devdoctor

CLI for scanning, explaining, and fixing developer environment issues.

The npm package is @bishalsingh/devdoctor. After it is installed, the command on your PATH is still devdoctor (from the package bin field). npx must use the scoped package name; a global install does not.

devdoctor scan runs checks against the current project and writes the last result to .devdoctor/last-scan.json. Use explain and fix on issue IDs from that scan. benchmark [options] measures token/time/estimated cost of proposing those fixes (read-only).

Install

Try without installing

npx @bishalsingh/devdoctor scan

Install globally from npm (for repeated use)

npm install -g @bishalsingh/devdoctor
devdoctor scan

Permission errors on Linux/macOS

If npm install -g @bishalsingh/devdoctor fails with EACCES: permission denied, npm's default global install location isn't writable by your user. Two options:

• Quick fix: run the install with sudo:

  sudo npm install -g @bishalsingh/devdoctor

• Recommended one-time fix: point npm's global installs at a directory your user owns, so this doesn't happen for any future global package:

  mkdir ~/.npm-global
  npm config set prefix '~/.npm-global'
  echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc
  source ~/.bashrc

  (use ~/.zshrc instead of ~/.bashrc if you're on zsh)

  Then retry npm install -g @bishalsingh/devdoctor without sudo.

Local development from source (clone this repo)

npm install
npm run build
npm install -g .
devdoctor scan

That last npm install -g . links the built bin so the bare devdoctor command works locally too.

Setting up explain and fix (Groq API key)

scan needs no API key. explain and fix call Groq’s API and require GROQ_API_KEY.

1. Sign up at https://console.groq.com and create an API key from the dashboard (free tier is enough).

2. Set it in your shell:

   export GROQ_API_KEY=your-key-here

   or put this in a .env file in the project root you are scanning:

   GROQ_API_KEY=your-key-here

This is a Groq account and key, not npm, Anthropic, or OpenAI.

Usage

Default command (no subcommand) runs scan. Examples below use npx @bishalsingh/devdoctor …. If you installed globally, drop npx @bishalsingh/ and run devdoctor … the same way.

npx @bishalsingh/devdoctor
npx @bishalsingh/devdoctor scan
npx @bishalsingh/devdoctor --version
npx @bishalsingh/devdoctor -V
npx @bishalsingh/devdoctor --help
npx @bishalsingh/devdoctor --verbose
npx @bishalsingh/devdoctor scan --verbose

Explain an issue from the last scan (needs GROQ_API_KEY):

npx @bishalsingh/devdoctor explain ISSUE_ID
npx @bishalsingh/devdoctor explain --help

Fix an issue (needs GROQ_API_KEY). fix and fix --interactive require an interactive terminal (a real TTY). They will not work when piped, run in CI, or run through non-interactive automation, because they prompt for confirmation before writing any files.

npx @bishalsingh/devdoctor fix ISSUE_ID
npx @bishalsingh/devdoctor fix --interactive
npx @bishalsingh/devdoctor fix -i
npx @bishalsingh/devdoctor fix --help

Benchmark (needs GROQ_API_KEY). Measures how many tokens, how much time, and estimated cost it takes an AI to propose fixes for the project's current issues. Read-only: it never writes files. The idea is to track whether a codebase gets more or less "AI-friendly" (cheaper/faster to fix) over time as code, specs, and conventions improve.

npx @bishalsingh/devdoctor benchmark
npx @bishalsingh/devdoctor benchmark --limit 5
npx @bishalsingh/devdoctor benchmark --history

Estimated cost uses Groq's published list pricing for the current model and is labeled as an estimate, not a live quote. Each run is appended to .devdoctor/benchmark-history.json; --history shows those trends over time.

--verbose on the root command or on scan prints full error details (including per-check failures).

Config

Optional project-root file: .devdoctorrc.json or .devdoctorrc.

{
  "disabledChecks": ["consoleLogLeftIn", "outdatedDependencies"]
}

Missing config: all checks run. Invalid JSON: a warning is printed to stderr (Ignoring invalid .devdoctorrc.json: …) and the scan continues with every check enabled.

Checks

| Name | Detects | | --- | --- | | noTestScript | Missing or npm-placeholder test script in package.json | | unusedDependencies | Production dependencies listed but not imported | | todoComments | TODO comments in source files | | missingErrorMiddleware | Express apps with no (err, req, res, next) handler | | largeComponents | Source files over 500 lines | | jwtHardcodedFallback | Hardcoded JWT secret fallback (JWT_SECRET \|\| '…') | | envFileCommitted | .env files tracked by git (not .env.example / sample / template) | | outdatedDependencies | Packages more than 2 major versions behind npm latest | | gitignoreGaps | Missing .gitignore, or missing relevant Node/TS ignore entries | | consoleLogLeftIn | console.log in src/ outside tests and CLI entry files |

Development

From a clone, run the CLI with tsx (no global install):

npm run dev
npm run dev -- scan
npm run dev -- --version
npm run dev -- --help
npm run dev -- explain ISSUE_ID
npm run dev -- fix ISSUE_ID
npm run dev -- fix --interactive

Build a single bundled file at dist/cli.js:

npm run build
node dist/cli.js scan