データ更新のエンドポイントを作成し，定期的に叩くGitHubワークフローを追加

.github/workflows/update_data.yaml

@@ -0,0 +1,18 @@
+name: update data
+on:
+  schedule:
+    - cron: '0 18 * * *' # 毎日3時（JST）に実行
+  workflow_dispatch:
+
+jobs:
+  call:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Trigger HF Space FastAPI Endpoint
+        run: |
+          curl --fail --show-error --silent -X 'POST' \
+            '${{ vars.FASTAPI_URL }}/tasks/update' \
+            -H 'accept: application/json' \
+            -H 'Authorization: Bearer ${{ secrets.HF_TOKEN }}' \
+            -H 'X-API-KEY: ${{ secrets.API_SECRET_KEY }}' \
+            -d ''

app/main.py

@@ -2,6 +2,8 @@ import os
 import sys
 import json
 import gzip
+import subprocess
+from pathlib import Path
 from contextlib import asynccontextmanager
 
 from fastapi import FastAPI, Response, Depends, HTTPException, Security, Query
@@ -154,6 +156,10 @@ class SearchRequest(BaseModel):
     debug: bool = False
 
 
+class TaskUpdateResponse(BaseModel):
+    message: str
+
+
 @app.post(
     "/api/search",
     response_model=ProjectIds,
@@ -195,3 +201,45 @@ def search(request: SearchRequest):
             pass
 
     return ProjectIds(projectIds=ids)
+
+
+@app.post(
+    "/tasks/update",
+    response_model=TaskUpdateResponse,
+    dependencies=[Depends(get_api_key)],
+)
+def update_tasks():
+    project_root = Path(__file__).resolve().parents[1]
+    script_path = project_root / "scripts" / "build_all.py"
+    if not script_path.exists():
+        log.error("Requested update script not found: %s", script_path)
+        raise HTTPException(status_code=500, detail="Update script not found.")
+
+    try:
+        result = subprocess.run(
+            [sys.executable, str(script_path)],
+            check=True,
+            capture_output=True,
+            text=True,
+            cwd=str(project_root),
+        )
+    except subprocess.CalledProcessError as exc:
+        stdout = exc.stdout.strip() if exc.stdout else ""
+        stderr = exc.stderr.strip() if exc.stderr else ""
+        if stdout:
+            log.error("build_all.py stdout:\n%s", stdout)
+        if stderr:
+            log.error("build_all.py stderr:\n%s", stderr)
+        raise HTTPException(
+            status_code=500,
+            detail="Data update failed while running build_all.py.",
+        ) from exc
+
+    stdout = result.stdout.strip() if result.stdout else ""
+    stderr = result.stderr.strip() if result.stderr else ""
+    if stdout:
+        log.info("build_all.py stdout:\n%s", stdout)
+    if stderr:
+        log.warning("build_all.py stderr:\n%s", stderr)
+
+    return TaskUpdateResponse(message="Data update completed.")

scripts/0_download_data.py

@@ -17,14 +17,24 @@ fast_text_vec_path = get_file_path_from_config("embeddings.fasttext_vec")
 
 def download_embeddings() -> None:
     """Hugging Face Datasets から Embeddings をダウンロードする"""
+    target_files = {
+        "cc.ja.300.bin": Path(fast_text_bin_path),
+        "cc.ja.300.vec": Path(fast_text_vec_path),
+    }
+
+    missing_files = [name for name, path in target_files.items() if not path.exists()]
+    if not missing_files:
+        log.info("Embeddings already exist. Skipping download.")
+        return
+
     from huggingface_hub import hf_hub_download
 
     # フォルダがなかったら新規作成
-    target_dir = "resources/embeddings"
+    target_dir = Path(fast_text_bin_path).parent
     os.makedirs(target_dir, exist_ok=True)
     # huggingface_hub は local_dir 配下に .cache/huggingface を作るため先に用意
     try:
-        os.makedirs(os.path.join(target_dir, ".cache", "huggingface"), exist_ok=True)
+        os.makedirs(target_dir / ".cache" / "huggingface", exist_ok=True)
     except PermissionError:
         log.error(
             "権限エラー: %s に .cache/huggingface を作成できません。Dockerfile の権限設定を確認してください。",
@@ -37,31 +47,31 @@ def download_embeddings() -> None:
     dotenv.load_dotenv()  # .envから環境変数をロード
 
     # ダウンロード
-    if os.path.exists(fast_text_bin_path):
+    if "cc.ja.300.bin" not in missing_files:
         log.info("Embeddings binary already exists. Skipping download.")
     else:
         path = hf_hub_download(
             repo_id=os.getenv("HF_EMBEDDINGS_REPO_ID"),
             repo_type="dataset",
             filename="cc.ja.300.bin",
-            local_dir=target_dir,
+            local_dir=str(target_dir),
             token=os.getenv("HF_TOKEN"),
         )
         log.info(f"Embeddings downloaded: {path}")
-        _ensure_in_target_dir("cc.ja.300.bin", target_dir)
+        _ensure_in_target_dir("cc.ja.300.bin", str(target_dir))
 
-    if os.path.exists(fast_text_vec_path):
+    if "cc.ja.300.vec" not in missing_files:
         log.info("Embeddings vector already exists. Skipping download.")
     else:
         path = hf_hub_download(
             repo_id=os.getenv("HF_EMBEDDINGS_REPO_ID"),
             repo_type="dataset",
             filename="cc.ja.300.vec",
-            local_dir=target_dir,
+            local_dir=str(target_dir),
             token=os.getenv("HF_TOKEN"),
         )
         log.info(f"Embeddings downloaded: {path}")
-        _ensure_in_target_dir("cc.ja.300.vec", target_dir)
+        _ensure_in_target_dir("cc.ja.300.vec", str(target_dir))
 
     return