LiftCanary

SDKs

Node.js

Server-side decisions for many users, plus AI config helpers.

@liftcanary/node decides on your server: one client serves every request, keeps the datafile fresh in the background and sends events in batches. Decisions are in-process function calls, so they add no latency to your requests.

Install#

Terminal
npm install @liftcanary/node

Requires Node.js 18 or newer (it uses the built-in fetch).

Usage#

liftcanary.ts
import { createClient } from "@liftcanary/node";

export const liftcanary = createClient({ sdkKey: process.env.LIFTCANARY_SDK_KEY! });
await liftcanary.ready(); // at startup
An Express route
app.post("/checkout", async (req, res) => {
  const user = liftcanary.createUser({
    id: req.user.id,
    attributes: { plan: req.user.plan, country: req.headers["cf-ipcountry"] as string },
  });

  const flow = user.getVariable("new_checkout", "flow", "classic");
  const order = await placeOrder(req.body, flow);
  user.track("purchase", { value: order.total });
  res.json(order);
});

Server defaults#

The Node.js client starts with defaults suited to one process deciding for many users:

streaming
true
Live updates: a published change arrives in about a second. The connection never keeps your process from exiting; false turns it off.
pollingIntervalMs
30000
Also checks for a new datafile every 30 s, in case the live connection drops (ETag, so unchanged datafiles cost almost nothing).
eventFlushIntervalMs
2000
Sends queued events every 2 s.
eventBatchSize
200
Events per request.
eventMaxQueueSize
20000
Events held in memory before the oldest are dropped (counted in client.events.dropped).
modelPrices
Record<model, { input, output }>
Price per 1M tokens, for AI configs.

Any of them can be overridden, along with the shared options.

Shutting down#

TypeScript
process.on("SIGTERM", async () => {
  await liftcanary.close(); // stops polling and flushes remaining events
  process.exit(0);
});

Pages, locations and request context#

Location conditions need the page: pass the full request URL as page, and the keys of activated triggered locations, if you track them, as activeLocations. Neither is sent with events.

At sign-in, pass the browser's anonymous id (the liftcanary_anon cookie, also exported as ANONYMOUS_ID_COOKIE) as anonymousId, so results count the landing-page visit and the signed-in user as one person. See From landing page to app.

TypeScript
liftcanary.createUser({ id, page: `https://${req.headers.host}${req.originalUrl}` });

AI configs#

The package also has aiConfig, measureLLM, measureLLMStream (streamed answers, with the time to first token), trackLLM, trackFeedback and renderPrompt for testing models and prompts. See AI configs and LLM experiments.