Update app.py

app.py

@@ -46,6 +46,46 @@ app.add_middleware(
 
 client = anthropic.Anthropic(api_key=os.environ.get("ANTHROPIC_API_KEY", ""))
 
+
+# ── PDF Text Extraction ───────────────────────────────────────────────────────
+
+def decode_content(raw: str) -> str:
+    """
+    If the client sent a PDF as __PDF_BASE64__<data>, decode and extract text.
+    Otherwise return the string unchanged.
+    """
+    PREFIX = "__PDF_BASE64__"
+    if not raw.startswith(PREFIX):
+        return raw
+
+    import base64
+    from pypdf import PdfReader
+
+    b64 = raw[len(PREFIX):]
+    try:
+        pdf_bytes = base64.b64decode(b64)
+    except Exception:
+        raise HTTPException(status_code=400, detail="Invalid base64 PDF data.")
+
+    try:
+        reader = PdfReader(io.BytesIO(pdf_bytes))
+        pages = []
+        for page in reader.pages:
+            text = page.extract_text()
+            if text:
+                pages.append(text.strip())
+        extracted = "\n\n".join(pages).strip()
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"Could not read PDF: {e}")
+
+    if not extracted:
+        raise HTTPException(
+            status_code=400,
+            detail="PDF appears to be scanned/image-based — no text found. Please paste the text manually.",
+        )
+    return extracted
+
+
 # ── Request / Response Models ─────────────────────────────────────────────────
 class ProblemInput(BaseModel):
     content: str
@@ -480,12 +520,15 @@ def analyze(body: ProblemInput):
     """Run full 5-agent pipeline. Returns structured JSON."""
     if not body.content.strip():
         raise HTTPException(status_code=400, detail="Content cannot be empty.")
-    if len(body.content) < 30:
+
+    content = decode_content(body.content)
+
+    if len(content) < 30:
         raise HTTPException(status_code=400, detail="Content too short for meaningful analysis.")
 
     try:
         result = run_pipeline(
-            content=body.content,
+            content=content,
             name=body.intern_name or "Intern",
             role=body.intern_role or "AI Developer Intern",
             goal=body.intern_goal or "",
@@ -505,6 +548,9 @@ def analyze_stream(body: ProblemInput):
     if not body.content.strip():
         raise HTTPException(status_code=400, detail="Content cannot be empty.")
 
+    # Decode PDF if needed before streaming starts
+    resolved_content = decode_content(body.content)
+
     def event_stream():
         agents = [
             ("analyst", AGENT_ANALYST, "Problem Analyst"),
@@ -515,7 +561,7 @@ def analyze_stream(body: ProblemInput):
         ]
 
         context = {
-            "content": body.content[:8000],
+            "content": resolved_content[:8000],
             "name": body.intern_name or "Intern",
             "role": body.intern_role or "AI Developer Intern",
             "goal": body.intern_goal or "",
@@ -528,7 +574,7 @@ def analyze_stream(body: ProblemInput):
 
             # Build context-aware prompt for this agent
             if key == "analyst":
-                user_msg = f"Intern: {context['name']} | Role: {context['role']} | Goal: {context['goal']}\n\nAnalyze:\n{context['content']}"
+                user_msg = f"Intern: {context['name']} | Role: {context['role']} | Goal: {context['goal']}\n\nAnalyze this content:\n{context['content']}"
             elif key == "root_cause":
                 user_msg = f"Problem:\n{accumulated.get('analyst','')}\n\nOriginal:\n{context['content'][:2000]}"
             elif key == "solutions":
@@ -571,9 +617,11 @@ def analyze_pdf(body: ProblemInput):
     if not body.content.strip():
         raise HTTPException(status_code=400, detail="Content cannot be empty.")
 
+    content = decode_content(body.content)
+
     try:
         analysis = run_pipeline(
-            content=body.content,
+            content=content,
             name=body.intern_name or "Intern",
             role=body.intern_role or "AI Developer Intern",
             goal=body.intern_goal or "",