citations in under five minutes.

Step 1

Create an index

Smart defaults: cosine, hybrid search on, 800/100 token chunking.

bash
curl https://api.dextome.com/v1/indexes \
  -H "Authorization: Bearer $DEXTOME_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name":"Docs"}'
Step 2

Add a document

Push text, a URL, or upload a PDF. Ingestion runs async with live status.

bash
curl https://api.dextome.com/v1/indexes/idx_abc/documents \
  -H "Authorization: Bearer $DEXTOME_KEY" \
  -H "Content-Type: application/json" \
  -d '{"source_type":"text","text":"Refunds within 30 days.","title":"Refunds"}'
Step 3

Get a grounded answer

Cited answer with the retrieved chunks. SSE streaming available.

bash
curl https://api.dextome.com/v1/indexes/idx_abc/answer \
  -H "Authorization: Bearer $DEXTOME_KEY" \
  -H "Content-Type: application/json" \
  -d '{"question":"What is the refund window?"}'

Everything you'd build yourself

Chunkers, embeddings, hybrid search, reranking, citations, and BYO model keys.

Managed embeddings

OpenAI, Cohere, Voyage — or BYO key. Pluggable per index.

Smart chunking

Recursive, sentence, markdown-aware, fixed. Tuned defaults out of the box.

Hybrid search

pgvector HNSW + tsvector keyword, fused with Reciprocal Rank Fusion.

Reranking

Cohere rerank-v3.5 on Pro+. The single biggest quality lever.

Citations & sources

Every grounded answer maps back to chunk anchors — page, heading, row.

Metadata filtering

Parameterized JSON-DSL → SQL. No string concat, no injection.

Async ingestion at scale

pg-boss queue with retries, content-hash idempotency, live progress.

BYO models

Your OpenAI / Anthropic / Cohere keys. We meter our infra, not your tokens.

Just Postgres

Vectors, keyword search, and metadata in one Postgres. No separate vector database to run or sync.

One pipeline. One Postgres

Documents flow into a single Index. Retrieval and answer happen behind one API.

Documents

text · url · pdf · docx

ingest

Index

pgvector HNSWtsvector FTSmetadata
rerank

Grounded answer

citations included

retrieve

One API

/v1/query/v1/answer

Your data stays yours

Dextome never trains on your documents. Every index is tenant-isolated, stored on encrypted infrastructure, and deletable on demand.

Predictable pricing

Generous free tier with hybrid search included. Upgrade when you ship.

Free

$0

 

  • 5K queries / mo
  • 200 answers / mo
  • 50K vectors
  • 3 indexes
  • Hybrid search
Most popular

Pro

$25

per month

  • 500K queries / mo
  • 50K answers / mo (10K on our model key)
  • 500K vectors
  • Unlimited indexes
  • Hybrid search
  • Reranking
  • Bring your own model keys

Team

$99

per month

  • 2M queries / mo
  • 250K answers / mo (50K on our model key)
  • 2.5M vectors
  • Unlimited indexes
  • Hybrid search
  • Reranking
  • Bring your own model keys
  • Roles and permissions

Pro and Team can go over their allowances, billed by usage. See full pricing →

Start building with Dextome

Five minutes from signup to a cited answer.