Code stats — every metric tool turned on (9-tool stack)
Code stats — every metric tool turned on
Decision
Every public repo in the chirag127/oriz* family runs nine
code-quality / code-stats tools — every metric the family can
auto-track for free, turned on, with no manual upkeep.
| # | Tool | What it owns | Where it renders |
|---|---|---|---|
| 1 | Sonarcloud | SAST + code smells + duplication + complexity | Quality gate on main |
| 2 | CodeRabbit | LLM-grade design + intent review | PR comments |
| 3 | Codecov | Per-PR coverage delta | PR comment + status check |
| 4 | Code Climate Quality | A — F maintainability per file | Dashboard + status check |
| 5 | DeepSource | Static analysis + autofix PRs | Issue list + auto-PR |
| 6 | biome (in repo, not a SaaS) | Lint + format + simple-bug catch | Local + CI |
| 7 | GitHub Insights | Native contributors / commits / code-frequency / dependents | Repo /pulse + /graphs/* |
| 8 | Tokei | Per-language line / file / blank / comment counts | CI artefact + family /stats |
| 9 | Lines of Code badge | Single LoC number | README badge per repo |
That is intentionally 9 tools. The user direction was: "ADD EVERYTHING — GitHub Insights + Tokei + CodeClimate + LinesOfCode badges." Builds on (does not supersede) the 5-tool code-quality decision — the 5 quality tools stay; the 4 stat-shaped tools (biome, GH Insights, Tokei, LoC badge) are the auto-tracking-friendly metric layer alongside.
Why every tool
Each tool catches or surfaces a different signal and renders to a different surface, so layering them costs nothing (all free for public OSS) and makes a different reviewer comfortable:
- SAST + smells = Sonarcloud (depth) + DeepSource (autofix) + biome (fast)
- Coverage = Codecov (PR delta), reinforced by Sonarcloud's gate
- Maintainability = Code Climate's A — F grade, the cheapest glance for "is this file getting worse?"
- Design + intent = CodeRabbit (LLM PR review)
- Native repo metrics = GitHub Insights — commit cadence, contributor list, code-frequency graph, dependency graph, dependents — none of which the SAST tools surface
- Line counts = Tokei (canonical, scriptable JSON output) + LoC badge (visible README at-a-glance)
Auto-tracked everywhere — every metric is computed without a human
pressing "refresh", in line with
auto-only-tracking. No manual
LOC counting, no manual contributor lists, no manual maintainability
calls.
The user posture was "use everything … so that everything is done best." For public-OSS repos, "everything" costs nothing.
Implications
templates/per-site-ci/.github/workflows/ci.ymlruns Tokei in CI as the line-counter step, emits JSON, uploads as a workflow artefact AND publishes to the family/statspage perfamily-wide-stats-page.- README badges — every repo's README ships the LoC badge from
lines-of-code-badge, alongside the existing quality / coverage / Sonarcloud / CodeRabbit badges. Templated viatemplates/per-site-ci/so new repos get the badge by default. - GitHub Insights is enabled by default on public repos — no
setup. The decision is to use it (link to
/pulsefrom family /stats; consume/graphs/contributorsdata via REST API at build time for the family stats page), not configure it. - Quota math — all 9 tools free for OSS / public repos under
the family's
public-everywhere posture;
no card on file required for any. Per
rules/never-hit-quotas: if any tool tightens its OSS tier in future, that single tool drops; the rest stay. - Build-time aggregation — the family
/statspage onoriz.in(perfamily-wide-stats-page) rolls up Tokei + GitHub Insights + Codecov + Code Climate / DeepSource badges across every repo into one public dashboard.
Cross-refs
- Code quality services index
- 5-tool code-quality decision (extended here)
- Family-wide /stats page decision
- GitHub Insights service
- Tokei service
- Lines of Code badge service
- Auto-only-tracking rule (forward ref — being added in parallel)
- Repos work independently rule
- Never hit quotas rule
- No card-on-file rule