Update app.py

app.py

@@ -8,61 +8,59 @@ from groq import Groq
 from tqdm.auto import tqdm
 import streamlit as st
 
+# Required imports
+import json
+import time
+import os
+from sentence_transformers import SentenceTransformer
+from pinecone import Pinecone, ServerlessSpec
+from groq import Groq
+from tqdm.auto import tqdm
+
 # Constants (hardcoded)
 FILE_PATH = "anjibot_chunks.json"
 BATCH_SIZE = 384
 INDEX_NAME = "groq-llama-3-rag"
-PINECONE_API_KEY = os.getenv("PINECONE_API_KEY") 
-GROQ_API_KEY = os.getenv("GROQ_API_KEY")
-DIMENSIONS = 768
-
-# Load data once at the start
-data = load_data(FILE_PATH)
-
-# Initialize Pinecone and SentenceTransformer once
-index = initialize_pinecone(PINECONE_API_KEY, INDEX_NAME, DIMENSIONS)
+PINECONE_API_KEY = os.getenv("PINECONE_API_KEY")  # Fixed syntax here
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")  # Fixed s
+DIMS = 768
 encoder = SentenceTransformer('dwzhu/e5-base-4k')
 
-def load_data(file_path: str) -> dict:
-    with open(file_path, 'r') as file:
-        return json.load(file)
+with open(FILE_PATH, 'r') as file:
+        data= json.load(file)
 
-def initialize_pinecone(api_key: str, index_name: str, dims: int) -> any:
-    pc = Pinecone(api_key=api_key)
-    spec = ServerlessSpec(cloud="aws", region='us-east-1')
+pc = Pinecone(api_key=PINECONE_API_KEY)
+spec = ServerlessSpec(cloud="aws", region='us-east-1')
+existing_indexes = [index_info["name"] for index_info in pc.list_indexes()]
+# Check if index already exists; if not, create it
+if INDEX_NAME not in existing_indexes:
+    pc.create_index(INDEX_NAME, dimension=DIMS, metric='cosine', spec=spec)
 
-    existing_indexes = [index_info["name"] for index_info in pc.list_indexes()]
+    # Wait for the index to be initialized
+    while not pc.describe_index(INDEX_NAME).status['ready']:
+        time.sleep(1)
 
-    # Check if index already exists; if not, create it
-    if index_name not in existing_indexes:
-        pc.create_index(index_name, dimension=dims, metric='cosine', spec=spec)
+index = pc.Index(INDEX_NAME)
 
-        # Wait for the index to be initialized
-        while not pc.describe_index(index_name).status['ready']:
-            time.sleep(1)
+for i in tqdm(range(0, len(data['id']), BATCH_SIZE)):
+    # Find end of batch
+    i_end = min(len(data['id']), i + BATCH_SIZE)
 
-    return pc.Index(index_name)
+    # Create batch
+    batch = {k: v[i:i_end] for k, v in data.items()}
 
-def upsert_data_to_pinecone(index: any, data: dict):
-    for i in tqdm(range(0, len(data['id']), BATCH_SIZE)):
-        # Find end of batch
-        i_end = min(len(data['id']), i + BATCH_SIZE)
-        
-        # Create batch
-        batch = {k: v[i:i_end] for k, v in data.items()}
-        
-        # Create embeddings
-        chunks = [f'{x["title"]}: {x["content"]}' for x in batch["metadata"]]
-        embeds = encoder.encode(chunks)
+    # Create embeddings
+    chunks = [f'{x["title"]}: {x["content"]}' for x in batch["metadata"]]
+    embeds = encoder.encode(chunks)
 
-        # Ensure correct length
-        assert len(embeds) == (i_end - i)
+    # Ensure correct length
+    assert len(embeds) == (i_end - i)
 
-        # Upsert to Pinecone
-        to_upsert = list(zip(batch["id"], embeds, batch["metadata"]))
-        index.upsert(vectors=to_upsert)
+    # Upsert to Pinecone
+    to_upsert = list(zip(batch["id"], embeds, batch["metadata"]))
+    index.upsert(vectors=to_upsert)
 
-def get_docs(query: str, index: any, encoder: any, top_k: int) -> list[str]:
+def get_docs(query: str, top_k: int) -> list[str]:
     xq = encoder.encode(query)
     res = index.query(vector=xq.tolist(), top_k=top_k, include_metadata=True)
     return [x["metadata"]['content'] for x in res["matches"]]
@@ -88,20 +86,22 @@ def get_response(query: str, docs: list[str], groq_client: any) -> str:
     )
     return chat_response.choices[0].message.content
 
+
+
 def handle_query(user_query: str):
-    # Upsert data into Pinecone (if necessary)
-    upsert_data_to_pinecone(index, data)
 
     # Initialize Groq client
     groq_client = Groq(api_key=GROQ_API_KEY)
 
     # Get relevant documents
-    docs = get_docs(user_query, index, encoder, top_k=5)
+    docs = get_docs(user_query, top_k=5)
 
     # Generate and return response
     response = get_response(user_query, docs, groq_client)
 
-    return response
+    for word in response.split():
+            yield word + " "
+            time.sleep(0.05)
 
 def main():
     st.title("Ask Anjibot 2.0")