Refactor dashboard charts into per-chart cards and refresh the header

Makefile

@@ -31,4 +31,10 @@ install-hooks: ## Point git at the tracked .githooks/ (auto-rebuilds dist on com
 	git config core.hooksPath .githooks
 	@echo "core.hooksPath -> .githooks"
 
-.PHONY: help web-build web-dev web-check format format-check build techdata dev install-hooks
+gc: ## Repack git history (committed dist/repo-intel deltas down to ~nothing)
+	@before=$$(git count-objects -vH | awk '/size-pack:/{print $$2 $$3}'); \
+	git gc --quiet; \
+	after=$$(git count-objects -vH | awk '/size-pack:/{print $$2 $$3}'); \
+	echo "pack: $$before -> $$after"
+
+.PHONY: help web-build web-dev web-check format format-check build techdata dev install-hooks gc

README.md

@@ -270,6 +270,18 @@ Pushing a `vX.Y.Z` tag by hand still works as a fallback and runs the same
 `Release` workflow — but it skips the pre-tag build gate, so prefer _Cut
 release_.
 
+**Syncing tags locally.** Because `Release` _force-moves_ the floating `vX`
+major tag onto each new release commit, a plain `git fetch --tags` refuses to
+update it (`! [rejected] vX -> vX (would clobber existing tag)`). Pull the
+realigned tags with:
+
+```sh
+git fetch --tags --force --prune origin
+```
+
+The `vX.Y.Z` tags are immutable and always fetch cleanly; only the floating
+`vX` tag needs `--force`.
+
 ### Detection data (`techdata.json`)
 
 Language detection (extension/filename → language, colors, vendored-path noise

gen_techdata.py

@@ -41,6 +41,12 @@
     "r": "R", "pl": "Perl", "t": "Perl", "l": "Common Lisp", "v": "Verilog",
     "f": "Fortran", "for": "Fortran", "cls": "Apex", "pro": "Prolog",
     "ts": "TypeScript", "rs": "Rust", "cs": "C#", "sql": "SQL",
+    # Linguist's "Gettext Catalog" (.po/.pot) is type:prose and ships no color,
+    # so it'd be dropped and translation catalogs would vanish into "Other".
+    # Re-pin them under Linguist's own name with the gold GitHub falls back to
+    # for colorless languages (Primer --bgColor-attention-emphasis; see
+    # SYNTHETIC_COLORS).
+    "po": "Gettext Catalog", "pot": "Gettext Catalog",
 }
 
 # Generic extensions whose canonical Linguist owner is the colorless "Text"
@@ -141,7 +147,7 @@
 
 # Colors for synthetic framework groups Linguist doesn't define a language for.
 # Purple keeps "Tools" distinct from the grey "Other" bucket on the same page.
-SYNTHETIC_COLORS = {"Tools": "#a371f7"}
+SYNTHETIC_COLORS = {"Tools": "#a371f7", "Gettext Catalog": "#9e6a03"}
 
 # Backend / non-JS sentinel files: basename (or sub-path) → (framework, language).
 # The "Tools" bucket surfaces build/devops tooling that's present as a config
@@ -254,12 +260,13 @@ def build_language_tables(langs):
                     ext_meta[key] = (rank, primary)
         for fn in info.get("filenames", []):
             filename_lang.setdefault(fn.lower(), eff)
+    name_color.update(SYNTHETIC_COLORS)  # synthetic buckets Linguist doesn't color;
+                                         # merged before EXT_OVERRIDE so its guard sees them
     for ext, lang in EXT_OVERRIDE.items():
         if lang in name_color:
             ext_lang[ext] = lang
     for ext in EXT_EXCLUDE:
         ext_lang.pop(ext, None)
-    name_color.update(SYNTHETIC_COLORS)  # synthetic buckets Linguist doesn't color
     return name_color, ext_lang, filename_lang
 
 

repo-intel.py

@@ -460,7 +460,7 @@ def classify_path(field, present=None, shebang=None):
     return OTHER_LANG
 
 
-def top_languages(langs, limit=6):
+def top_languages(langs, limit=10):
     """Build a sorted language-bar list from {name: [added, deleted, files]}.
 
