Update app.py

app.py

@@ -1,43 +1,48 @@
 # ============================================
-# Civil Engineering RAG (ASTM) - Uploadable Version
+# Civil Engineering RAG (ASTM) - Hugging Face Version
 # ============================================
 import os
-import fitz                      # PyMuPDF
+import fitz  # PyMuPDF
 import faiss
 import numpy as np
 import gradio as gr
+import tempfile
 from typing import List
 from groq import Groq
 from sentence_transformers import SentenceTransformer
-import tempfile
 
 # --------------------------
-# Config
+# 🔑 API Key
 # --------------------------
-GROQ_API_KEY = os.environ.get("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", "")
+GROQ_API_KEY = os.environ.get("GROQ_API_KEY", "")
 if not GROQ_API_KEY:
-    raise RuntimeError("❌ Missing GROQ_API_KEY. Add it in Hugging Face → Settings → Secrets.")
+    raise RuntimeError("❌ Missing GROQ_API_KEY. Please add it in Hugging Face → Settings → Secrets.")
 
+# Initialize Groq client and embedding model
 client = Groq(api_key=GROQ_API_KEY)
-embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+embedder = SentenceTransformer("all-MiniLM-L6-v2")
 
 INDEX, CORPUS = None, []
 
 # --------------------------
-# PDF text extraction
+# 📄 Safe PDF Text Extraction
 # --------------------------
 def extract_text_from_pdf(file_path: str) -> str:
-    text = []
-    with fitz.open(file_path) as doc:
-        for page in doc:
-            text.append(page.get_text("text"))
-    return "\n".join(text)
+    try:
+        text = ""
+        with fitz.open(file_path) as doc:
+            for page in doc:
+                text += page.get_text("text")
+        return text
+    except Exception as e:
+        return f"Error extracting text from {file_path}: {e}"
 
 # --------------------------
-# Chunking helper
+# ✂️ Chunking Function
 # --------------------------
 def chunk_text(text: str, chunk_size: int = 800, overlap: int = 120) -> List[str]:
-    chunks, start, n = [], 0, len(text)
+    chunks = []
+    start, n = 0, len(text)
     while start < n:
         end = min(start + chunk_size, n)
         chunk = text[start:end].strip()
@@ -49,44 +54,72 @@ def chunk_text(text: str, chunk_size: int = 800, overlap: int = 120) -> List[str
     return chunks
 
 # --------------------------
-# Build FAISS index from uploaded PDFs
+# 🔢 Build FAISS Index
 # --------------------------
 def build_faiss_index(paths: List[str]):
     texts, vectors = [], []
     for p in paths:
-        raw = extract_text_from_pdf(p)
-        chunks = chunk_text(raw)
+        text = extract_text_from_pdf(p)
+        if text.startswith("Error extracting text"):
+            raise RuntimeError(text)
+        chunks = chunk_text(text)
+        if not chunks:
+            continue
         embs = embedder.encode(chunks, convert_to_numpy=True, show_progress_bar=False)
         texts.extend(chunks)
         vectors.append(embs.astype("float32"))
+
+    if not texts:
+        raise RuntimeError("❌ No valid text extracted from PDFs.")
+
     vectors = np.vstack(vectors).astype("float32")
     index = faiss.IndexFlatL2(vectors.shape[1])
     index.add(vectors)
     return index, texts
 
+# --------------------------
+# 📤 Rebuild Index from Upload
+# --------------------------
 def rebuild_index_from_upload(files):
     if not files:
         return "⚠️ Please upload at least one PDF."
+
     paths = []
     for f in files:
-        temp_path = os.path.join(tempfile.gettempdir(), f.name)
-        f.save(temp_path)
-        paths.append(temp_path)
-    global INDEX, CORPUS
-    INDEX, CORPUS = build_faiss_index(paths)
-    return f"✅ Indexed {len(paths)} file(s). You can now ask questions!"
+        try:
+            # Gradio provides a temp file path automatically (f.name)
+            if hasattr(f, "name") and os.path.exists(f.name):
+                temp_path = f.name
+            else:
+                # fallback in rare case
+                with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
+                    tmp.write(f.read())
+                    temp_path = tmp.name
+            paths.append(temp_path)
+        except Exception as e:
+            return f"❌ Error while saving uploaded file: {e}"
+
+    try:
+        global INDEX, CORPUS
+        INDEX, CORPUS = build_faiss_index(paths)
+        return f"✅ Successfully indexed {len(paths)} PDF(s). You can now ask questions!"
+    except Exception as e:
+        return f"❌ Error while building index: {e}"
 
 # --------------------------
-# Retrieval + Groq LLM
+# 🔍 Retrieve Context
 # --------------------------
 def retrieve_context(query: str, top_k: int = 4) -> str:
     if INDEX is None:
-        return "⚠️ Please upload PDFs first."
+        return "⚠️ Please upload and index PDFs first."
     q_emb = embedder.encode([query], convert_to_numpy=True).astype("float32")
     distances, indices = INDEX.search(q_emb, top_k)
     selected = [CORPUS[i] for i in indices[0] if 0 <= i < len(CORPUS)]
     return "\n\n---\n\n".join(selected)
 
+# --------------------------
+# 🧠 Query via Groq LLM
+# --------------------------
 SYSTEM_PROMPT = (
     "You are a helpful Civil Engineering assistant. "
     "Use ONLY the provided ASTM or uploaded document context to answer. "
@@ -96,7 +129,11 @@ SYSTEM_PROMPT = (
 def ask_groq(query: str, top_k: int = 4, model: str = "llama-3.3-70b-versatile") -> str:
     if INDEX is None:
         return "⚠️ Please upload PDFs first."
+
     context = retrieve_context(query, top_k)
+    if not context.strip():
+        return "⚠️ No relevant information found in the uploaded PDFs."
+
     prompt = f"""{SYSTEM_PROMPT}
 
 Context:
@@ -105,15 +142,19 @@ Context:
 Question:
 {query}
 """
-    completion = client.chat.completions.create(
-        model=model,
-        messages=[{"role": "user", "content": prompt}],
-        temperature=0.2,
-    )
-    return completion.choices[0].message.content
+
+    try:
+        completion = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": prompt}],
+            temperature=0.2,
+        )
+        return completion.choices[0].message.content
+    except Exception as e:
+        return f"❌ LLM Error: {e}"
 
 # --------------------------
-# Gradio UI
+# 🎨 Gradio UI
 # --------------------------
 def ui_ask(query: str, top_k: int):
     try:
@@ -123,13 +164,14 @@ def ui_ask(query: str, top_k: int):
 
 with gr.Blocks(title="Civil Engineering RAG (ASTM)") as demo:
     gr.Markdown("## 🏗️ Civil Engineering RAG\nUpload ASTM or civil-engineering PDFs, build an index, and ask questions.")
+    
     with gr.Row():
-        uploader = gr.File(label="Upload PDFs", file_count="multiple", file_types=[".pdf"])
+        uploader = gr.File(label="📄 Upload PDFs", file_count="multiple", file_types=[".pdf"])
         status = gr.Textbox(label="Status", interactive=False)
     uploader.upload(rebuild_index_from_upload, uploader, status)
 
     gr.Markdown("---")
-    inp = gr.Textbox(label="Your Question", placeholder="e.g., What is the standard curing time for concrete?")
+    inp = gr.Textbox(label="Your Question", placeholder="e.g., What is the curing time for concrete as per ASTM?")
     k = gr.Slider(1, 10, value=4, step=1, label="Top-K passages")
     out = gr.Textbox(label="Answer")
     btn = gr.Button("Ask")