Home

Serverless autocompletion: an embedded engine whose database is a bucket.

completr is a library, not a service. Your application embeds it (in Rust, or in Python through first-class bindings), and an object-store bucket or a local directory is the database. There is no cluster to deploy, scale or keep alive: every process that serves completions reads the index straight from storage, and writers coordinate through the storage itself.

• Every match kind

  ---

  Exact, prefix, infix, abbreviation, spelling-tolerant and word-decomposing completion, ranked together with popularity.

  Completion

• Serverless by design

  ---

  Storage is the only shared component. Commits are create-only writes, and ingestion is a role any process can take.

  Serverless deployment

• Semantic and hybrid

  ---

  Embeddings from any model, quantised to 2 to 4 bits, fused with lexical results by reciprocal rank, weighted blending or lexical-first ordering.

  Semantic and hybrid

• Override layers

  ---

  Search a tenant's, a user's or an experiment's index on top of shared data, per document id, without copying it.

  Layers

• Fast and deterministic

  ---

  Memory-mapped, checksummed segments. Typed queries take around 0.1 ms at p50, and identical inputs give bit-identical results.

  Architecture

• Rust and Python

  ---

  Documents as dicts, Document objects, or pandas, polars and Arrow tables. Typed stubs, an asyncio API and a command-line tool.

  Python API

Install

PythonRustCommand line

pip install completr

cargo add completr                     # engine
cargo add completr --features store    # plus databases on local disk and in memory

cargo install completr-cli

Quick start

PythonRust

from completr import Index

index = Index.from_documents([
    {"id": "ml", "text": "Machine Learning", "popularity": 0.9, "abbreviations": ["ML"]},
    {"id": "mv", "text": "Machine Vision", "popularity": 0.4},
    {"id": "ds", "text": "Data Science", "popularity": 0.7},
])

for query in ["mach", "ML", "vison", "science"]:
    print(query, [(s.text, s.kind) for s in index.complete(query, limit=3)])

use completr::{Document, Index};

let index = Index::from_documents([
    Document::keyed("ml", "Machine Learning", 0.9).with_abbreviation("ML"),
    Document::keyed("mv", "Machine Vision", 0.4),
    Document::keyed("ds", "Data Science", 0.7),
])?;

for s in index.complete("mach", 10) {
    println!("{} {} {:.3}", s.text, s.kind.as_str(), s.score);
}

mach [('Machine Learning', 'prefix'), ('Machine Vision', 'prefix')]
ML [('Machine Learning', 'abbreviation')]
vison [('Machine Vision', 'fuzzy')]
science [('Data Science', 'infix')]

Next, Getting started walks from a first index to a live database, and Concepts explains the model behind it.