Instrument your app
Where traces go
A sink is the function a finished trace is handed to. Send it to ChunkTrace, write it to a file, or both.
To ChunkTrace#
import { createProfiler, httpSink } from "@chunktrace/sdk";
const profiler = createProfiler({
sink: httpSink({
endpoint: "https://api.chunktrace.com",
apiKey: process.env.CHUNKTRACE_KEY, // a full-access key
}),
});httpSink posts each trace to the HTTP API. It never throws. If the request fails, it logs a warning and your app carries on.
To handle failures yourself:
httpSink({ endpoint, apiKey, onError: (error) => logger.warn({ error }, "chunktrace ingest failed") });Sending the same trace twice is safe. The server keeps one copy per trace_id.
To a file#
import { createProfiler, fileSink } from "@chunktrace/sdk";
const profiler = createProfiler({ sink: fileSink("./traces") });Writes each trace to ./traces/<trace_id>.json. No account, no network. Read them back with the CLI:
npx chunktrace diagnose traces/tr_mwk0ppb3_unefr7.jsonTo both#
import { fileSink, httpSink, multiSink } from "@chunktrace/sdk";
const sink = multiSink(fileSink("./traces"), httpSink({ endpoint, apiKey }));Nowhere#
With no sink, traces are kept in memory on profiler.traces. That suits tests:
const profiler = createProfiler();
// … run a query …
expect(profiler.traces).toHaveLength(1);Your own#
A sink is a function that takes a trace. It can be async.
const profiler = createProfiler({
sink: async (trace) => {
await queue.publish("traces", JSON.stringify(trace));
},
});The trace is plain JSON in the shape described in Trace schema.