Upload 3 files

Dockerfile

@@ -1,20 +1,27 @@
-FROM python:3.13.5-slim
-
-WORKDIR /app
-
-RUN apt-get update && apt-get install -y \
-    build-essential \
-    curl \
-    git \
-    && rm -rf /var/lib/apt/lists/*
-
-COPY requirements.txt ./
-COPY src/ ./src/
-
-RUN pip3 install -r requirements.txt
-
-EXPOSE 8501
-
-HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health
-
-ENTRYPOINT ["streamlit", "run", "src/streamlit_app.py", "--server.port=8501", "--server.address=0.0.0.0"]
+FROM python:3.10-slim
+
+# Set working directory
+WORKDIR /app
+
+# Create writable cache directories
+RUN mkdir -p /app/.cache/huggingface/hub /app/.cache/torch
+RUN chmod -R 777 /app/.cache
+
+# Set environment variables for caches
+ENV HF_HOME=/app/.cache/huggingface
+ENV TRANSFORMERS_CACHE=/app/.cache/huggingface/transformers
+ENV TORCH_HOME=/app/.cache/torch
+ENV XDG_CACHE_HOME=/app/.cache
+
+# Install dependencies
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy app files
+COPY . .
+
+# Expose Streamlit port
+EXPOSE 7860
+
+# Run Streamlit
+CMD ["streamlit", "run", "app4.py", "--server.port=7860", "--server.address=0.0.0.0"]

