Update app.py

app.py

@@ -16,13 +16,12 @@ import requests
 from pdf2image import convert_from_bytes
 from google.cloud import vision_v1 as vision
 
-# ───── Load .env locally (in HF Spaces secrets are injected automatically) ─────
+# ───── Load .env (locally) ────────────────────────────────────
+# In HF Spaces you'll set these as Secrets under Settings → Variables & secrets
 load_dotenv()
-
-# ───── Configure your API keys ─────────────────────────────────────────────────
-openai.api_key       = os.getenv("OPENAI_API_KEY")
-gcv_api_key          = os.getenv("GCV_API_KEY")
-proxycurl_api_key    = os.getenv("PROXYCURL_API_KEY")
+openai.api_key    = os.getenv("OPENAI_API_KEY")
+gcv_api_key       = os.getenv("GCV_API_KEY")
+proxycurl_api_key = os.getenv("PROXYCURL_API_KEY")
 
 if not openai.api_key:
     raise RuntimeError("Missing OPENAI_API_KEY")
@@ -31,9 +30,8 @@ if not gcv_api_key:
 if not proxycurl_api_key:
     raise RuntimeError("Missing PROXYCURL_API_KEY")
 
-# ───── FastAPI Setup ───────────────────────────────────────────────────────────
+# ───── FastAPI setup ──────────────────────────────────────────
 app = FastAPI(title="Aliro Data Extraction API")
-
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"],
@@ -41,17 +39,15 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# ───── Helper: Try to pull raw bytes out of whatever UploadFile gives you ───────
-def get_bytes(f: UploadFile) -> bytes:
-    data = f.file.read()
-    # rewind so Starlette can close cleanly
-    f.file.seek(0)
-    return data
+# ───── A simple root so GET / won’t 404 ───────────────────────
+@app.get("/")
+def read_root():
+    return {"message": "Aliro Data Extraction API – POST your files to /extract"}
 
-# ───── Extract text from PDF or image buffers ─────────────────────────────────
+# ───── PDF/OCR processing ─────────────────────────────────────
 def process_buffers(buffers: List[BytesIO]) -> List[str]:
     client = vision.ImageAnnotatorClient(client_options={"api_key": gcv_api_key})
-    results = []
+    results: List[str] = []
 
     for buf in buffers:
         name = getattr(buf, "name", "").lower()
@@ -59,47 +55,46 @@ def process_buffers(buffers: List[BytesIO]) -> List[str]:
 
         # PDF: try text → fallback to page-by-page OCR
         if name.endswith(".pdf"):
-            # 1) text via PyPDF2
+            # 1) PyPDF2 text
             try:
                 reader = PyPDF2.PdfReader(io.BytesIO(data))
-                txt = "".join(page.extract_text() or "" for page in reader.pages)
-                if txt.strip():
-                    results.append(txt)
+                text = "".join(page.extract_text() or "" for page in reader.pages)
+                if text.strip():
+                    results.append(text)
                     continue
             except:
                 pass
 
-            # 2) fallback: render each page → OCR
+            # 2) Fallback: render pages → OCR
             try:
-                images = convert_from_bytes(data)
-                for img in images:
+                for img in convert_from_bytes(data):
                     img_buf = io.BytesIO()
                     img.save(img_buf, format="JPEG")
                     resp = client.text_detection(image=vision.Image(content=img_buf.getvalue()))
-                    text = resp.text_annotations[0].description or ""
-                    if text:
-                        results.append(text)
+                    desc = resp.text_annotations[0].description if resp.text_annotations else ""
+                    if desc:
+                        results.append(desc)
                 continue
             except:
                 pass
 
-        # Image → OCR
+        # Image → always OCR
         if name.endswith((".png", ".jpg", ".jpeg")):
             try:
                 resp = client.text_detection(image=vision.Image(content=data))
-                text = resp.text_annotations[0].description or ""
-                if text:
-                    results.append(text)
+                desc = resp.text_annotations[0].description if resp.text_annotations else ""
+                if desc:
+                    results.append(desc)
                     continue
             except:
                 pass
 
-        # unsupported
+        # Otherwise unsupported
         results.append(f"[Unsupported file type: {name}]")
 
     return results
 
-# ───── Scrape LinkedIn via Proxycurl ────────────────────────────────────────────
+# ───── LinkedIn scrape via Proxycurl ─────────────────────────
 def scrape_linkedin(url: str) -> str:
     if not url:
         return ""
@@ -112,7 +107,7 @@ def scrape_linkedin(url: str) -> str:
     resp.raise_for_status()
     return json.dumps(resp.json(), indent=2)
 
-# ───── Summarize via OpenAI ────────────────────────────────────────────────────
+# ───── Summarize via OpenAI ─────────────────────────────────
 def make_summary(chunks: List[str]) -> str:
     if not chunks:
         return "No data extracted."
@@ -124,24 +119,24 @@ def make_summary(chunks: List[str]) -> str:
                 "role": "system",
                 "content": (
                     "Extract and structure personal, educational, and professional details "
-                    "into a clear, logical hierarchy. Translate non-English content, remove "
-                    "duplicates, and output in English."
+                    "into a clear, logical and relational hierarchy. Translate non-English "
+                    "content, remove duplicates, and output in English."
                 )
             },
-            {"role": "user", "content": f"Summarize the following data:\n\n{prompt}"}
+            {"role": "user", "content": f"Summarize the following data:\n\n{prompt}"},
         ],
-        max_tokens=1500
+        max_tokens=1500,
     )
     return resp.choices[0].message.content.strip()
 
-# ───── The /extract endpoint ───────────────────────────────────────────────────
+# ───── The POST /extract endpoint ─────────────────────────────
 @app.post("/extract")
 async def extract_endpoint(
     files: List[UploadFile] = File(default=[]),
-    linkedin_url: str            = Form(default="")
+    linkedin_url: str       = Form(default="")
 ):
     try:
-        # read & wrap each UploadFile in a BytesIO
+        # Wrap each UploadFile in a BytesIO (so we can peek at .name & .getvalue())
         buffers: List[BytesIO] = []
         for f in files:
             data = await f.read()
@@ -149,18 +144,17 @@ async def extract_endpoint(
             bio.name = f.filename
             buffers.append(bio)
 
-        # extract and summarize
+        # 1) OCR/PDF text
         texts = process_buffers(buffers)
-
-        # LinkedIn scrape, if provided
+        # 2) LinkedIn JSON
         if linkedin_url:
             texts.append(scrape_linkedin(linkedin_url))
-
+        # 3) Summarize
         summary = make_summary(texts)
         return {"summary": summary}
 
     except requests.HTTPError as e:
         raise HTTPException(status_code=502, detail=f"LinkedIn scrape failed: {e}")
-    except Exception as e:
+    except Exception:
         tb = traceback.format_exc()
         raise HTTPException(status_code=500, detail=f"Processing error:\n\n{tb}")