feat: pzmm conflict detection + content-type categorization

- mod_files manifest table populated at parse time
- POST /api/conflicts endpoint
- mod_types fingerprinting feeds derive_category
- DD filelist regex broadened to cover conflict-eligible exts
- media/maps/<*>/* excluded from manifest (per-mod namespaced,
  no conflict value, can be tens of MB per mod)

Plan: docs/plans/2026-05-04-pzmm-conflict-and-typing.md

api/app.py

@@ -26,6 +26,7 @@ from pydantic import BaseModel, Field
 
 import adapters
 import db
+import diagnostics
 import expansion
 import jobs
 import steam
@@ -191,6 +192,7 @@ def _row_to_modinfo(r) -> ModInfo:
         maps=list(r["maps"] or []),
         is_addon=bool(r["is_addon"]) if "is_addon" in r else False,
         workshop_tags=list(r["workshop_tags"] or []) if "workshop_tags" in r else [],
+        mod_types=list(r["mod_types"] or []) if "mod_types" in r else [],
     )
 
 
@@ -684,7 +686,7 @@ async def _build_result_for_job(
         SELECT mp.workshop_id, mp.mod_id, mp.name, mp.category,
                mp.requirements, mp.load_after, mp.load_before,
                mp.incompatible_mods, mp.load_first, mp.load_last,
-               mp.tags, mp.maps, mp.is_addon, wm.tags AS workshop_tags
+               mp.tags, mp.maps, mp.is_addon, mp.mod_types, wm.tags AS workshop_tags
           FROM mod_parsed mp
           JOIN workshop_meta wm ON wm.workshop_id = mp.workshop_id
          WHERE mp.workshop_id = ANY($1::text[])
@@ -1001,7 +1003,7 @@ async def sort_endpoint(req: SortRequest, request: Request) -> Dict[str, Any]:
                 SELECT mp.workshop_id, mp.mod_id, mp.name, mp.category,
                        mp.requirements, mp.load_after, mp.load_before,
                        mp.incompatible_mods, mp.load_first, mp.load_last,
-                       mp.tags, mp.maps, mp.is_addon, wm.tags AS workshop_tags
+                       mp.tags, mp.maps, mp.is_addon, mp.mod_types, wm.tags AS workshop_tags
                 FROM mod_parsed mp
                 JOIN workshop_meta wm ON wm.workshop_id = mp.workshop_id
                 WHERE mp.workshop_id = ANY($1::text[])
@@ -1215,7 +1217,7 @@ async def resort_endpoint(req: ResortRequest, request: Request) -> Dict[str, Any
             SELECT mp.workshop_id, mp.mod_id, mp.name, mp.category,
                    mp.requirements, mp.load_after, mp.load_before,
                    mp.incompatible_mods, mp.load_first, mp.load_last,
-                   mp.tags, mp.maps, mp.is_addon, wm.tags AS workshop_tags
+                   mp.tags, mp.maps, mp.is_addon, mp.mod_types, wm.tags AS workshop_tags
             FROM mod_parsed mp
             JOIN workshop_meta wm ON wm.workshop_id = mp.workshop_id
             WHERE mp.workshop_id IN (SELECT workshop_id FROM selected_wsids)
@@ -1469,6 +1471,71 @@ async def vote_broken_mod(
     return {"upvotes": int(row["upvotes"]), "downvotes": int(row["downvotes"])}
 
 
+@app.post("/api/conflicts")
+async def conflicts_endpoint(req: SortRequest, request: Request) -> Dict[str, Any]:
+    """Detect rel_paths claimed by ≥2 input mods with non-equal sha1.
+
+    v1: bare wsids only. Collection input returns 400 so the caller can
+    resolve via /api/sort first (where the async-job + drain-progress
+    plumbing already lives). Mods whose `files_manifest_built` is false
+    cannot be analyzed and are reported in `missing_manifests` instead of
+    silently ignored.
+    """
+    bare_wsids, collection_ids = parse_with_collections(req.input or "")
+    if collection_ids:
+        raise HTTPException(
+            status_code=400,
+            detail="conflict scan does not support collection input; resolve via /api/sort first",
+        )
+    if not bare_wsids:
+        raise HTTPException(status_code=400, detail="no workshop ids found in input")
+    if len(bare_wsids) > MAX_IDS:
+        raise HTTPException(
+            status_code=413,
+            detail=f"too many workshop ids ({len(bare_wsids)} > {MAX_IDS})",
+        )
+
+    pool = request.app.state.db
+    async with pool.acquire() as conn:
+        rows = await conn.fetch(
+            """
+            SELECT mp.workshop_id, mp.mod_id, mp.name, mp.category,
+                   mp.requirements, mp.load_after, mp.load_before,
+                   mp.incompatible_mods, mp.load_first, mp.load_last,
+                   mp.tags, mp.maps, mp.is_addon, mp.mod_types,
+                   mp.files_manifest_built, wm.tags AS workshop_tags
+              FROM mod_parsed mp
+              JOIN workshop_meta wm ON wm.workshop_id = mp.workshop_id
+             WHERE mp.workshop_id = ANY($1::text[])
+               AND mp.parsed_at_time_updated = wm.time_updated
+             ORDER BY mp.workshop_id, mp.mod_id
+            """,
+            bare_wsids,
+        )
+
+        mods: List[ModInfo] = [_row_to_modinfo(r) for r in rows]
+
+        # Missing-manifest wsids: input wsids that have no mod_parsed rows
+        # OR whose rows all have files_manifest_built=false. Any single
+        # built row in a multi-mod wsid counts as "manifest available".
+        wsid_has_manifest: Dict[str, bool] = {}
+        for r in rows:
+            w = r["workshop_id"]
+            built = bool(r["files_manifest_built"])
+            wsid_has_manifest[w] = wsid_has_manifest.get(w, False) or built
+        missing_manifests = [w for w in bare_wsids if not wsid_has_manifest.get(w, False)]
+
+        conflicts = await diagnostics.scan_file_conflicts(conn, mods)
+
+    return {
+        "conflicts": [
+            {"rel_path": c.rel_path, "providers": c.providers, "winner": c.winner}
+            for c in conflicts
+        ],
+        "missing_manifests": missing_manifests,
+    }
+
+
 # ── static frontend ────────────────────────────────────────────────────────
 # Mount LAST so all API routes win path resolution.
 _FRONTEND_DIR = Path(__file__).resolve().parent.parent / "frontend"