Initial commit: E-commerce RAG Docker demo

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+ecom_chroma_db/chroma.sqlite3 filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -0,0 +1,8 @@
+FROM python:3.10-slim
+WORKDIR /app
+RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
+COPY requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -r /app/requirements.txt
+COPY . /app
+EXPOSE 7860
+CMD ["streamlit", "run", "app.py", "--server.port=7860", "--server.address=0.0.0.0"]

README.md

@@ -1,10 +1,19 @@
 ---
-title: Ecom Rag
-emoji: 🐠
+title: E-commerce RAG Demo
+emoji: 🛍️
 colorFrom: pink
-colorTo: green
+colorTo: purple
 sdk: docker
+app_file: app.py
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# E-commerce RAG Demo (Streamlit inside Docker)
+
+This Space was auto-created from Colab. It ships with:
+- Tiny multi-source dataset (descriptions/specs/reviews)
+- Embeddings via `all-MiniLM-L6-v2` into **ChromaDB** (bundled in repo)
+- Sentiment analysis model (bundled in repo)
+- Streamlit app for **recommendations** and **comparisons**
+
+## Run locally

app.py

@@ -0,0 +1,108 @@
+import os
+import json
+import chromadb
+from sentence_transformers import SentenceTransformer
+import streamlit as st
+from typing import List
+import google.generativeai as genai
+from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
+
+st.set_page_config(page_title='E-commerce RAG Demo', layout='wide')
+st.title('🛍️ E-commerce RAG Demo (Recommendations & Comparisons)')
+
+# Configure Gemini (optional)
+GEMINI_API_KEY = os.environ.get('GEMINI_API_KEY', '')
+if GEMINI_API_KEY:
+    genai.configure(api_key=GEMINI_API_KEY)
+
+@st.cache_resource(show_spinner=False)
+def get_clients():
+    client = chromadb.PersistentClient(path='ecom_chroma_db')
+    collection = client.get_or_create_collection('products', metadata={"hnsw:space": "cosine"})
+    embedder = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
+    return client, collection, embedder
+
+@st.cache_resource(show_spinner=False)
+def get_sentiment_pipeline():
+    model_dir = 'sentiment_model' # Load from saved directory
+    tokenizer = AutoTokenizer.from_pretrained(model_dir)
+    model = AutoModelForSequenceClassification.from_pretrained(model_dir)
+    return pipeline('sentiment-analysis', model=model, tokenizer=tokenizer)
+
+_, collection, embedder = get_clients()
+sa_pipeline = get_sentiment_pipeline()
+
+
+def retrieve(query: str, k: int = 5):
+    qemb = embedder.encode([query]).tolist()
+    out = collection.query(query_embeddings=qemb, n_results=k, include=['documents', 'metadatas', 'distances'])
+    items = []
+    for doc, meta, dist in zip(out['documents'][0], out['metadatas'][0], out['distances'][0]):
+        items.append({'doc': doc, 'meta': meta, 'score': 1 - dist})
+    return items
+
+def llm_generate(prompt: str) -> str:
+    if GEMINI_API_KEY:
+        model = genai.GenerativeModel('gemini-1.5-flash')
+        resp = model.generate_content(prompt)
+        return resp.text
+    # Fallback if no key: return prompt tail as simple echo
+    return 'LLM disabled. Showing retrieved context only.\n\n' + prompt[-1500:]
+
+st.sidebar.header('Preferences')
+prefs_cat = st.sidebar.multiselect('Preferred categories', ['Audio', 'Wearables', 'Computers'])
+price_min, price_max = st.sidebar.slider('Price range', 0, 50000, (0, 50000), step=500)
+
+mode = st.radio('Mode', ['Recommend Products', 'Compare Products'])
+query = st.text_input('Describe what you need (e.g., "lightweight earbuds for calls and gym")')
+topk = st.slider('Top K', 1, 10, 5)
+
+if st.button('Run'):
+    if not query:
+        st.warning('Enter a query first.')
+    else:
+        results = retrieve(query, k=topk)
+        # Simple personalization: filter by category and price range
+        filtered = []
+        for r in results:
+            cat_ok = (not prefs_cat) or (r['meta']['category'] in prefs_cat)
+            price_ok = (price_min <= r['meta']['price'] <= price_max)
+            if cat_ok and price_ok:
+                filtered.append(r)
+        if not filtered:
+            filtered = results
+
+        if mode == 'Recommend Products':
+            ctx = '\n\n'.join([f"[Score={round(x['score'],3)}] {x['doc']}" for x in filtered])
+            prompt = f"""
+You are an assistant that recommends e-commerce products. Based on the retrieved context below, recommend 3 products and explain why each fits the user's query. Summarize pros/cons succinctly. If information is missing, say so.
+USER QUERY: {query}
+CONTEXT:\n{ctx}
+"""
+            answer = llm_generate(prompt)
+            st.markdown(answer)
+            st.subheader('Retrieved Items')
+            for r in filtered:
+                st.write(r['meta'])
+                with st.expander('Context'):
+                    st.write(r['doc'])
+        else:
+            # Compare top 2-4
+            comps = filtered[:4]
+            if not comps:
+                st.info('No items to compare.')
+            else:
+                cols = st.columns(len(comps))
+                for c, r in zip(cols, comps):
+                    with c:
+                        st.metric(r['meta']['title'], f"₹{int(r['meta']['price'])}")
+                        st.caption(f"Category: {r['meta']['category']} | Score: {r['score']:.3f}")
+                        with st.expander('Details'):
+                            st.write(r['doc'])
+                ctx = '\n\n'.join([r['doc'] for r in comps])
+                prompt = f"""
+Create a concise comparison table (Markdown) for the products in the context. Rows: Price, Category, Best for, Not ideal for, Key specs. Then a 3-bullet summary of trade-offs.
+USER QUERY: {query}
+CONTEXT:\n{ctx}
+"""
+                st.markdown(llm_generate(prompt))

ecom_chroma_db/0adce968-463d-42bc-be27-acfba3a21a21/data_level0.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8146ecc3e4c3a36ea9b3edc3778630c452f483990ec942d38e8006f4661e430
+size 16760000

ecom_chroma_db/0adce968-463d-42bc-be27-acfba3a21a21/header.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18f1e924efbb5e1af5201e3fbab86a97f5c195c311abe651eeec525884e5e449
+size 100

ecom_chroma_db/0adce968-463d-42bc-be27-acfba3a21a21/length.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7e2dcff542de95352682dc186432e98f0188084896773f1973276b0577d5305
+size 40000

ecom_chroma_db/chroma.sqlite3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8da19cc93a21395e2af3990299ac6bf726485aabe0c44cc0fb22b1e2362e5220
+size 245760

requirements.txt

@@ -0,0 +1,8 @@
+
+sentence-transformers
+chromadb==0.5.3
+transformers
+streamlit
+google-generativeai
+tiktoken
+rapidfuzz