Continuous codebase judgment

Understand what matters in a repository before you change it.

Solvix is a CLI that scans a codebase, explains the important targets, records traceable evidence, and compares findings across future runs.

Start using Solvix Read the report guide

Terminal brief

Solvix Maintainer Brief
Identity       : http_client_library (system)
Default lens   : api_stability
Judgment       : needs_benchmark / build good
Primary target : lib/helpers/formDataToJSON.js -> formDataToJSON()
Confidence     : low
Trace ID       : svx-52422549735e
History        : compared with previous run; new=0, repeated=2
Next check     : Benchmark nested FormData input shapes.

Use cases

When Solvix helps

Before changing unfamiliar code

Learn the project identity, central files, risky paths, and safest first target.

Before a pull request

Check whether the change touches a high-relevance path or needs a benchmark.

During maintenance reviews

Run history to see what is new, repeated, changed, or resolved since the last run.

Before writing an issue

Use the trace ID, confidence level, verification step, and no-send reasoning.

Quick start

Run it on a project

Start with parser health, then run project mode. Add history when you want Solvix to compare the current run with the previous trace ledger.

solvix doctor
solvix analyze . --project
solvix analyze . --project --history
solvix analyze . --project --history --json --output solvix-report.json

Report reading

What to look at first

Maintainer Brief Read the identity, default lens, judgment, primary target, trace ID, and next check.
Judgment Use sendable, needs benchmark, or no-send to decide whether to inspect, measure, or hold back.
Trace Ledger Follow the evidence chain from inventory and identity to graph, lens, judgment, and verification.
Run History Compare this run with the previous baseline and review changed or new findings first.

Judgment status

Solvix does not treat every signal as a bug

sendable

A bounded target exists. Review it seriously, then verify before making claims.

needs_benchmark

The target is plausible, but measurement or source review should come before action.

no_send

The signal is too weak, low-impact, already calibrated, or not worth maintainer attention.

Documentation