     Ranks by lines touched (added + deleted); languages past `limit` collapse
@@ -1017,6 +1017,15 @@ def collect_local(cwd=None, suppress_current_user=False):
         "frameworks": detect_frameworks(present, cwd=cwd),
         "file_count": len(present),
         "branch_count": count_branches(cwd=cwd),
+        # Whole-repo merge count for the header's "(N merges)" note. The stats
+        # pipeline runs --no-merges, so the headline total is rebuilt as
+        # no-merge commits + this. It can't be filtered (merges aren't in
+        # commits_meta to recount), so main() drops it under any --since/--until/
+        # --commits filter rather than inflate the now-filtered headline.
+        "merge_count": int(git("rev-list", "--merges", "--count", "HEAD", cwd=cwd).strip() or 0),
+        # HEAD's commit date (includes merges) for the header's "updated N ago".
+        # Whole-repo like the counts above; the browser renders it relative to now.
+        "last_commit_iso": git("log", "-1", "--format=%cI", cwd=cwd).strip(),
         "largest_files": head_file_sizes(cwd=cwd),
         "disk_by_path": history_disk_by_path(cwd=cwd),
     }
@@ -1122,6 +1131,33 @@ def probe_remote_total(owner, repo, token):
     return total if isinstance(total, int) else None
 
