Update app.py

app.py

@@ -4,6 +4,7 @@ import gradio as gr
 import faiss
 import numpy as np
 from pypdf import PdfReader
+from docx import Document
 from sentence_transformers import SentenceTransformer
 from transformers import pipeline
 
@@ -18,14 +19,69 @@ embedder = SentenceTransformer(EMBED_MODEL_NAME)
 generator = pipeline("text2text-generation", model=GEN_MODEL_NAME)
 
 # ---- PDF to text ----
-def pdfs_to_texts(files):
+def read_pdf_from_path_or_bytes(file_obj_or_path):
+
+    path = getattr(file_obj_or_path, "path", None)
+    if isinstance(file_obj_or_path, str) and os.path.exists(file_obj_or_path):
+        path = file_obj_or_path
+    if path and os.path.exists(path):
+        reader = PdfReader(path)
+        return "\n".join((p.extract_text() or "") for p in reader.pages)
+
+    data = None
+    if hasattr(file_obj_or_path, "read"):
+        data = file_obj_or_path.read()
+    elif hasattr(file_obj_or_path, "bytes"):
+        data = file_obj_or_path.bytes
+    if data:
+        reader = PdfReader(io.BytesIO(data))
+        return "\n".join((p.extract_text() or "") for p in reader.pages)
+
+    return ""
+
+
+def read_docx_text(path):
+    doc = Document(path)
+    return "\n".join(p.text for p in doc.paragraphs)
+
+
+def load_files_to_texts(files):
+    """
+    Accepts mixed uploads (.pdf, .docx, .txt).
+    Returns a list[str] of raw texts (one per file).
+    """
     texts = []
-    for f in files:
-        # f is an object from Gradio that read bytes for pypdf
-        reader = PdfReader(io.BytesIO(f.read()))
-        pages = [page.extract_text() or "" for page in reader.pages]
-        texts.append("\n".join(pages))
+    for f in files or []:
+        path = getattr(f, "path", None) or getattr(f, "name", None)
+        name = (path or str(f)).lower()
+
+        if name.endswith(".pdf"):
+            texts.append(read_pdf_from_path_or_bytes(f if path is None else path))
+
+        elif name.endswith(".docx"):
+            if path:
+                texts.append(read_docx_text(path))
+            else:
+                # Need a real path for python-docx
+                data = f.read() if hasattr(f, "read") else getattr(f, "bytes", b"")
+                import tempfile
+                with tempfile.NamedTemporaryFile(delete=False, suffix=".docx") as tf:
+                    tf.write(data)
+                    tmp_path = tf.name
+                texts.append(read_docx_text(tmp_path))
+                os.unlink(tmp_path)
+
+        elif name.endswith(".txt"):
+            if path and os.path.exists(path):
+                with open(path, "r", errors="ignore") as fh:
+                    texts.append(fh.read())
+            else:
+                data = f.read().decode("utf-8", errors="ignore") if hasattr(f, "read") else ""
+                texts.append(data)
+        else:
+            continue
     return texts
+
     
 
 # ---- Chunking ----
@@ -46,30 +102,32 @@ corpus_chunks = []
 
 def build_index(files, progress=gr.Progress()):
     global index, corpus_chunks
-    texts = pdfs_to_texts(files)
-    
-    # basic cleanup + chunk
-    corpus_chunks = []
-    for t in texts:
-        if not t.strip():
-            continue
-        corpus_chunks += chunk_text(t)
+    try:
+        texts = load_files_to_texts(files)
+        corpus_chunks = []
+        for t in texts:
+            if t and t.strip():
+                corpus_chunks += chunk_text(t)
+
+        if not corpus_chunks:
+            return "No text extracted from files.", 0
+
+        progress(0.3, desc="Embedding chunks…")
+        embeddings = embedder.encode(corpus_chunks, convert_to_numpy=True, show_progress_bar=False)
+        d = embeddings.shape[1]
 
-    if not corpus_chunks:
-        return "No text extracted from PDFs.", None
+        # Normalize for cosine sim with inner product
+        norms = np.linalg.norm(embeddings, axis=1, keepdims=True) + 1e-10
+        embeddings = embeddings / norms
 
-    progress(0.3, desc="Embedding chunks…")
-    embeddings = embedder.encode(corpus_chunks, convert_to_numpy=True, show_progress_bar=False)
-    d = embeddings.shape[1]
+        progress(0.6, desc="Creating FAISS index…")
+        index = faiss.IndexFlatIP(d)
+        index.add(embeddings.astype(np.float32))
 
-    progress(0.6, desc="Creating FAISS index…")
-    index = faiss.IndexFlatIP(d)  # cosine via inner product on normalized vectors
-    # normalize to unit length to approximate cosine similarity
-    norms = np.linalg.norm(embeddings, axis=1, keepdims=True) + 1e-10
-    embeddings = embeddings / norms
-    index.add(embeddings.astype(np.float32))
+        return f"Indexed {len(corpus_chunks)} chunks.", len(corpus_chunks)
+    except Exception as e:
+        return f"Build failed: {e}", 0
 
-    return f"Indexed {len(corpus_chunks)} chunks.", len(corpus_chunks)
 
 # ---- RAG query -> retrieve -> generate ----
 def answer_question(question, top_k=5, max_new_tokens=256):
@@ -99,7 +157,7 @@ with gr.Blocks(title="Group 5 Study Helper (RAG)") as demo:
     gr.Markdown("# Group 5 Study Helper (RAG)\nUpload PDFs → Build Index → Ask questions.")
 
     with gr.Row():
-        file_in = gr.Files(file_types=[".pdf"], label="Upload PDF files")
+        file_in = gr.Files(file_types=[".pdf", ".docx", ".txt"], label="Upload PDF/DOCX/TXT files")
     with gr.Row():
         build_btn = gr.Button("Build Index", variant="primary")
         status = gr.Markdown()