Upload 4 files

Dockerfile

@@ -0,0 +1,28 @@
+# Use official Python image
+FROM python:3.10-slim
+
+# Set environment variables
+ENV PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PIP_NO_CACHE_DIR=1
+
+# Set work directory
+WORKDIR /app
+
+# Copy all files into the container
+COPY . /app
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install Python dependencies
+RUN pip install --upgrade pip
+RUN pip install -r requirements.txt
+
+# Expose port for Hugging Face
+EXPOSE 7860
+
+# Start the Flask app
+CMD ["python", "app.py"]

app.py

@@ -0,0 +1,199 @@
+from flask import Flask, request, render_template, send_file, redirect, url_for
+import os
+import re
+import uuid
+import numpy as np
+import faiss
+from sentence_transformers import SentenceTransformer
+from transformers import pipeline
+from PyPDF2 import PdfReader
+
+print("✅ App starting...")
+print("⏳ Loading SentenceTransformer model...")
+model = SentenceTransformer('all-MiniLM-L6-v2')
+print("✅ Model loaded.")
+
+print("⏳ Loading NLI pipeline...")
+nli = pipeline("text-classification", model="microsoft/deberta-large-mnli")
+print("✅ NLI pipeline loaded.")
+
+app = Flask(__name__)
+
+# ── base folders ───────────────────────────────────────────────────────────────
+BASE_UPLOADS = os.path.join(os.path.dirname(__file__), "uploads")
+BASE_RESULTS = os.path.join(os.path.dirname(__file__), "results")
+os.makedirs(BASE_UPLOADS, exist_ok=True)
+os.makedirs(BASE_RESULTS, exist_ok=True)
+
+# ── clear uploads at launch ────────────────────────────────────────────────────
+def clear_uploads_folder():
+    """Remove all files and subfolders inside the uploads folder on app launch."""
+    for entry in os.listdir(BASE_UPLOADS):
+        path = os.path.join(BASE_UPLOADS, entry)
+        if os.path.isdir(path):
+            for root, dirs, files in os.walk(path, topdown=False):
+                for fname in files:
+                    os.remove(os.path.join(root, fname))
+                for dname in dirs:
+                    os.rmdir(os.path.join(root, dname))
+            os.rmdir(path)
+        else:
+            os.remove(path)
+clear_uploads_folder()
+print("✅ Uploads folder cleared.")
+
+# runtime cache keyed by search‑id → (paragraphs, embeddings, faiss‑index)
+index_data = {}
+
+# ── helpers ────────────────────────────────────────────────────────────────────
+def get_paths(sid: str):
+    """Return per‑search folders & files, creating them if needed."""
+    up_folder   = os.path.join(BASE_UPLOADS, sid)
+    res_folder  = os.path.join(BASE_RESULTS, sid)
+    os.makedirs(up_folder,  exist_ok=True)
+    os.makedirs(res_folder, exist_ok=True)
+    merged_file = os.path.join(res_folder, "merged.txt")
+    result_file = os.path.join(res_folder, "results.txt")
+    return up_folder, res_folder, merged_file, result_file
+
+def extract_text(file_path):
+    if file_path.endswith('.txt'):
+        with open(file_path, 'r', encoding='utf-8') as f:
+            return f.read()
+    elif file_path.endswith('.pdf'):
+        reader = PdfReader(file_path)
+        full_text = " ".join(page.extract_text() for page in reader.pages if page.extract_text())
+        full_text = re.sub(r'(?<=[.!?])\s{2,}', '\n\n', full_text)
+        full_text = re.sub(r'(?<=[a-z])\.\s+(?=[A-Z])', '.\n\n', full_text)
+        full_text = re.sub(r'(\n\s*){2,}', '\n\n', full_text)
+        return full_text
+    return ""
+
+def rebuild_merged_and_index(sid: str):
+    """Re‑embed everything for *this* search id."""
+    up_folder, _, merged_file, _ = get_paths(sid)
+
+    merged_text = ""
+    for filename in os.listdir(up_folder):
+        if filename.lower().endswith((".pdf", ".txt")):
+            merged_text += extract_text(os.path.join(up_folder, filename)) + "\n\n"
+
+    with open(merged_file, "w", encoding='utf-8') as f:
