Surface agent code in OverallBestEntry for AgentDetailPage

CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## [0.15.12] — 2026-06-04
+
+### Changed
+- **`OverallBestEntry` model** (`app/models.py`): added `agent_path: str` and `commit_hash: str` to each per-spec best record. `GET /leaderboard/overall` and `GET /rounds/{id}/leaderboard` now carry the data needed to deep-link from an agent's profile straight to the exact GitHub commit of the code that achieved each score. This closes a flywheel gap — the dashboard's AgentDetailPage shows per-problem results but had no way to surface the code that earned them, so a contributor who wanted to fork the entry had to manually browse the repo. Both query paths in `app/routes/leaderboard.py` and `app/routes/rounds.py` updated to SELECT `s.agent_path, s.commit_hash` and pass them through to the model.
+
+### Tests
+- Existing 24 tests in `test_overall_leaderboard.py` + `test_rounds.py` continue to pass — the new fields are populated from columns the test fixtures already provide.
+
+---
+
 ## [0.15.11] — 2026-06-04
 
 ### Changed

app/models.py

@@ -140,6 +140,8 @@ class OverallBestEntry(BaseModel):
     normalized_score: float  # rank / (N+1) where N = entries for this spec; 1.0 = not entered
     submission_id: str
     submitted_at: datetime
+    agent_path: str  # repo path to this submission's agent dir (e.g. "agents/foo/v1")
+    commit_hash: str  # commit hash for deep-linking to the exact code on GitHub
 
 
 class OverallLeaderboardEntry(BaseModel):

app/routes/leaderboard.py

@@ -97,6 +97,7 @@ async def _compute_overall_leaderboard() -> OverallLeaderboard:
         SELECT s.id as submission_id, s.spec_id, s.contributor,
                s.mass_grams, COALESCE(s.score, s.mass_grams) as score,
                COALESCE(s.score_metric, 'mass_grams') as score_metric,
+               s.agent_path, s.commit_hash,
                s.submitted_at
         FROM submissions s
         WHERE s.spec_id IN ({spec_id_placeholders}) AND s.passed = 1
@@ -152,6 +153,8 @@ async def _compute_overall_leaderboard() -> OverallLeaderboard:
                     normalized_score=normalized,
                     submission_id=row["submission_id"],
                     submitted_at=row["submitted_at"],
+                    agent_path=row["agent_path"],
+                    commit_hash=row["commit_hash"],
                 )
             )
 

app/routes/rounds.py

@@ -168,6 +168,7 @@ async def get_round_leaderboard(round_id: str):
         SELECT s.id as submission_id, s.spec_id, s.contributor,
                s.mass_grams, COALESCE(s.score, s.mass_grams) as score,
                COALESCE(s.score_metric, 'mass_grams') as score_metric,
+               s.agent_path, s.commit_hash,
                s.submitted_at
         FROM submissions s
         WHERE s.spec_id IN ({placeholders}) AND s.passed = 1
@@ -216,6 +217,8 @@ async def get_round_leaderboard(round_id: str):
                     normalized_score=normalized,
                     submission_id=row["submission_id"],
                     submitted_at=row["submitted_at"],
+                    agent_path=row["agent_path"],
+                    commit_hash=row["commit_hash"],
                 )
             )