app4.py

@@ -0,0 +1,326 @@
+import os
+import streamlit as st
+import numpy as np
+import time
+from sentence_transformers import SentenceTransformer
+import datetime
+import feedparser
+from huggingface_hub import hf_hub_download
+import faiss, pickle
+import aiohttp
+import asyncio
+import sqlite3
+
+# -------------------------------
+# CONFIG & PRIVATE STORAGE PATHS
+# -------------------------------
+PRIVATE_DIR = ".internal_data"
+os.makedirs(PRIVATE_DIR, exist_ok=True)
+
+CACHE_DB_PATH = os.path.join(PRIVATE_DIR, "query_cache.db")
+ADMIN_DB_PATH = os.path.join(PRIVATE_DIR, "cache_admin.db")
+
+# -------------------------------
+# DATABASE INITIALIZATION
+# -------------------------------
+def init_cache_db(db_path=CACHE_DB_PATH):
+    conn = sqlite3.connect(db_path, check_same_thread=False)
+    c = conn.cursor()
+    c.execute("""
+        CREATE TABLE IF NOT EXISTS cache (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            query TEXT UNIQUE,
+            answer TEXT,
+            embedding BLOB,
+            frequency INTEGER DEFAULT 1
+        )
+    """)
+    conn.commit()
+    return conn
+
+def init_admin_db(db_path=ADMIN_DB_PATH):
+    conn = sqlite3.connect(db_path, check_same_thread=False)
+    c = conn.cursor()
+    c.execute("""
+        CREATE TABLE IF NOT EXISTS logs (
+            id INTEGER PRIMARY KEY AUTOINCREMENT,
+            ts TEXT,
+            query TEXT,
+            cached INTEGER,
+            faiss_score REAL
+        )
+    """)
+    conn.commit()
+    return conn
+
+cache_conn = init_cache_db()
+admin_conn = init_admin_db()
+
+# -------------------------------
+# ADMIN LOGGING (PRIVATE DB)
+# -------------------------------
+def admin_log(query, cached: bool, faiss_score: float | None):
+    c = admin_conn.cursor()
+    c.execute(
+        "INSERT INTO logs (ts, query, cached, faiss_score) VALUES (?, ?, ?, ?)",
+        (datetime.datetime.now().isoformat(), query, int(bool(cached)), None if faiss_score is None else float(faiss_score))
+    )
+    admin_conn.commit()
+
+# -------------------------------
+# SIMPLE CACHE HELPERS (EXACT MATCH)
+# -------------------------------
+def store_in_cache(query, answer, embedding: np.ndarray):
+    c = cache_conn.cursor()
+    c.execute("""
+        INSERT OR REPLACE INTO cache (query, answer, embedding, frequency)
+        VALUES (?, ?, ?, COALESCE((SELECT frequency FROM cache WHERE query=?), 0) + 1)
+    """, (query, answer, embedding.astype(np.float32).tobytes(), query))
+    cache_conn.commit()
+
+def search_cache_exact(query):
+    c = cache_conn.cursor()
+    c.execute("SELECT answer FROM cache WHERE query = ?", (query,))
+    row = c.fetchone()
+    return row[0] if row else None
+
+def get_top_cached_queries(limit=5):
+    c = cache_conn.cursor()
+    c.execute("SELECT query, frequency FROM cache ORDER BY frequency DESC LIMIT ?", (limit,))
+    return c.fetchall()
+
+# -------------------------------
+# Load FAISS index + embedder
+# -------------------------------
+@st.cache_resource
+def load_index_and_model():
+    faiss_path = hf_hub_download(
+        repo_id="krishnasimha/health-chatbot-data",
+        filename="health_index.faiss",
+        repo_type="dataset"
+    )
+    pkl_path = hf_hub_download(
+        repo_id="krishnasimha/health-chatbot-data",
+        filename="health_metadata.pkl",
+        repo_type="dataset"
+    )
+
+    index = faiss.read_index(faiss_path)
+    with open(pkl_path, "rb") as f:
+        metadata = pickle.load(f)
+
+    embed_model = SentenceTransformer("all-MiniLM-L6-v2")
+    return index, metadata, embed_model
+
+index, metadata, embed_model = load_index_and_model()
+
+# -------------------------------
+# FAISS benchmark (sidebar)
+# -------------------------------
+def benchmark_faiss(n_queries=40, k=3):
+    queries = ["What is diabetes?", "How to prevent malaria?", "Symptoms of dengue?"]
+    query_embs = embed_model.encode(queries, convert_to_numpy=True)
+    times = []
+    for _ in range(n_queries):
+        q = query_embs[np.random.randint(0, len(query_embs))].reshape(1, -1)
+        start = time.time()
+        index.search(q, k)
+        times.append(time.time() - start)
+    avg_time = np.mean(times) * 1000
+    st.sidebar.write(f"⚡ FAISS Speed: {avg_time:.2f} ms/query")
+
+# -------------------------------
+# RSS News fetcher (async)
+# -------------------------------
+RSS_URL = "https://news.google.com/rss/search?q=health+disease+awareness&hl=en-IN&gl=IN&ceid=IN:en"
+
+async def fetch_rss_url(url):
+    async with aiohttp.ClientSession() as session:
+        async with session.get(url) as resp:
+            return await resp.text()
+
+def fetch_news():
+    try:
+        xml = asyncio.run(fetch_rss_url(RSS_URL))
+        feed = feedparser.parse(xml)
+        return [{"title": e.title, "link": e.link, "published": getattr(e, "published", "")} for e in feed.entries[:5]]
+    except Exception:
+        return []
+
+def update_news_hourly():
+    now = datetime.datetime.now()
+    if "last_news_update" not in st.session_state or (now - st.session_state.last_news_update).seconds > 3600:
+        st.session_state.last_news_update = now
+        st.session_state.news_articles = fetch_news()
+
+# -------------------------------
+# Together API async call
+# -------------------------------
+async def async_together_chat(messages):
+    url = "https://api.together.xyz/v1/chat/completions"
+    headers = {
+        "Authorization": f"Bearer {os.environ.get('TOGETHER_API_KEY','')}",
+        "Content-Type": "application/json",
+    }
+    payload = {"model": "deepseek-ai/DeepSeek-V3", "messages": messages}
+
+    async with aiohttp.ClientSession() as session:
+        async with session.post(url, headers=headers, json=payload) as resp:
+            resp.raise_for_status()
+            data = await resp.json()
+            return data["choices"][0]["message"]["content"]
+
+# -------------------------------
+# Main retrieval flow (exact-match cache -> faiss -> API)
+# -------------------------------
+def retrieve_answer(query, k=3):
+    # 1) exact cache
+    cached = search_cache_exact(query)
+    if cached:
+        admin_log(query, True, None)
+        st.sidebar.success("⚡ From cache")
