Spaces:

lvvignesh2122
/

Gemini-Rag-Fastapi-Pro

Running

App Files Files Community

lvvignesh2122 commited on 8 days ago

Commit

9d21791

1 Parent(s): 775a7d0

Add frontend UI and document upload for RAG app

Browse files

Files changed (4) hide show

frontend/index.html +257 -0
main.py +111 -48
rag_store.py +45 -34
requirements.txt +1 -0

frontend/index.html ADDED Viewed

	@@ -0,0 +1,257 @@

+<!DOCTYPE html>
+<html lang="en">
+<head>
+  <meta charset="UTF-8" />
+  <title>Gemini RAG Assistant</title>
+  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+  <!-- Fonts -->
+  <link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap" rel="stylesheet">
+  <style>
+    :root {
+      --bg: radial-gradient(1200px 600px at top, #e0e7ff 0%, #f8fafc 60%);
+      --card: rgba(255,255,255,0.85);
+      --border: rgba(15,23,42,0.08);
+      --primary: #4f46e5;
+      --secondary: #0ea5e9;
+      --text: #0f172a;
+      --muted: #64748b;
+      --error: #dc2626;
+    }
+    * { box-sizing: border-box; font-family: Inter, sans-serif; }
+    body {
+      margin: 0;
+      min-height: 100vh;
+      background: var(--bg);
+      display: flex;
+      justify-content: center;
+      padding: 40px 16px;
+      color: var(--text);
+    }
+    .container {
+      width: 100%;
+      max-width: 980px;
+      background: var(--card);
+      backdrop-filter: blur(16px);
+      border-radius: 24px;
+      padding: 36px;
+      border: 1px solid var(--border);
+      box-shadow: 0 40px 120px rgba(15,23,42,.15);
+    }
+    h1 {
+      font-size: 2.2rem;
+      margin: 0;
+      font-weight: 700;
+      background: linear-gradient(135deg, #4f46e5, #06b6d4);
+      -webkit-background-clip: text;
+      -webkit-text-fill-color: transparent;
+    }
+    .subtitle {
+      margin-top: 8px;
+      color: var(--muted);
+      font-size: 1rem;
+    }
+    .card {
+      margin-top: 28px;
+      background: white;
+      border-radius: 18px;
+      padding: 24px;
+      border: 1px solid var(--border);
+    }
+    .card h3 {
+      margin-top: 0;
+      margin-bottom: 16px;
+      font-size: 1.1rem;
+    }
+    input[type="file"], textarea {
+      width: 100%;
+      padding: 14px;
+      border-radius: 14px;
+      border: 1px solid var(--border);
+      font-size: 0.95rem;
+    }
+    textarea {
+      min-height: 120px;
+      resize: vertical;
+    }
+    .row {
+      display: flex;
+      gap: 12px;
+      margin-top: 12px;
+      flex-wrap: wrap;
+    }
+    button {
+      padding: 12px 18px;
+      border-radius: 14px;
+      border: none;
+      background: var(--primary);
+      color: white;
+      font-weight: 600;
+      cursor: pointer;
+      transition: all .2s ease;
+    }
+    button.secondary { background: var(--secondary); }
+    button:disabled {
+      opacity: .5;
+      cursor: not-allowed;
+    }
+    button:hover:not(:disabled) {
+      transform: translateY(-1px);
+      box-shadow: 0 10px 25px rgba(79,70,229,.35);
+    }
+    .status {
+      margin-top: 10px;
+      font-size: .9rem;
+      color: var(--muted);
+    }
+    .answer {
+      margin-top: 24px;
+      padding: 20px;
+      border-radius: 16px;
+      background: #f8fafc;
+      border: 1px solid var(--border);
+      white-space: pre-wrap;
+      line-height: 1.6;
+    }
+    .error {
+      color: var(--error);
+      margin-top: 10px;
+      font-weight: 500;
+    }
+    .loader {
+      font-weight: 600;
+      color: var(--primary);
+      animation: pulse 1.2s infinite;
+    }
+    @keyframes pulse {
+      0% { opacity: .4 }
+      50% { opacity: 1 }
+      100% { opacity: .4 }
+    }
+    footer {
+      text-align: center;
+      margin-top: 28px;
+      font-size: .8rem;
+      color: var(--muted);
+    }
+  </style>
+</head>
+<body>
+  <div class="container">
+    <h1>Gemini RAG Assistant</h1>
+    <div class="subtitle">
+      Upload documents · Ask questions · Get grounded answers
+    </div>
+    <!-- Upload -->
+    <div class="card">
+      <h3>📄 Upload documents</h3>
+      <input type="file" id="files" multiple />
+      <div class="row">
+        <button id="uploadBtn" onclick="upload()">Upload & Index</button>
+      </div>
+      <div id="uploadStatus" class="status"></div>
+    </div>
+    <!-- Ask -->
+    <div class="card">
+      <h3>💬 Ask or summarize</h3>
+      <textarea id="question" placeholder="Ask something about your documents…"></textarea>
+      <div class="row">
+        <button id="askBtn" onclick="ask()">Ask</button>
+        <button class="secondary" id="sumBtn" onclick="summarize()">Summarize</button>
+      </div>
+    </div>
+    <!-- Answer -->
+    <div id="answerBox" class="answer" style="display:none;"></div>
+    <div id="errorBox" class="error"></div>
+    <footer>
+      Built with FastAPI · FAISS · Gemini
+    </footer>
+  </div>
+<script>
+  let busy = false;
+  function setBusy(state) {
+    busy = state;
+    document.getElementById("askBtn").disabled = state;
+    document.getElementById("sumBtn").disabled = state;
+    document.getElementById("uploadBtn").disabled = state;
+  }
+  async function upload() {
+    const files = document.getElementById("files").files;
+    if (!files.length) return;
+    setBusy(true);
+    document.getElementById("uploadStatus").innerText = "Indexing documents…";
+    const fd = new FormData();
+    for (let f of files) fd.append("files", f);
+    const res = await fetch("/upload", { method: "POST", body: fd });
+    const data = await res.json();
+    document.getElementById("uploadStatus").innerText = data.message || "Done ✅";
+    setBusy(false);
+  }
+  async function ask() {
+    const q = document.getElementById("question").value.trim();
+    if (!q || busy) return;
+    setBusy(true);
+    document.getElementById("errorBox").innerText = "";
+    document.getElementById("answerBox").style.display = "block";
+    document.getElementById("answerBox").innerHTML = "<span class='loader'>Thinking…</span>";
+    try {
+      const res = await fetch("/ask", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ prompt: q })
+      });
+      const data = await res.json();
+      document.getElementById("answerBox").innerText = data.answer;
+    } catch {
+      document.getElementById("errorBox").innerText =
+        "⚠️ LLM quota exceeded. Please wait ~1 minute and retry.";
+    }
+    setBusy(false);
+  }
+  function summarize() {
+    document.getElementById("question").value =
+      "Summarize the uploaded documents in 5 bullet points.";
+    ask();
+  }
+</script>
+</body>
+</html>

