commit 546727726bc5f74d8c7f601d1df9c05db0b6a8a2
parent a77fea2239f377360119cca978f85b5ee019a29f
Author: Amit Dutta <mail@amit.is-a.dev>
Date: Fri, 4 Sep 2026 16:13:50 +0530
Improve error handling and response for static files
Refactor error handling and response types for static files.
Diffstat:
1 file changed, 2 insertions(+), 17 deletions(-)
diff --git a/app/main.py b/app/main.py
@@ -18,9 +18,6 @@ templates = Jinja2Templates(directory="app/templates")
STATIC_DIR = Path(__file__).parent / "static"
-# manifest.json and the service worker must always be fetched fresh —
-# the app can change at any time and we never want a stale PWA shell
-# or a stale install prompt. Icons are safe to cache normally.
_NO_STORE_FILES = {"manifest.json", "sw.js"}
@app.get("/static/{filename}")
@@ -58,9 +55,6 @@ RAW_BASE_URL = os.getenv("RAW_BASE_URL", f"https://{RAW_DOMAIN}")
RAW_PREFIX = "raw/"
def render_context(extra: dict) -> dict:
- """Merges page-specific context with repo-wide file count / size stats
- (shown in the footer on every page). repo_stats() is cached, so this
- doesn't add a fresh full-repo scan on every request."""
stats = repo_stats()
ctx = dict(extra)
ctx["repo_file_count"] = stats["file_count"] if stats else None
@@ -69,10 +63,8 @@ def render_context(extra: dict) -> dict:
@app.exception_handler(404)
async def not_found_handler(request: Request, exc: HTTPException):
- # Keep the raw subdomain returning plain text errors
if request.url.hostname == RAW_DOMAIN:
return PlainTextResponse("404: File Not Found", status_code=404)
- # Render UI 404 page for the main CDN domain
return templates.TemplateResponse(request, "index.html", render_context({"page": "404"}), status_code=404)
async def stream_raw(path: str):
@@ -88,17 +80,14 @@ async def stream_raw(path: str):
filename = path.split("/")[-1].lower()
guessed_type, _ = mimetypes.guess_type(filename)
- # Allow media and PDFs to render in-browser securely
if guessed_type and (
guessed_type.startswith("image/") or
guessed_type.startswith("video/") or
- guessed_type.startswith("audio/") or
+ guessed_type.startswith("audio/") or
guessed_type == "application/pdf"
):
content_type = guessed_type
else:
- # Force all other files (code, HTML, JSON, unknown) to plain text
- # This prevents XSS attacks on the raw domain
content_type = "text/plain; charset=utf-8"
headers = {
@@ -187,9 +176,7 @@ async def download_zip(path: str):
hf_url = f"https://huggingface.co/datasets/{HF_REPO_ID}/resolve/main/{f['path']}"
r = await client.get(hf_url)
if r.status_code != 200:
- continue # skip files that fail rather than aborting the whole zip
- # arcname: path relative to the requested folder, so the zip
- # doesn't contain the full repo path
+ continue
arcname = f["path"][prefix_len:] if prefix_len else f["path"]
zf.writestr(arcname, r.content)
@@ -210,13 +197,11 @@ async def ping():
async def serve(request: Request, path: str):
clean_path = path.strip("/")
- # 1. Intercept requests coming to the raw subdomain
if request.url.hostname == RAW_DOMAIN:
if not clean_path:
return HTMLResponse("Specify a file path.", status_code=200)
return await stream_raw(clean_path)
- # 2. Keep old /raw/ prefix working on main domain by redirecting to subdomain
if clean_path == RAW_PREFIX.rstrip("/") or clean_path.startswith(RAW_PREFIX):
raw_path = clean_path[len(RAW_PREFIX):]
if not raw_path:
