Upload app.py with huggingface_hub

app.py

@@ -2,8 +2,10 @@ from __future__ import annotations
 
 import csv
 import hashlib
+import io
 import json
 import random
+import re
 import time
 from pathlib import Path
 
@@ -30,6 +32,75 @@ FIELDNAMES = [
 USE_GITHUB = github_configured()
 
 
+def _load_reviewer_tokens() -> dict[str, str]:
+    import os
+    tokens: dict[str, str] = {}
+    for key, val in os.environ.items():
+        if key.startswith("REVIEWER_TOKEN_"):
+            reviewer_id = key.replace("REVIEWER_TOKEN_", "").lower()
+            tokens[val] = reviewer_id
+    try:
+        reviewer_secrets = st.secrets.get("reviewers", {})
+        for reviewer_id, token in reviewer_secrets.items():
+            tokens[str(token)] = str(reviewer_id)
+    except (KeyError, FileNotFoundError):
+        pass
+    return tokens
+
+
+def _authenticate(token: str) -> str | None:
+    valid_tokens = _load_reviewer_tokens()
+    if not valid_tokens:
+        if re.match(r"^[a-zA-Z0-9_-]{2,30}$", token):
+            return token
+        return None
+    return valid_tokens.get(token)
+
+
+def _safe_reviewer_id(reviewer_id: str) -> str:
+    return hashlib.sha256(reviewer_id.encode()).hexdigest()[:12]
+
+
+def _safe_filename_id(value: str) -> str:
+    safe = re.sub(r"[^a-zA-Z0-9_-]+", "_", value).strip("_")
+    return safe or _safe_reviewer_id(value)
+
+
+def _reviews_to_csv(reviews: dict[int, dict]) -> str:
+    output = io.StringIO()
+    writer = csv.DictWriter(output, fieldnames=FIELDNAMES, extrasaction="ignore")
+    writer.writeheader()
+    for idx in sorted(reviews.keys()):
+        writer.writerow(reviews[idx])
+    return output.getvalue()
+
+
+def _strip_dicom_metadata(file_path: Path) -> bytes | None:
+    suffix = file_path.suffix.lower()
+    if suffix in (".png", ".jpg", ".jpeg", ".bmp", ".tiff", ".tif"):
+        return file_path.read_bytes()
+    if suffix in (".dcm", ".dicom"):
+        try:
+            import pydicom
+            from PIL import Image
+
+            ds = pydicom.dcmread(str(file_path))
+            pixel_array = ds.pixel_data
+            if hasattr(ds, "pixel_array"):
+                pixel_array = ds.pixel_array
+            else:
+                return None
+            img = Image.fromarray(pixel_array)
+            if img.mode not in ("L", "RGB"):
+                img = img.convert("L")
+            buf = io.BytesIO()
+            img.save(buf, format="PNG")
+            return buf.getvalue()
+        except Exception:
+            return None
+    return file_path.read_bytes() if file_path.exists() else None
+
+
 def load_pairs() -> list[dict]:
     if USE_GITHUB:
         import github_storage
@@ -45,7 +116,7 @@ def load_pairs() -> list[dict]:
 
 def get_results_path(reviewer_id: str) -> Path:
     RESULTS_DIR.mkdir(parents=True, exist_ok=True)
-    safe_id = hashlib.sha256(reviewer_id.encode()).hexdigest()[:12]
+    safe_id = _safe_reviewer_id(reviewer_id)
     return RESULTS_DIR / f"reviews_{safe_id}.csv"
 
 
@@ -73,10 +144,7 @@ def save_review(reviewer_id: str, pair_index: int, review: dict) -> None:
     existing[pair_index] = review
 
     with open(path, "w", newline="") as f:
-        writer = csv.DictWriter(f, fieldnames=FIELDNAMES)
-        writer.writeheader()
-        for idx in sorted(existing.keys()):
-            writer.writerow(existing[idx])
+        f.write(_reviews_to_csv(existing))
 
     append_audit_log(reviewer_id, pair_index, "update" if is_update else "submit")
 
@@ -121,19 +189,25 @@ def render_login() -> str | None:
         "the best of your clinical judgment."
     )
     st.markdown("---")
-    reviewer_id = st.text_input("Enter your reviewer ID to begin")
-    if reviewer_id.strip():
-        return reviewer_id.strip()
+    token = st.text_input("Enter your access token to begin", type="password")
+    if token.strip():
+        reviewer_id = _authenticate(token.strip())
+        if reviewer_id:
+            return reviewer_id
+        st.error("Invalid access token.")
     return None
 
 
-def render_image(path: str | Path, label: str) -> None:
+def render_image(path: Path, label: str) -> None:
     st.subheader(label)
-    p = Path(path) if not isinstance(path, Path) else path
-    if p.exists():
-        st.image(str(p), use_container_width=True)
+    if not path.exists():
+        st.warning(f"Image not found: {path.name}")
+        return
+    image_bytes = _strip_dicom_metadata(path)
+    if image_bytes:
+        st.image(image_bytes, use_container_width=True)
     else:
-        st.warning(f"Image not found: {p.name}")
+        st.warning(f"Could not load image: {path.name}")
 
 
 def main() -> None:
@@ -164,16 +238,13 @@ def main() -> None:
     st.sidebar.markdown(f"**Reviewer:** {reviewer_id}")
     st.sidebar.metric("Progress", f"{reviewed} / {total}")
     st.sidebar.progress(reviewed / total if total > 0 else 0)
-
-    if not USE_GITHUB:
-        path = get_results_path(reviewer_id)
-        if path.exists():
-            st.sidebar.download_button(
-                "Export Results CSV",
-                data=path.read_text(),
-                file_name=f"validation_{reviewer_id}.csv",
-                mime="text/csv",
-            )
+    st.sidebar.download_button(
+        "Export Results CSV",
+        data=_reviews_to_csv(reviews),
+        file_name=f"validation_{_safe_filename_id(reviewer_id)}.csv",
+        mime="text/csv",
+        disabled=reviewed == 0,
+    )
 
     if st.sidebar.button("Log out"):
         append_audit_log(reviewer_id, -1, "logout")
@@ -223,7 +294,7 @@ def main() -> None:
     existing = reviews.get(pair_index, {})
 
     plausible = st.radio(
-        "1. Is the counterfactual clinically plausible as a real spine X-ray?",
+        "1. Do both images appear clinically plausible as real spine X-rays?",
         ["Yes", "No"],
         index=0 if existing.get("clinically_plausible", "Yes") == "Yes" else 1,
         horizontal=True,
@@ -231,7 +302,7 @@ def main() -> None:
     )
 
     preserved = st.radio(
-        "2. Is the pathology preserved across both images?",
+        "2. Is the pathology consistent across both images?",
         ["Yes", "No", "Uncertain"],
         index=["Yes", "No", "Uncertain"].index(
             existing.get("pathology_preserved", "Yes")
@@ -266,7 +337,11 @@ def main() -> None:
             "comments": comments,
             "timestamp": time.strftime("%Y-%m-%d %H:%M:%S UTC", time.gmtime()),
         }
-        save_review(reviewer_id, pair_index, review)
+        try:
+            save_review(reviewer_id, pair_index, review)
+        except Exception as e:
+            st.error(f"Failed to save review: {e}. Please try again.")
+            return
 
         if pos < total - 1:
             st.session_state.position = pos + 1