fix: survive an LFS pointer at data/metrics.db in the daily job

.gitattributes tracks data/*.db under Git LFS and actions/checkout@v4
defaults to lfs: false, so on the CI runner data/metrics.db is a ~130-byte
pointer: DB_PATH.exists() is True and sqlite3 then raises
"file is not a database". The daily consensus job would traceback every
single day, on the one archive whose gaps cannot be backfilled from any
public source.

Both halves fixed:

- the workflow pulls only data/metrics.db from LFS. `lfs: true` on the
checkout would fetch ~100 MB of LFS objects daily (80 MB of it
data/chroma/chroma.sqlite3, unread by this job) and exhaust GitHub's LFS
bandwidth quota within weeks — and an exhausted quota reinstates the
pointer, i.e. the bug.
- list_tickers() catches sqlite3.DatabaseError and returns [], so the
caller's existing "no covered tickers found" path fires with a readable
message and exit code 1 instead of an unhandled traceback.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.github/workflows/snapshot-consensus.yml

@@ -20,6 +20,22 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
+      # data/metrics.db is tracked by Git LFS (.gitattributes: `data/*.db`),
+      # and actions/checkout@v4 defaults to `lfs: false` — so without this the
+      # file on the runner is a ~130-byte pointer, storage/metrics_db.py opens
+      # it, and sqlite3 raises "file is not a database". The job would fail
+      # every day, on the one archive whose missed days cannot be recovered
+      # from any public source.
+      #
+      # This pulls ONLY metrics.db (~1 MB) rather than setting `lfs: true` on
+      # the checkout, which would fetch every LFS object in the repo — ~100 MB,
+      # 80 MB of it data/chroma/chroma.sqlite3, none of which this job reads.
+      # At a daily cadence that would exhaust GitHub's LFS bandwidth quota
+      # within a couple of weeks, and an exhausted quota fails the LFS fetch,
+      # which puts the pointer file straight back. Fetch what the job needs.
+      - name: Fetch the metrics DB from LFS
+        run: git lfs pull --include="data/metrics.db"
+
       - uses: actions/setup-python@v5
         with:
           python-version: "3.11"

storage/metrics_db.py

@@ -289,16 +289,28 @@ def list_tickers() -> list[str]:
     This is the app's de facto coverage universe: there is no explicit
     coverage table, and scripts/warm_runtime_cache.py and deploy_check.py
     have both derived it with this same query.
+
+    A path that exists but is not a SQLite database yields [] rather than
+    raising. That is not hypothetical: `.gitattributes` tracks `data/*.db`
+    under Git LFS, and `actions/checkout@v4` defaults to `lfs: false`, so on
+    a CI runner without an LFS pull this file is a ~130-byte pointer and
+    `sqlite3.connect(...).execute(...)` raises
+    `sqlite3.DatabaseError: file is not a database`. Callers already handle
+    "no covered tickers" with a readable message; an unhandled traceback
+    would just be a noisier way of saying the same thing.
     """
     if not DB_PATH.exists():
         return []
-    with sqlite3.connect(DB_PATH) as conn:
-        return [
-            row[0]
-            for row in conn.execute(
-                "SELECT DISTINCT ticker FROM metrics ORDER BY ticker"
-            )
-        ]
+    try:
+        with sqlite3.connect(DB_PATH) as conn:
+            return [
+                row[0]
+                for row in conn.execute(
+                    "SELECT DISTINCT ticker FROM metrics ORDER BY ticker"
+                )
+            ]
+    except sqlite3.DatabaseError:
+        return []
 
 
 def get_all_metrics(ticker: str) -> list[dict]:

tests/test_metrics_db.py

@@ -168,3 +168,25 @@ def test_list_tickers_on_a_missing_database_returns_empty(tmp_path, monkeypatch)
     import storage.metrics_db as m
     monkeypatch.setattr(m, "DB_PATH", tmp_path / "absent.db")
     assert m.list_tickers() == []
+
+
+def test_list_tickers_on_a_non_database_file_returns_empty(tmp_path, monkeypatch):
+    """A Git LFS pointer is a text file that exists but is not SQLite.
+
+    `.gitattributes` tracks `data/*.db` under LFS, and `actions/checkout@v4`
+    defaults to `lfs: false`, so on a CI runner without an LFS pull the path
+    exists and holds a ~130-byte pointer. Opening it raises
+    `sqlite3.DatabaseError: file is not a database`. list_tickers must degrade
+    to "no coverage" so the caller prints its own message and exits 1, rather
+    than dying with an unhandled traceback.
+    """
+    import storage.metrics_db as m
+    pointer = tmp_path / "metrics.db"
+    pointer.write_text(
+        "version https://git-lfs.github.com/spec/v1\n"
+        "oid sha256:0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef\n"
+        "size 4194304\n",
+        encoding="utf-8",
+    )
+    monkeypatch.setattr(m, "DB_PATH", pointer)
+    assert m.list_tickers() == []