+        f.write(merged_text)
+
+    paras = re.split(r'\n\s*\n+', merged_text)
+    paras = [p.strip().replace('\n', ' ') for p in paras if len(p.strip().split()) > 4]
+    if not paras:
+        index_data[sid] = ([], None, None)
+        return
+
+    embed = model.encode(paras, batch_size=32, show_progress_bar=False)
+    embed = np.asarray(embed)
+    if embed.ndim == 1:
+        embed = embed[np.newaxis, :]
+    faiss.normalize_L2(embed)
+    idx = faiss.IndexFlatIP(embed.shape[1])
+    idx.add(embed)
+
+    index_data[sid] = (paras, embed, idx)
+
+# ── routes ─────────────────────────────────────────────────────────────────────
+@app.route("/", methods=["GET", "POST"])
+def index():
+    # Each *page load* gets its own UUID, preserved via hidden form fields/URLs
+    sid = request.args.get("sid") or request.form.get("sid")
+    if not sid:
+        sid = str(uuid.uuid4())
+
+    up_folder, _, _, _ = get_paths(sid)           # ensure dirs exist
+    paragraphs, embeddings, index_faiss = index_data.get(sid, ([], None, None))
+
+    results = []
+    query = ""
+    k = 5
+
+    if request.method == "POST":
+        query = request.form.get("query", "").strip()
+        try:
+            k = int(request.form.get("topk", 5))
+        except ValueError:
+            k = 5
+
+        if paragraphs and query:
+            q_embed = model.encode([query])
+            q_embed = np.asarray(q_embed)
+            if q_embed.ndim == 1:
+                q_embed = q_embed[np.newaxis, :]
+            faiss.normalize_L2(q_embed)
+            D, I = index_faiss.search(q_embed, k=min(k, len(paragraphs)))
+            results = [paragraphs[i] for i in I[0]]
+
+            _, res_folder, _, result_file = get_paths(sid)
+            with open(result_file, "w", encoding='utf-8') as f:
+                for para in results:
+                    f.write(para + "\n\n")
+
+    return render_template("index.html", results=results, query=query, topk=k, sid=sid)
+
+@app.route("/upload", methods=["POST"])
+def upload_file():
+    sid = request.args.get("sid")
+    if not sid:
+        return ("Missing sid", 400)
+
+    up_folder, _, _, _ = get_paths(sid)
+    uploaded_files = request.files.getlist("file")
+    for file in uploaded_files:
+        if file and file.filename.lower().endswith((".pdf", ".txt")):
+            file.save(os.path.join(up_folder, file.filename))
+
+    rebuild_merged_and_index(sid)
+    return ("", 204)
+
+@app.route("/download")
+def download():
+    sid = request.args.get("sid")
+    if not sid:
+        return ("Missing sid", 400)
+
+    _, _, _, result_file = get_paths(sid)
+    if not os.path.exists(result_file):
+        return ("Nothing to download", 404)
+    return send_file(result_file, as_attachment=True)
+
+@app.route("/download_merged")
+def download_merged():
+    sid = request.args.get("sid")
+    if not sid:
+        return ("Missing sid", 400)
+
+    _, _, merged_file, _ = get_paths(sid)
+    if not os.path.exists(merged_file):
+        return ("Nothing to download", 404)
+    return send_file(merged_file, as_attachment=True)
+
+
+@app.route("/reset")
+def reset():
+    sid = request.args.get("sid")
+    if not sid:
+        return redirect(url_for('index'))
+
+    up_folder, res_folder, _, _ = get_paths(sid)
+    for folder in [up_folder, res_folder]:
+        if os.path.exists(folder):
+            for f in os.listdir(folder):
+                os.remove(os.path.join(folder, f))
+
+    index_data.pop(sid, None)  # drop cached embeddings
+    return redirect(url_for('index'))
+
+#if __name__ == "__main__":
+#    from waitress import serve
+#    # Use threads to approximate “workers” on Windows (Waitress is single‑process).
+#    serve(app, host="0.0.0.0", port=9001, threads=4)
+
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=7860)
+