main.py CHANGED Viewed

@@ -1,75 +1,138 @@
 import os
-from fastapi import FastAPI
 from pydantic import BaseModel
 from dotenv import load_dotenv
 import google.generativeai as genai
-from rag_store import search_knowledge
-load_dotenv()
 genai.configure(api_key=os.getenv("GEMINI_API_KEY"))
-app = FastAPI(title="AI RAG Backend with Gemini")
 class PromptRequest(BaseModel):
     prompt: str
-@app.get("/")
-def home():
-    return {"message": "AI backend is running 🚀"}
 @app.post("/ask")
 async def ask(data: PromptRequest):
-    results = search_knowledge(data.prompt)
     if not results:
-        return {
             "answer": "I don't know based on the provided documents.",
             "confidence": 0.0,
             "citations": []
         }
-    # -------- Context
-    context_text = "\n".join(r["text"] for r in results)
     prompt = f"""
-Answer the question strictly using the context.
-If unsure, say "I don't know".
 Question:
 {data.prompt}
-Context:
-{context_text}
 """
-    model = genai.GenerativeModel("gemini-2.5-flash")
-    response = model.generate_content(prompt)
-    # -------- Confidence scoring
-    avg_distance = sum(r["distance"] for r in results) / len(results)
-    if avg_distance < 0.6:
-        confidence = 0.9
-    elif avg_distance < 1.2:
-        confidence = 0.7
-    else:
-        confidence = 0.4
-    # -------- Citations
-    citations = []
-    seen = set()
-    for r in results:
-        key = (r["metadata"]["source"], r["metadata"]["page"])
-        if key not in seen:
-            seen.add(key)
-            citations.append({
-                "source": r["metadata"]["source"],
-                "page": r["metadata"]["page"]
-            })
-    return {
-        "answer": response.text,
-        "confidence": round(confidence, 2),
-        "citations": citations
-    }

