rinha4.c

DOCS COMMAND CENTER

Read the C scorer by evidence path, not by scroll depth.

This hub maps the implementation to the source files and benchmark artifacts behind each claim: fd-passing transport, raw HTTP parsing, 14-dimensional vector scoring, and archived official/CI results.

entrypoint:9999standalone yolo FD-pass LB
services2 API + LBroot docker-compose.yml
budget1 CPU / 350 MBofficial resource envelope
evidencereports/candidate, experiment, calibrated

Fast paths

Contract review

Start with endpoints, score failure modes, and the 1 CPU / 350 MB envelope.

 Hot-path review

Trace the handoff from port 9999 through Unix control sockets into the C API.

 Local loop

Build, run tests, launch Compose, hit /ready, and smoke /fraud-score.

 Evidence check

Open archived benchmark JSON/HTML before trusting any metric claim.

Source-backed map

Challenge

Endpoint contract, resource envelope, and scoring pressure.

 Architecture

FD-passing topology, request path, data pipeline, and classifier boundaries.

 Rules

Repository operating rules and promotion constraints.

 Getting Started

Build, test, Docker smoke, endpoint sample, and docs loop.

 Performance

Hot-path choices, current bottleneck, competitor signals, and search instrumentation.

 CI/CD Pipeline

Image build, benchmark archive, report publishing, and Pages deployment.
Where each docs claim should point back to source truth.
Docs areaSource of truthReview note
Challenge docker-compose.yml, official result JSON Resource budget and required endpoints.
Architecture src/api, src/common, src/preprocess Transport, vectorization, index, and response path.
Performance docs/public/reports, search.c, comparison branch notes CI evidence, p99 caveats, and active optimization lane.
CI/CD .github/workflows, docs/public/reports Immutable images, report kinds, archive semantics.