 
+def fetch_repo_social(github_base, token):
+    """Stars / watchers / forks for a github.com repo via REST; None otherwise
+    (non-GitHub origin, offline, private repo without a token, etc.)."""
+    m = ORIGIN_RE.match(github_base or "")
+    host = (m.group("https_host") or m.group("ssh_host")) if m else ""
+    if not m or not (host == "github.com" or host.endswith(".github.com")):
+        return None
+    url = f"https://api.github.com/repos/{m.group('owner')}/{m.group('repo')}"
+    headers = {"Accept": "application/vnd.github+json", "User-Agent": "repo-intel"}
+    if token:
+        headers["Authorization"] = f"Bearer {token}"
+    try:
+        with urllib.request.urlopen(
+            urllib.request.Request(url, headers=headers), timeout=5
+        ) as resp:
+            body = json.loads(resp.read().decode())
+    except (urllib.error.URLError, TimeoutError, ValueError):
+        return None
+    return {
+        "stars": body.get("stargazers_count"),
+        # REST quirk: "Watchers" in GitHub's UI is subscribers_count;
+        # watchers_count is a legacy alias of the star count.
+        "watchers": body.get("subscribers_count"),
+        "forks": body.get("forks_count"),
+    }
+
+
 def get_github_token():
     try:
         token = subprocess.check_output(
@@ -1196,7 +1232,10 @@ def fetch_user_profiles(logins, token):
     unique = []
     seen = set()
     for login in logins:
-        if login and login not in seen:
+        # Bot accounts (e.g. "github-actions[bot]") are GraphQL Bot nodes, not
+        # User nodes — user(login:) can never resolve them, so querying just
+        # yields a NOT_FOUND error. Skip them rather than emit a noisy warning.
+        if login and login not in seen and not login.endswith("[bot]"):
             seen.add(login)
             unique.append(login)
     if not unique:
@@ -1219,8 +1258,12 @@ def fetch_user_profiles(logins, token):
     except urllib.error.URLError as exc:
         print(f"  warning: profile fetch failed: {exc}", file=sys.stderr)
         return {}
-    if "errors" in body:
-        print(f"  warning: profile fetch errors: {body['errors']}", file=sys.stderr)
+    # NOT_FOUND just means a login no longer resolves (deleted/renamed account);
+    # the per-alias `if not node: continue` below already skips those. Only warn
+    # about other errors (auth, rate limits) that signal a real fetch problem.
+    real_errors = [e for e in body.get("errors", []) if e.get("type") != "NOT_FOUND"]
+    if real_errors:
+        print(f"  warning: profile fetch errors: {real_errors}", file=sys.stderr)
     data = body.get("data") or {}
     out = {}
     for i, login in enumerate(unique):
@@ -1778,6 +1821,12 @@ def build_data(
     # Total branch count: a snapshot stat like file_count. None on paths that
     # can't determine it.
     branch_count = (extras or {}).get("branch_count")
+    merge_count = (extras or {}).get("merge_count") or 0
+    # HEAD commit date for "updated N ago"; fall back to the newest commit we
+    # have (the remote GraphQL path supplies no extras).
+    last_commit_iso = (extras or {}).get("last_commit_iso") or max(
+        (m.get("iso") or "" for m in commits_meta.values()), default=""
+    )
     # File-size sunbursts: blob sizes at HEAD and accumulated on-disk size per
     # path across history. None on the remote GraphQL path (no tree fetched).
     largest_files = (extras or {}).get("largest_files")
@@ -1920,8 +1969,15 @@ def build_data(
         "largestFiles": largest_files,
         "diskByPath": disk_by_path,
         "dateRange": date_range,
+        "lastCommit": last_commit_iso,
+        "generatedAt": datetime.now(timezone.utc).isoformat(),
         "totals": {
+            # Authored (no-merge) commits — equals the sum of contributors'
+            # commits, so it's the denominator for every share/percentage.
+            # Merges are tracked separately; the GitHub-comparable headline
+            # (commits + merges) is composed at display time.
             "commits": total_commits,
+            "merges": merge_count,
             "added": total_added,
             "deleted": total_deleted,
             "contributors": total_contributors,
@@ -2073,15 +2129,20 @@ def render_markdown(data):
     out.append(f"# repo-intel — {f'[{name}]({base})' if base else name}")
     out.append("")
     n_contrib = t["contributors"]
+    # Headline commit count matches GitHub (authored + merges); the per-author
+    # stats below use the merge-free t["commits"].
+    merges = t.get("merges") or 0
+    commits_all = t["commits"] + merges
+    merge_note = f" ({merges:,} merge{'' if merges == 1 else 's'})" if merges else ""
     out.append(
-        f"_{t['commits']:,} commits · {dr['start'] or '?'} — {dr['end'] or '?'} · "
+        f"_{commits_all:,} commits · {dr['start'] or '?'} — {dr['end'] or '?'} · "
         f"{n_contrib:,} contributor{'' if n_contrib == 1 else 's'}_"
     )
     out.append("")
 
     out.append("## Totals")
     out.append("")
-    out.append(f"- Commits: {t['commits']:,}")
+    out.append(f"- Commits: {commits_all:,}{merge_note}")
     out.append(f"- Lines: +{t['added']:,} / -{t['deleted']:,}")
     out.append(f"- Contributors: {t['contributors']:,}")
     out.append(f"- Default branch: `{data['defaultBranch']}`")
@@ -2251,6 +2312,13 @@ def main():
         )
         if not commits_meta:
             sys.exit("error: no commits match the given filters")
+        # The merge tally is whole-history (collected before filtering, and
+        # merges aren't in commits_meta to recount), so adding it to the now
+        # date/subset-filtered authored total would inflate the "N commits"
+        # headline. Drop it under any filter — the headline then reads as just
+        # the commits in range.
+        if extras:
+            extras.pop("merge_count", None)
 
     tags = filter_tags_to_range(tags, commits_meta)
 
@@ -2271,6 +2339,10 @@ def main():
 
     enrich_contributor_profiles(data["contributors"], commits_meta, github_base, token=token)
 