+        return cached, []
+
+    # 2) FAISS retrieval on KB
+    q_emb = embed_model.encode([query], convert_to_numpy=True)
+    D, I = index.search(q_emb, k)
+    retrieved = [metadata["texts"][i] for i in I[0]]
+    sources = [metadata["sources"][i] for i in I[0]]
+    context = "\n".join(retrieved)
+
+    user_message = {"role": "user", "content": f"Use the context to answer:\n\n{context}\n\nQuestion: {query}"}
+    # append to session chat
+    if "chats" not in st.session_state:
+        st.session_state.chats = {}
+    if st.session_state.current_chat not in st.session_state.chats:
+        st.session_state.chats[st.session_state.current_chat] = [{"role":"system","content":"You are a helpful public health bot."}]
+    st.session_state.chats[st.session_state.current_chat].append(user_message)
+
+    # 3) Call LLM
+    answer = asyncio.run(async_together_chat(st.session_state.chats[st.session_state.current_chat]))
+
+    # 4) store exact query -> answer mapping
+    store_in_cache(query, answer, q_emb[0])
+    # record admin log with FAISS distance score (D contains distances; lower means closer for many indexes)
+    # For readability we convert the first distance to float if present
+    faiss_score = float(D[0][0]) if (D is not None and len(D) and len(D[0])) else None
+    admin_log(query, False, faiss_score)
+
+    st.session_state.chats[st.session_state.current_chat].append({"role":"assistant","content":answer})
+    return answer, sources
+
+# -------------------------------
+# Background news updater (fire-and-forget)
+# -------------------------------
+async def news_updater_loop():
+    while True:
+        st.session_state.news_articles = fetch_news()
+        await asyncio.sleep(3600)
+
+if "news_task" not in st.session_state:
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    st.session_state.news_task = loop.create_task(news_updater_loop())
+
+# -------------------------------
+# Streamlit UI (main)
+# -------------------------------
+st.title("Health Awareness Chatbot")
+
+# Chat initialization (session)
+if "chats" not in st.session_state:
+    st.session_state.chats = {}
+if "current_chat" not in st.session_state:
+    st.session_state.current_chat = "New Chat 1"
+    st.session_state.chats["New Chat 1"] = [{"role":"system","content":"You are a helpful public health awareness chatbot."}]
+
+# Sidebar: Chat manager + benchmark + top cached queries (fast, read-only)
+st.sidebar.header("Chat Manager")
+if st.sidebar.button("➕ New Chat"):
+    name = f"New Chat {len(st.session_state.chats) + 1}"
+    st.session_state.chats[name] = [{"role":"system","content":"You are a helpful public health awareness chatbot."}]
+    st.session_state.current_chat = name
+
+benchmark_faiss()
+
+# Show top cached queries in the sidebar (helpful to the user to pick exact strings)
+st.sidebar.subheader("🔥 Most Asked Questions")
+top_qs = get_top_cached_queries(limit=5)
+if top_qs:
+    for q, freq in top_qs:
+        st.sidebar.write(f"**{q}** — used {freq} times")
+else:
+    st.sidebar.write("_No cached queries yet._")
+
+# Chat selector (unique key to avoid duplicate-element error)
+chat_list = list(st.session_state.chats.keys())
+selected_chat = st.sidebar.selectbox(
+    "Your chats:",
+    chat_list,
+    index=chat_list.index(st.session_state.current_chat),
+    key="chat_selector"
+)
+st.session_state.current_chat = selected_chat
+
+# Rename chat
+new_name = st.sidebar.text_input("Rename Chat:", st.session_state.current_chat, key="rename_chat")
+if new_name and new_name != st.session_state.current_chat:
+    if new_name not in st.session_state.chats:
+        st.session_state.chats[new_name] = st.session_state.chats.pop(st.session_state.current_chat)
+        st.session_state.current_chat = new_name
+
+# -------------------------------
+# News section & query input
+# -------------------------------
+update_news_hourly()
+st.subheader("📰 Latest Health News")
+if "news_articles" in st.session_state:
+    for art in st.session_state.news_articles:
+        st.markdown(f"**{art['title']}**  \n[Read more]({art['link']})  \n*{art['published']}*")
+        st.write("---")
+
+user_query = st.text_input("Ask me about health, prevention, or awareness:")
+
+if user_query:
+    with st.spinner("Searching knowledge base..."):
+        answer, sources = retrieve_answer(user_query)
+    st.write("### 💡 Answer")
+    st.write(answer)
+
+    st.write("### 📖 Sources")
+    for src in sources:
+        st.write(f"- {src}")
+
+# Render chat history
+for msg in st.session_state.chats[st.session_state.current_chat]:
+    if msg["role"] == "user":
+        st.write(f"🧑 **You:** {msg['content']}")
+    elif msg["role"] == "assistant":
+        st.write(f"🤖 **Bot:** {msg['content']}")
+
+# -------------------------------
+# ADMIN (hidden) - only visible if admin key provided
+# -------------------------------
+# Get admin key from secrets or environment
+ADMIN_KEY_SECRET = None
+try:
+    ADMIN_KEY_SECRET = st.secrets["ADMIN_KEY"]
+except Exception:
+    ADMIN_KEY_SECRET = os.environ.get("ADMIN_KEY", None)
+
+# Admin panel in sidebar (password protected). Only you should know the key.
+with st.sidebar.expander("🔐 Admin (hidden)"):
+    dev_key = st.text_input("Admin key (password):", type="password", key="admin_key_input")
+    if dev_key and ADMIN_KEY_SECRET and dev_key == ADMIN_KEY_SECRET:
+        st.success("Admin access granted — private logs shown below.")
+        # Show private logs from admin DB (careful: this is only visible to whoever knows the key)
+        c = admin_conn.cursor()
+        rows = c.execute("SELECT id, ts, query, cached, faiss_score FROM logs ORDER BY id DESC LIMIT 200").fetchall()
+        st.write(f"Showing {len(rows)} recent log rows (private).")
+        for rid, ts, q, cached, score in rows:
+            st.text(f"[{rid}] {ts} | cached={bool(cached)} | faiss_score={score}")
+            st.markdown(f"- **Q:** {q}")
+            st.write("---")
+    elif dev_key:
+        st.error("Wrong admin key.")
+

requirements.txt

@@ -1,3 +1,12 @@
-altair
-pandas
-streamlit
+streamlit
+numpy
+sentence-transformers
+together
+aiohttp
+feedparser
+huggingface_hub
+faiss-cpu
+pickle5
+sqlite3-binary
+
+