 import os
+from time import time
+from fastapi import FastAPI, UploadFile, File
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import HTMLResponse, JSONResponse
+from fastapi.staticfiles import StaticFiles
 from pydantic import BaseModel
 from dotenv import load_dotenv
 import google.generativeai as genai
+from rag_store import ingest_documents, search_knowledge
+# -----------------------
+# Setup
+# -----------------------
+load_dotenv()
 genai.configure(api_key=os.getenv("GEMINI_API_KEY"))
+app = FastAPI(
+    title="Gemini RAG FastAPI",
+    docs_url="/docs",
+    redoc_url="/redoc"
+)
+# -----------------------
+# CORS
+# -----------------------
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+# -----------------------
+# Frontend
+# -----------------------
+app.mount("/frontend", StaticFiles(directory="frontend"), name="frontend")
+# -----------------------
+# Cache (protect quota)
+# -----------------------
+CACHE_TTL = 300  # seconds
+answer_cache = {}
+# -----------------------
+# Models
+# -----------------------
 class PromptRequest(BaseModel):
     prompt: str
+# -----------------------
+# Routes
+# -----------------------
+@app.get("/", response_class=HTMLResponse)
+def serve_ui():
+    with open("frontend/index.html", "r", encoding="utf-8") as f:
+        return f.read()
+# -----------------------
+# Upload
+# -----------------------
+@app.post("/upload")
+async def upload(files: list[UploadFile] = File(...)):
+    try:
+        chunks = ingest_documents(files)
+        return {"message": f"Indexed {chunks} chunks from {len(files)} file(s)."}
+    except Exception as e:
+        return JSONResponse(status_code=400, content={"error": str(e)})
+# -----------------------
+# Ask
+# -----------------------
 @app.post("/ask")
 async def ask(data: PromptRequest):
+    prompt_key = data.prompt.strip().lower()
+    now = time()
+    # 🔁 Cache
+    if prompt_key in answer_cache:
+        ts, cached = answer_cache[prompt_key]
+        if now - ts < CACHE_TTL:
+            return cached
+    results = search_knowledge(data.prompt)
     if not results:
+        response = {
             "answer": "I don't know based on the provided documents.",
             "confidence": 0.0,
             "citations": []
         }
+        answer_cache[prompt_key] = (now, response)
+        return response
+    context = "\n\n".join(r["text"] for r in results)
     prompt = f"""
+Answer strictly using the context below.
+If not found, say "I don't know".
+Context:
+{context}
 Question:
 {data.prompt}
 """
+    try:
+        model = genai.GenerativeModel("gemini-2.5-flash")
+        llm_response = model.generate_content(prompt)
+        response = {
+            "answer": llm_response.text,
+            "confidence": round(min(1.0, len(results) / 5), 2),
+            "citations": [
+                {"source": r["metadata"]["source"], "page": r["metadata"]["page"]}
+                for r in results
+            ]
+        }
+        answer_cache[prompt_key] = (now, response)
+        return response
+    except Exception as e:
+        return JSONResponse(
+            status_code=429,
+            content={"error": "LLM quota exceeded. Please wait and retry."}
+        )
+# -----------------------
+# Summarize
+# -----------------------
+@app.post("/summarize")
+async def summarize():
+    return await ask(PromptRequest(
+        prompt="Summarize the uploaded documents in 5 concise bullet points."
+    ))

