Updates to handle history

app.py → app01-simpleRag.py

@@ -11,16 +11,12 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_google_genai import ChatGoogleGenerativeAI
 
 # Pinecone vector database
-# import langchain.vectorstores as vs
-# from langchain_pinecone import Pinecone
-# import pinecone
 from pinecone import Pinecone, ServerlessSpec
 from langchain_pinecone import PineconeVectorStore
 
 
 setid = "global"
 
-#EMBEDDINGS_MODEL = "BAAI/bge-base-en-v1.5"  # Ranking 8, 768
 embeddings = HuggingFaceEmbeddings(model_name=os.getenv("EMBEDDINGS_MODEL"))
 
 # model = ChatOpenAI(temperature=0.0)

app02-chatRag.py

@@ -0,0 +1,95 @@
+# Run with reload mode:
+# gradio app02-chatRag.py
+
+import os
+import gradio as gr
+
+# Langchain
+from langchain.chains import RetrievalQA
+from langchain.prompts import ChatPromptTemplate
+from langchain_core.runnables import RunnablePassthrough
+
+# HuggingFace
+from langchain_community.embeddings import HuggingFaceEmbeddings
+
+# GeminiPro
+from langchain_google_genai import ChatGoogleGenerativeAI
+
+# Pinecone vector database
+from pinecone import Pinecone, ServerlessSpec
+from langchain_pinecone import PineconeVectorStore
+
+
+setid = "global"
+
+embeddings = HuggingFaceEmbeddings(model_name=os.getenv("EMBEDDINGS_MODEL"))
+
+# model = ChatOpenAI(temperature=0.0)
+model = ChatGoogleGenerativeAI(
+    model="gemini-pro", temperature=0.1, convert_system_message_to_human=True
+)
+
+pc = Pinecone(
+        api_key=os.getenv("PINECONE_API_KEY")
+    )
+index = pc.Index(setid)
+vectorstore = PineconeVectorStore(index, embeddings, "text")
+
+
+template_no_history = """Answer the question based only on the following context:
+{context}
+
+Question: {question}
+"""
+PROMPT_NH = ChatPromptTemplate.from_template(template_no_history)
+
+template_with_history = """Given the following conversation history, answer the follow up question:
+Chat History:
+{chat_history}
+
+Question: {question}
+"""
+PROMPT_WH = ChatPromptTemplate.from_template(template_with_history)
+
+
+def pipeLog(x):
+    print("***", x)
+    return x
+
+
+
+def rag_query(question: str, history: list[list[str]]):
+    if len(history)==0:
+        chain = (
+            pipeLog 
+            | { "context": vectorstore.as_retriever(kwargs={"k":5}), "question": RunnablePassthrough() }
+            | PROMPT_NH 
+            | pipeLog 
+            | model
+        )
+        response = chain.invoke(question)
+        print(response)
+        return response
+    else:
+        chat_history = ""
+        for l in history:
+            chat_history += " : ".join(l)
+        chain = (
+            pipeLog 
+            | { "chat_history": chat_history, "question": RunnablePassthrough() }
+            | PROMPT_WH 
+            | pipeLog 
+            | model
+        )
+        response = chain.invoke(question)
+        return response
+
+
+
+gr.ChatInterface(
+    rag_query,
+    title="RAG Chatbot demo",
+    description="A chatbot doing Retrieval Augmented Generation, backed by a Pinecone vector database"
+    ).launch()
+
+

modules.md

@@ -8,6 +8,8 @@ pip install \
     langchain-pinecone              \
     huggingface_hub
 
+pip install ipykernel IProgress ipywidgets --upgrade
+
     # python-dotenv                   \
     # pinecone-client==2.2.4          \
 ```

test.ipynb

@@ -0,0 +1,130 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import gradio as gr\n",
+    "\n",
+    "# Langchain\n",
+    "from langchain.chains import RetrievalQA\n",
+    "from langchain.prompts import ChatPromptTemplate\n",
+    "from langchain_core.runnables import RunnablePassthrough\n",
+    "\n",
+    "# HuggingFace\n",
+    "from langchain_community.embeddings import HuggingFaceEmbeddings\n",
+    "\n",
+    "# GeminiPro\n",
+    "from langchain_google_genai import ChatGoogleGenerativeAI\n",
+    "\n",
+    "# Pinecone vector database\n",
+    "from pinecone import Pinecone, ServerlessSpec\n",
+    "from langchain_pinecone import PineconeVectorStore\n",
+    "\n",
+    "\n",
+    "setid = \"global\"\n",
+    "\n",
+    "embeddings = HuggingFaceEmbeddings(model_name=os.getenv(\"EMBEDDINGS_MODEL\"))\n",
+    "\n",
+    "# model = ChatOpenAI(temperature=0.0)\n",
+    "model = ChatGoogleGenerativeAI(\n",
+    "    model=\"gemini-pro\", temperature=0.1, convert_system_message_to_human=True\n",
+    ")\n",
+    "\n",
+    "pc = Pinecone(\n",
+    "        api_key=os.getenv(\"PINECONE_API_KEY\")\n",
+    "    )\n",
+    "index = pc.Index(setid)\n",
+    "vectorstore = PineconeVectorStore(index, embeddings, \"text\")\n",
+    "\n",
+    "\n",
+    "template_no_history = \"\"\"Answer the question based only on the following context:\n",
+    "{context}\n",
+    "\n",
+    "Question: {question}\n",
+    "\"\"\"\n",
+    "PROMPT_NH = ChatPromptTemplate.from_template(template_no_history)\n",
+    "\n",
+    "template_with_history = \"\"\"Given the following conversation history, answer the follow up question:\n",
+    "Chat History:\n",
+    "{chat_history}\n",
+    "\n",
+    "Question: {question}\n",
+    "\"\"\"\n",
+    "PROMPT_WH = ChatPromptTemplate.from_template(template_with_history)\n",
+    "\n",
+    "\n",
+    "def pipeLog(x):\n",
+    "    print(\"***\", x)\n",
+    "    return x\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "content='A blockchain is a distributed ledger technology that enables secure and immutable record-keeping of digital transactions. It comprises a chain of blocks, each containing a list of validated and time-stamped transactions.'\n"
+     ]
+    }
+   ],
+   "source": [
+    "question = \"What is a blockchain?\"\n",
+    "\n",
+    "# chain = (\n",
+    "#     pipeLog \n",
+    "#     | { \"context\": vectorstore.as_retriever(kwargs={\"k\":5}), \"question\": RunnablePassthrough() }\n",
+    "#     | PROMPT_NH \n",
+    "#     | pipeLog \n",
+    "#     | model\n",
+    "# )\n",
+    "\n",
+    "\n",
+    "chain = (\n",
+    "    { \"context\": vectorstore.as_retriever(kwargs={\"k\": 5}), \"question\": RunnablePassthrough() }\n",
+    "    | PROMPT_NH\n",
+    "    | model\n",
+    ")\n",
+    "\n",
+    "\n",
+    "response = chain.invoke(question)\n",
+    "print(response)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": ".venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}