+    social = fetch_repo_social(github_base, token or get_github_token())
+    if social:
+        data.update(social)
+
     multi = len(formats) > 1
     builders = {
         "html": lambda: render_html(data),
@@ -2287,7 +2359,7 @@ def main():
         print(f"Wrote {out_path}")
 
     print(
-        f"  {data['totals']['commits']} commits · "
+        f"  {data['totals']['commits'] + (data['totals'].get('merges') or 0)} commits · "
         f"{data['dateRange']['start']} — {data['dateRange']['end']} · "
         f"{data['totals']['contributors']} contributor"
         f"{'' if data['totals']['contributors'] == 1 else 's'}"

techdata.json

@@ -489,6 +489,7 @@
    "Gentoo Ebuild": "#9400ff",
    "Gentoo Eclass": "#9400ff",
    "Gerber Image": "#d20b00",
+   "Gettext Catalog": "#9e6a03",
    "Gherkin": "#5B2063",
    "Git Attributes": "#F44D27",
    "Git Commit": "#F44D27",
@@ -1790,12 +1791,14 @@
    "pm6": "Raku",
    "pml": "Promela",
    "pmod": "Pike",
+   "po": "Gettext Catalog",
    "podsl": "Common Lisp",
    "podspec": "Ruby",
    "pogo": "PogoScript",
    "polar": "Polar",
    "por": "Portugol",
    "postcss": "CSS",
+   "pot": "Gettext Catalog",
    "pov": "POV-Ray SDL",
    "pp": "Puppet",
    "pprx": "REXX",

web/index.html

@@ -3,6 +3,10 @@
   <head>
     <meta charset="UTF-8" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <!-- Referenced as a build asset (not a public/ file) so vite-plugin-singlefile
+         inlines it as a base64 data URI — the generated dashboard is a single HTML
+         opened over file:// from /tmp, so a sibling/remote icon would never load. -->
+    <link rel="icon" type="image/png" href="./src/assets/favicon.png" sizes="96x96" />
     <title>Repo Intel</title>
     <!-- repo-intel.py replaces this marker with `window.__DATA__ = {...};`.
          The injectionMarker() Vite plugin guarantees it survives the build. -->

web/src/App.svelte

@@ -11,7 +11,8 @@
     createTimelineTooltip,
     buildPunchPoints,
   } from "$lib/popovers";
-  import { fmtTimelineDuration } from "$lib/format";
+  import { authorUrl, fmtTimelineDuration, relativeTime, fmtDateTime } from "$lib/format";
+  import { setAuthorTotalCommits } from "$lib/popover-store.svelte";
   import { registerEchartsTheme } from "$lib/theme";
   import { buildTimeline } from "$lib/timeline";
   import { dragScroll, scrollSpy } from "$lib/actions";
@@ -41,6 +42,10 @@
     timelineDur ? `Commit timeline: ${timelineDur}` : "Commit timeline",
   );
 