rag_store.py CHANGED Viewed

@@ -1,67 +1,78 @@
 import os
 import faiss
 import numpy as np
-from sentence_transformers import SentenceTransformer
 from pypdf import PdfReader
-DATA_DIR = "data"
-INDEX_FILE = "vector.index"
-DOCS_FILE = "documents.npy"
-META_FILE = "metadata.npy"
-model = SentenceTransformer("all-MiniLM-L6-v2")
-# -------------------------
-# Load or build index
-# -------------------------
-if os.path.exists(INDEX_FILE):
-    print("🔁 Loading FAISS index from disk...")
-    index = faiss.read_index(INDEX_FILE)
-    documents = np.load(DOCS_FILE, allow_pickle=True)
-    metadata = np.load(META_FILE, allow_pickle=True)
-else:
-    print("🧠 Building FAISS index...")
     texts = []
     meta = []
-    for file in os.listdir(DATA_DIR):
-        if file.endswith(".pdf"):
-            reader = PdfReader(os.path.join(DATA_DIR, file))
             for i, page in enumerate(reader.pages):
                 text = page.extract_text()
                 if text:
                     texts.append(text)
                     meta.append({
-                        "source": file,
                         "page": i + 1
                     })
-    embeddings = model.encode(texts)
     index = faiss.IndexFlatL2(embeddings.shape[1])
     index.add(np.array(embeddings))
-    np.save(DOCS_FILE, texts)
-    np.save(META_FILE, meta)
-    faiss.write_index(index, INDEX_FILE)
     documents = texts
     metadata = meta
-    print("✅ FAISS index saved to disk.")
-# -------------------------
 # Search
-# -------------------------
 def search_knowledge(query, top_k=5):
-    query_vec = model.encode([query])
     distances, indices = index.search(query_vec, top_k)
     results = []
-    for dist, idx in zip(distances[0], indices[0]):
         results.append({
             "text": documents[idx],
-            "metadata": metadata[idx],
-            "distance": float(dist)
         })
     return results

 import os
 import faiss
 import numpy as np
 from pypdf import PdfReader
+from sentence_transformers import SentenceTransformer
+# -----------------------
+# Global in-memory state
+# -----------------------
+index = None
+documents = []
+metadata = []
+embedder = SentenceTransformer("all-MiniLM-L6-v2")
+# -----------------------
+# Ingest uploaded files
+# -----------------------
+def ingest_documents(files):
+    global index, documents, metadata
     texts = []
     meta = []
+    for file in files:
+        filename = file.filename
+        if filename.endswith(".pdf"):
+            reader = PdfReader(file.file)
             for i, page in enumerate(reader.pages):
                 text = page.extract_text()
                 if text:
                     texts.append(text)
                     meta.append({
+                        "source": filename,
                         "page": i + 1
                     })
+        elif filename.endswith(".txt"):
+            content = file.file.read().decode("utf-8")
+            texts.append(content)
+            meta.append({
+                "source": filename,
+                "page": "N/A"
+            })
+    if not texts:
+        raise ValueError("No readable text found.")
+    embeddings = embedder.encode(texts)
     index = faiss.IndexFlatL2(embeddings.shape[1])
     index.add(np.array(embeddings))
     documents = texts
     metadata = meta
+    return len(texts)
+# -----------------------
 # Search
+# -----------------------
 def search_knowledge(query, top_k=5):
+    if index is None:
+        return []
+    query_vec = embedder.encode([query])
     distances, indices = index.search(query_vec, top_k)
     results = []
+    for idx, dist in zip(indices[0], distances[0]):
         results.append({
             "text": documents[idx],
+            "distance": float(dist),
+            "metadata": metadata[idx]
         })
     return results

requirements.txt CHANGED Viewed

@@ -6,3 +6,4 @@ faiss-cpu
 sentence-transformers
 pypdf
 numpy

 sentence-transformers
 pypdf
 numpy
+python-multipart