Website: https://hotspots.dev | Docs: https://docs.hotspots.dev | Crates.io:
Install: brew install Stephen-Collins-tech/tap/hotspots | npm install -g @stephencollinstech/hotspots | pip install hotspots-cli | cargo install hotspots-cli
Find the code that's actually causing problems.
Your codebase has thousands of functions. Some are messy but never break. Others are complex AND change constantly — those are your hotspots, the 20% of code causing 80% of your bugs, incidents, and slowdowns.
brew install Stephen-Collins-tech/tap/hotspots # macOS
npm install -g @stephencollinstech/hotspots # any platform
pip install hotspots-cli # any platform
cargo install hotspots-cli # Rust toolchain
curl -fsSL https://raw.githubusercontent.com/Stephen-Collins-tech/hotspots/main/install.sh | sh # LinuxWindows: download the binary from GitHub Releases.
New here? Start with the Quick Start guide.
Verify: hotspots --version
GitHub Action:
- uses: Stephen-Collins-tech/hotspots/action@v1
with:
github-token: ${{ secrets.GITHUB_TOKEN }}# Find your hotspots
hotspots analyze src/
# Critical (LRS ≥ 9.0):
# processPlanUpgrade src/api/billing.ts:142 LRS 12.4 CC 15 ND 4 FO 8 NS 3
#
# High (6.0 ≤ LRS < 9.0):
# validateSession src/auth/session.ts:67 LRS 9.8 CC 11 ND 3 FO 7 NS 2Critical = refactor now. High = refactor next time you touch it. Moderate = block increases. Low = leave it alone.
# Per-function explanations with refactoring advice
hotspots analyze . --format text --explain --top 10
# Block complexity regressions in CI
hotspots diff --auto-analyze --policy
# Compare any two git refs
hotspots diff main HEAD --top 10 --policy
# Interactive HTML report
hotspots analyze src/ --format html
# JSON for tooling/AI
hotspots analyze src/ --format json
# Train a repo-specific ranker from your bug history (always runs its own P@K eval)
hotspots train . --blame
# Driving this CLI from a coding agent? Start here
hotspots analyze --help-agentHotspots computes a Local Risk Score (LRS) per function from four structural metrics:
| Metric | What it measures |
|---|---|
| CC — Cyclomatic Complexity | Independent decision paths (if/loop/catch/&&/||) |
| ND — Nesting Depth | Maximum depth of nested control structures |
| FO — Fan-Out | Distinct functions called |
| NS — Non-Structured Exits | Early returns, throws, breaks |
LRS = 1.0×R_cc + 0.8×R_nd + 0.6×R_fo + 0.7×R_ns
where each component is log-scaled and capped to prevent outliers from dominating (full formula, caps, and weight rationale).
Every analyze run also combines LRS with the file's git history (churn, touch frequency, recency) into an absolute Activity Risk Score, and marks each function active (its file is touched more than the repo median, or changed in the last 30 days) or quiet:
Activity Risk = LRS/20.2 + churn + touch + recency (each term scaled to 0–1)
Activity Risk says where attention is likely needed next, not how good the code is. Editing a file raises it for every function in the file and it decays as the file goes quiet, so use the risk band (from LRS) to judge structure and Activity Risk to decide where to look first. See Activity Risk and Activity tier.
Risk bands: Low (< 3) · Moderate (3–6) · High (6–9) · Critical (≥ 9)
These formulas aren't hand-tuned guesses — every weight, cutoff, and signal that made it into the CLI was validated (or rejected) against real repo history and ground-truth bug/fix data in the open hotspots-public-research repo. See its findings index for the evidence behind each shipped feature, including the negative results that kept things out.
TypeScript · JavaScript · Go · Python · Rust · Java · C/C headers · C++ · C# · PHP · Vue · Astro
All supported file extensions (.ts, .tsx, .mts, .cts, .js, .jsx, .mjs, .cjs, .go, .py, .rs, .java, .c, .h, .cpp, .cc, .cxx, .hpp, .hxx, .hh, .cs, .php, .vue, .astro) work out of the box.
name: Hotspots
on: [pull_request, push]
jobs:
analyze:
runs-on: ubuntu-latest
permissions:
contents: read
pull-requests: write
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- uses: Stephen-Collins-tech/hotspots-action@v1
with:
github-token: ${{ secrets.GITHUB_TOKEN }}Posts PR comments, generates HTML reports, fails builds on policy violations.
# Fail CI if new critical functions introduced or LRS increases ≥ 1.0
hotspots diff --auto-analyze --policyExit code 1 on blocking violations, 0 on warnings only.
Policy engine — blocks: new critical-risk functions and LRS regressions ≥ 1.0.
Suppression comments — exclude functions from CI failures while keeping them visible:
// hotspots-ignore: legacy payment processor, rewrite scheduled Q2 2026
function legacyBillingLogic() { ... }Trained ranker — fit a RandomForest from your repo's bug-fix history (training always checks P@K vs. base rate before saving the model):
hotspots train . --blameOutput formats — text (terminal), json (machine), html (interactive), sarif (GitHub Code Scanning).
Configuration — .hotspotsrc.json in project root (auto-discovered):
{
"include": ["src/**/*.ts"],
"exclude": ["**/*.test.ts"],
"min_lrs": 0.0,
"policy": { "critical_introduction": "block" }
}- docs/COMMANDS.md — every command: why it exists, when to use it, how to read its output
- docs/USAGE.md — workflows, CI setup, output formats, policy engine, suppression, training, snapshot management
- docs/REFERENCE.md — complete CLI reference, all flags, config options, JSON schema, metrics formula
- docs/ARCHITECTURE.md — system design, analysis pipeline, invariants, design decisions
- docs/CONTRIBUTING.md — dev setup, adding languages, release process
- hotspots-public-research — the published evidence behind the scoring formulas: what was tested, what shipped, and what was rejected
Off by default. If you choose to enable it (set HOTSPOTS_TELEMETRY=on), hotspots sends one anonymous ping a day (version, OS, architecture, install channel, subcommand name, random install id) so I can see real usage and decide what to improve. No code, paths or results. Run hotspots analyze --show-config to see whether it's currently on, and unset HOTSPOTS_TELEMETRY (or set it to anything other than on) to turn it off. Details: docs/TELEMETRY.md.
vs ESLint complexity rules — ESLint checks one metric in isolation with no git context. Hotspots combines four metrics, git history, and call graph topology into a single prioritized list.
vs SonarQube / CodeClimate — Enterprise platforms requiring server infrastructure. Hotspots is a single binary, zero config, works offline, results in seconds.
vs code reviews — Reviews catch complexity subjectively and miss gradual drift. Hotspots enforces objective thresholds automatically on every commit.
- Bug reports: GitHub Issues
- Feature requests: GitHub Issues (describe the use case)
- PRs: see docs/CONTRIBUTING.md
MIT — see LICENSE-MIT.