+  // "Last commit N ago" shown in the Contributions header.
+  const lastCommitAgo = $derived(relativeTime(data.lastCommit));
+  const lastCommitFull = $derived(fmtDateTime(data.lastCommit));
+
   // The author popover (shared by the table and the timeline lane labels) and the
   // commit-bucket popover (opened by the punch-card cells) write the shared
   // popover store rendered by <AuthorPopover/> / <CommitPopover/>. Created in
@@ -61,6 +66,7 @@
   // The timeline is still rendered imperatively into the container elements below
   // (it's a hand-drawn canvas); wire it once the static layout is mounted.
   onMount(() => {
+    setAuthorTotalCommits(data.totals.commits);
     authorPopover = createAuthorPopover(data.contributors);
     commitPopover = createCommitPopover(data);
     buildTimeline(data, authorPopover, createTimelineTooltip());
@@ -89,7 +95,16 @@
       <div class="section" id="contributions">
         <div class="card">
           <div class="contributions-header">
-            <h2>Contributions</h2>
+            <div class="contributions-title">
+              <h2>Contributions</h2>
+              {#if lastCommitAgo}
+                <span class="last-commit"
+                  >Last commit <time datetime={data.lastCommit} title={lastCommitFull}
+                    >{lastCommitAgo}</time
+                  ></span
+                >
+              {/if}
+            </div>
             <YearToggles {data} onSelect={(mode) => (heatmapMode = mode)} />
           </div>
           <Heatmap {data} mode={heatmapMode} />
@@ -119,14 +134,16 @@
       </div>
       <div class="section" id="overall">
         <h2>Overall</h2>
-        <OverallCharts {data} />
+        <OverallCharts {data} {authorPopover} />
       </div>
       <div class="section" id="commit-frequency">
         <h2>Commit frequency over time</h2>
         <div class="grid-5">
           {#each data.contributors as c, i (c.email)}
             <ContributorCard
+              {authorPopover}
               contributor={c}
+              url={authorUrl(data, c)}
               index={i}
               weeks={data.weeks}
               weekly={data.weeklyData[c.email]}
@@ -139,7 +156,9 @@
         <div class="scroll-row" use:dragScroll>
           {#each data.contributors as c, i (c.email)}
             <PatternCard
+              {authorPopover}
               contributor={c}
+              url={authorUrl(data, c)}
               index={i}
               points={punchData[c.email] ?? []}
               {commitPopover}

web/src/lib/chart-helpers.ts

@@ -0,0 +1,44 @@
+// Shared scaffolding for the repo-wide OverallCharts cards. Contributors are
+// aggregated by email, so one person who committed under several addresses (old
+// work emails, a noreply alias) shows up as several rows that all share one
+// display name. ECharts keys legend selection by the series/slice name, so
+// identical names would collapse into a single legend entry and toggle together.
+// Every card therefore names its series/slice by the (unique) email under the
+// hood, hides ECharts' own legend, and renders the HTML legend (ChartLegend)
+// below — plain names, one entry per identity. These helpers map email ↔ name ↔
+// original contributor index so the cards can share that mapping.
+import { clr } from "$lib/theme";
+import type { Contributor } from "$types";
+
+// Legend rows for the weekly-commits line and the commit-share pie. `idx` is the
+// original contributor index (so clr() and the author popover line up); the pie
+// gains a non-author "Others" row (idx -1 → no popover) when the top-N don't
+// cover every commit.
+export type LegendItem = { key: string; name: string; color: string; idx: number };
+
+// Email → display name, for tooltips that key series by the unique email.
+export const buildNameByEmail = (contributors: Contributor[]): Map<string, string> =>
+  new Map(contributors.map((c) => [c.email, c.name] as const));
+
+// Email → original contributor index, so a chart that filters/re-sorts a local
+// copy (the churn bars) can still resolve the right person's colour and popover.
+export const buildEmailToOrig = (contributors: Contributor[]): Map<string, number> =>
+  new Map(contributors.map((c, i) => [c.email, i] as const));
+
+export const buildContribLegend = (contributors: Contributor[]): LegendItem[] =>
+  contributors.map((c, i) => ({ key: c.email, name: c.name, color: clr(i), idx: i }));
+
+// Contributors with their original index, minus bots. A `[bot]` login (the ones
+// repo-intel.py skips for profiles — Renovate, CI accounts) churns nothing like a
+// human and only flattens everyone else's scale, so the churn and commit-style
+// charts drop them. `origIdx` is preserved so each row still resolves the right
+// identity colour (clr) and author popover after the chart re-sorts its own copy.
+export const humanContribRows = (
+  contributors: Contributor[],
+): { c: Contributor; origIdx: number }[] =>
+  contributors.map((c, origIdx) => ({ c, origIdx })).filter((r) => !r.c.login.endsWith("[bot]"));
+
+// Half-transparent dark inner border on the treemap tiles, so each grey/brand
+// tile reads against the gap around it (shared by the languages and files
+// treemaps).
+export const tileInnerBorder = "rgba(0, 0, 0, 0.6)";