requirements.txt

@@ -0,0 +1,19 @@
+# Core UI
+gradio>=4.37.0
+flask
+# Embeddings / transformers
+sentence-transformers==2.7.0
+transformers>=4.40.0
+
+# Torch backend (CPU is fine on HF Spaces free tier)
+torch>=2.2.0
+
+# PDF + utils
+PyPDF2==3.0.1
+numpy<2
+
+# Vector search
+faiss-cpu==1.8.0
+
+accelerate>=0.28.0
+scikit-learn>=1.3.0

templates/index.html

@@ -0,0 +1,103 @@
+<!doctype html>
+<html>
+<head>
+  <title>Retriever</title>
+  <style>
+    body { font-family:sans-serif; max-width:800px; margin:40px auto; }
+    progress { width:300px; height:20px; vertical-align:middle; }
+    #uploadStatus { margin-left:10px; font-weight:bold; }
+    #uploadProgress::-webkit-progress-value { background:#4caf50; }
+    #uploadProgress::-webkit-progress-bar  { background:#eee; }
+  </style>
+</head>
+<body>
+  <h2>Upload Text or PDF Files</h2>
+
+  <input type="file" id="fileUpload" multiple><br><br>
+  <progress id="uploadProgress" value="0" max="100" style="display:none;"></progress>
+  <span id="uploadStatus"></span>
+
+  <script>
+    // Embed search‑id from the server
+    const SID = "{{ sid }}";
+
+    document.getElementById("fileUpload").addEventListener("change", function () {
+      const files = this.files;
+      if (files.length === 0) return;
+
+      const formData = new FormData();
+      for (const file of files) {
+        formData.append("file", file);
+      }
+
+      const xhr = new XMLHttpRequest();
+      const progressBar = document.getElementById("uploadProgress");
+      const statusText  = document.getElementById("uploadStatus");
+
+      // ⬇️ send sid with the upload
+      xhr.open("POST", `/upload?sid=${SID}`, true);
+
+      xhr.upload.onprogress = function (e) {
+        if (e.lengthComputable) {
+          const percent = Math.round((e.loaded / e.total) * 100);
+          progressBar.value = percent;
+          progressBar.style.display = "inline-block";
+          statusText.textContent = `Uploading: ${percent}%`;
+        }
+      };
+
+      xhr.onload  = function () {
+        progressBar.value = 100;
+        statusText.textContent = "✅ Upload complete!";
+        setTimeout(() => { progressBar.style.display="none"; statusText.textContent=""; }, 1500);
+      };
+      xhr.onerror = function () { statusText.textContent = "❌ Upload failed."; };
+      xhr.send(formData);
+    });
+
+    document.querySelector('form[method="post"]').addEventListener("submit", function () {
+      const progressBar = document.getElementById("searchProgress");
+      const statusText  = document.getElementById("searchStatus");
+      progressBar.removeAttribute("value"); // indeterminate mode
+      progressBar.style.display = "inline-block";
+      statusText.textContent = "🔍 Processing search...";
+    });
+  </script>
+
+  <hr>
+
+  <!-- ⬇️ the hidden sid travels with every form submit -->
+  <form method="post">
+    <input type="hidden" name="sid" value="{{ sid }}">
+    <label>Enter your question or claim:</label><br>
+    <textarea name="query" rows="4" cols="80" required>{{ query }}</textarea><br><br>
+
+    <label>Number of paragraphs to return:</label>
+    <input type="number" name="topk" value="{{ topk }}" min="1" max="50"><br><br>
+
+    <button type="submit">Retrieve</button>
+    <progress id="searchProgress" max="100" style="width:300px; display:none;"></progress>
+    <span id="searchStatus" style="margin-left:10px;"></span>
+  </form>
+
+  <br>
+  <form method="get" action="/reset" onsubmit="return confirm('Clear all uploaded files and results?');">
+    <input type="hidden" name="sid" value="{{ sid }}">
+    <button type="submit" style="color:red;">Start New Search</button>
+  </form>
+
+  {% if results %}
+<br>
+<a href="{{ url_for('download', sid=sid) }}">Download these results</a> |
+<a href="{{ url_for('download_merged', sid=sid) }}">Download full merged text</a>
+<h3>Matching Paragraphs</h3>
+<ol>
+  {% for para in results %}
+    <li><p>{{ para }}</p></li>
+  {% endfor %}
+</ol>
+{% endif %}
+
+
+</body>
+</html>