Size and performance

Brevity aims to keep framework overhead small. Here are the measurements we have, the code behind them, and what they can tell you.

Dependencies

Core and web have no external runtime dependencies. Views adds Eta 3.5.0; native islands adds no external runtime dependency. Svelte is a peer of the separate Svelte adapter, not part of native islands.

Count your app's database client, auth SDK, and browser runtime too. A small framework doesn't remove those costs.

Second application: support inbox

The support inbox example implements the same app in Brevity and Express 5.1.0: validated forms, persistent requests, details, search, pagination, and status updates. Both work without JavaScript and share the business logic.

Authored source	Brevity	Express
Routes and templates, nonblank lines	117	150
Whole runnable app, nonblank lines	309	342
Whole runnable app, bytes	17,271	19,593

That's 33 fewer nonblank lines in this example, mostly HTML/form glue. It's a modest reduction, not a claim that all Brevity apps take less code or time. Both serve the same 322-byte browser script. The local store is single-process and the example has no authentication.

Reproduce the source comparison

From examples/support-inbox, after installing and building packages:

npm run typecheck
npm run build
npm test
npm run smoke
node scripts/measure.mjs > /tmp/brevity-inbox-current.json
node scripts/measure.mjs | cmp - /tmp/brevity-inbox-current.json

The historical results include the file inventory, versions, dependency graph, and counting rules from before the npm scope migration. Current imports are longer, so byte counts change; rerun the script to measure your checkout. Each app total includes shared code once. Generated output, dependencies, lockfiles, and runtime data aren't counted as authored source. With common build configuration included, totals are 359 nonblank lines for Brevity and 392 for Express.

Report bytes alongside lines: compressed templates and CSS can hide a lot of code on one line. Blank lines improve readability and aren't a cost to eliminate.

Synthetic HTTP benchmark

Small JSON routes were measured against Express 5.3.0, Fastify 5.12.5, and Nest 12.1.2 using its Express adapter:

Framework	GET requests/s	POST requests/s	GET p95
Brevity	41,800	38,857	0.637 ms
Express	16,542	14,311	1.745 ms
Fastify	40,937	38,793	0.672 ms
Nest	14,955	13,005	1.927 ms

Brevity and Fastify had similar small-route throughput in this local run. Fastify used less sampled memory. These routes don't include a database, auth, templates, or overload, so don't use these numbers to size a production app.

Machine, method, and full report

The snapshot used Node 22.21.1 on an Apple M1 Max running macOS ARM64. Each scenario started a new process, warmed up for 500 ms, and measured for two seconds with 16 concurrent keep-alive clients. Values above are medians of three rotated-order runs. The client ran on the same machine as the servers.

Framework	Median process-to-ready	Maximum sampled RSS
Brevity	50.2 ms	91.5 MiB
Express	71.1 ms	126.6 MiB
Fastify	96.8 ms	79.8 MiB
Nest	237.0 ms	176.0 MiB

Startup includes process creation, imports, and listening, not cold filesystem caches. RSS is the child process's sampled peak. Startup/RSS combine both methods and all repeats. Brevity POST runs ranged from 34,091 to 40,508 requests/s, including a run below Fastify. Nest returns 201 for POST; the others return 200.

See the benchmark README for commands and the JSON report with individual runs, p50/p95/p99, event-loop delay, machine details, versions, and compiled-source hashes. Responses are checked; failures abort the run. A shared machine and short closed-loop run limit how much we can infer.

Test your workload

Compare equivalent behavior: the same validation, errors, escaping, storage, and browser experience. Include templates, configuration, and client code when comparing source. Measure development time with real developers, not line counts.

For hosted application results and outstanding load/recovery checks, see production readiness.