Rust API

The Rust API reference is on docs.rs/typemind. Enable the store feature for databases, and aws (or aws-credentials), gcp or azure for cloud storage. The database API is async and runs on Tokio. From synchronous code, typemind::block_on runs one of its futures on typemind's internal runtime; do not call it from async code.

[dependencies]
typemind = { version = "0.1", features = ["store"] }
tokio = { version = "1", features = ["macros", "rt-multi-thread"] }

Indexes and engines

Item	Feature	Role
Document		A document: Document::new(id, text, popularity) or Document::keyed(key, text, popularity), with with_synonym, with_abbreviation, with_context and with_vector.
Index		Segments searched as one: from_documents, complete, complete_aliases, vector_search, hybrid_search, and the _with variants taking options.
Segment, SegmentWriter		An immutable, memory-mapped set of documents and deletes; and a writer of segment files in bounded memory.
Engine		Indexes by name, published atomically, searched as layers. Returns LayeredSuggestions.
Suggestion, AliasSuggestion, HybridSuggestion		Results, with id, key, text, score and, for completions, kind and byte-range highlights.
MatchKind, Fusion		Match kinds, and hybrid fusion methods.
BuildOptions, IndexOptions, SearchOptions, HybridOptions		Settings, with chainable setters.
Error		Io, Corrupt, InvalidInput, NotFound, Conflict and Store, among others.

use std::sync::Arc;
use typemind::{AliasKind, Document, Index, IndexOptions, MatchKind, Segment};

let docs = [
    Document::new(1, "Under the Bridge – Red Hot Chili Peppers", 0.75)
        .with_alias("RHCP", AliasKind::Abbreviation),
    Document::new(2, "Bohemian Rhapsody – Queen", 0.95),
];
let index = Index::new(vec![Arc::new(Segment::build(docs, [])?)], IndexOptions::default())?;
let hits = index.complete("rhcp", 10);
assert_eq!((hits[0].id, hits[0].kind), (1, MatchKind::Abbreviation));

Databases

Item	Feature	Role
Database	store	Versioned manifests of named indexes in a store: Database::open(url, options), begin, open_index, compact, cleanup, acquire_lease, submit.
Transaction	store	An optimistic, atomic commit.
Replica	store	Publishes a database's latest version into an Engine: sync, and follow, which syncs from a background thread.
Follower, FollowStatus	store	The handle Replica::follow returns; dropping it stops the thread. status() returns synced_at_ms and error.
Lease	store	An expiring, fenced lock in the database.
ChangeSet	store	Upserts and deletes submitted to the inbox.
Ingestor, IngestStep	store	The lease-elected committer of change sets, and the result of a round.
CompactionPolicy, CleanupPolicy	store	Maintenance settings.
Store, BlockingStore	store	Object-store access, async and blocking.

use std::{sync::Arc, time::Duration};
use typemind::{Database, Document, Engine, IndexOptions, Replica};

#[tokio::main]
async fn main() -> Result<(), typemind::Error> {
    let database = Database::open("./data", Vec::<(String, String)>::new()).await?;
    let mut txn = database.begin().await?;
    txn.append_documents("songs", [Document::keyed("bohemian", "Bohemian Rhapsody – Queen", 0.95)], [])?;
    txn.commit().await?;

    let engine = Arc::new(Engine::new());
    let replica = Arc::new(Replica::new(database, IndexOptions::default()));
    replica.sync(&engine).await?;
    let _follower = replica.follow(&engine, Duration::from_secs(5));   // syncs until dropped
    let hits = engine.complete(&["songs"], "boh", 10)?;
    Ok(())
}

Runnable examples: quickstart.rs for an index and a database, and live_updates.rs for a followed database with an ingestor.

Collections

The collections layer. In Rust, typemind::connect returns this client; a database is opened with Database::open.

Item	Feature	Role
connect, ConnectOptions	store	Opens a client on the database at a URL: sync_every, cache_dir, storage options, build and index options.
Client	store	A database and its collections: collections, collection, create_collection, get_or_create_collection, drop_collection, sync, and database().
Collection	store	Named documents: add, delete, complete, optimize and stats.
Query	store	limit, aliases, layers, contexts and vector, with chainable setters.
Completion	store	A result: id, key, text, score, kind, byte-range highlights and collection.
CollectionStats	store	Size and freshness, from Collection::stats.
Optimize	store	Auto { policy, min_interval } or Off.

use typemind::{connect, ConnectOptions, Document, MatchKind, Optimize, Query};

#[tokio::main]
async fn main() -> Result<(), typemind::Error> {
    let client = connect("memory://", ConnectOptions::default()).await?;
    let songs = client.get_or_create_collection("songs", Optimize::default()).await?;
    songs
        .add([
            Document::keyed("bridge", "Under the Bridge – Red Hot Chili Peppers", 0.75)
                .with_abbreviation("RHCP"),
        ])
        .await?;
    let hits = songs.complete("rhcp", &Query::default())?;
    assert_eq!((hits[0].key.as_deref(), hits[0].kind), (Some("bridge"), MatchKind::Abbreviation));
    Ok(())
}