Skip to content
MicroResolve

Rust — Connect to Server

The embedded MicroResolve instance runs entirely in-process. For teams that want a shared training pipeline — auto-learn, Studio UI, review queue — you can point the engine at a self-hosted server using connected mode.

What connected mode does

In connected mode the local engine:

  1. Still classifies locally (zero added latency on the hot path)
  2. Pulls approved training updates from the server on a background thread at a configurable interval
  3. Buffers low-confidence queries and flushes them to the server’s ingest endpoint on each sync tick
  4. Pushes corrections to the server immediately when correct() is called

Training happens in the server (Studio UI, auto-learn pipeline, review queue). All connected clients receive the update on the next sync tick. No manual sync call is required — the background thread handles it.

Configure

use microresolve::{MicroResolve, MicroResolveConfig, ServerConfig};


let engine = MicroResolve::new(MicroResolveConfig {
    data_dir: Some("~/.local/share/microresolve".into()),
    default_threshold: 0.3,
    server: Some(ServerConfig {
        url: "http://localhost:3001".into(),
        api_key: std::env::var("MICRORESOLVE_API_KEY").ok(),
        subscribe: vec!["support".into(), "security".into()],
        ..Default::default()
    }),
    ..Default::default()
})?;


// Namespaces listed in `subscribe` are pulled from the server on startup
// and kept in sync by the background thread. Use them like any other namespace:
let ns = engine.namespace("support");
let matches = ns.resolve("cancel my order");

The subscribe list tells the engine which namespaces to pull. On startup, each is fetched from GET /api/sync. The background poll runs every sync_interval_secs (default: 60).

Corrections are pushed immediately

// This call is forwarded to POST /api/correct on the server in addition to
// updating the local engine. No extra step required.
ns.correct("stop shipment", Some("cancel_order"), "track_order")?;

Calling the server directly over HTTP

For operations not covered by the embedded API (reviewing the queue, triggering auto-learn), call the server’s REST API directly.

Terminal window
curl -X POST http://localhost:3001/api/learn/now \
  -H "X-Namespace-ID: support" \
  -H "X-Api-Key: $MICRORESOLVE_API_KEY"

Next