added intent detection code

Author: amitavroy

.gitignore

@@ -1,3 +1,4 @@
 .venv/*
 __pycache__
 db/*
+db-intents/*

data/intent.csv

@@ -0,0 +1,3 @@
+sentence,intent
+where is my order, order_status
+i want to check my order status, order_status

detect.py

@@ -0,0 +1,52 @@
+from langchain_ollama import OllamaEmbeddings, ChatOllama
+from langchain_chroma import Chroma
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import StrOutputParser
+
+def format_docs(docs):
+    return "\n\n".join(f"Sentence: {doc.page_content}\nIntent: {doc.metadata.get('intent', 'Unknown')}" for doc in docs)
+
+def detect_intent_with_context(sentence):
+    messages = []
+    persist_directory = "db-intents"
+    local_embeddings = OllamaEmbeddings(model="llama3.1:8b")
+    
+    vectorstore = Chroma(persist_directory=persist_directory, embedding_function=local_embeddings)
+    
+    docs = vectorstore.similarity_search(sentence)
+    if not docs:
+        messages.append("No relevant intent was found.")
+        return {"response": "No matching intent found.", "messages": messages}
+    
+    # Define the RAG prompt template for intent detection
+    INTENT_TEMPLATE = """
+    You are an assistant for detecting the intent of a sentence. Use the following pieces of retrieved context to determine the intent. If you cannot determine the intent, just say that you don't know. 
+
+    <context>
+    {context}
+    </context>
+
+    What is the intent of the following sentence? Just reply with the intent name and nothing else as response.
+
+    {sentence}"""
+    
+    intent_prompt = ChatPromptTemplate.from_template(INTENT_TEMPLATE)
+    model = ChatOllama(model="llama3.1:8b")
+    
+    chain = (
+        RunnablePassthrough.assign(context=lambda input: format_docs(input["context"]))
+        | intent_prompt
+        | model
+        | StrOutputParser()
+    )
+    
+    response = chain.invoke({"context": docs, "sentence": sentence})
+    return {"response": response, "messages": messages}
+
+# Example usage
+test_sentence = "Where is it?"
+result = detect_intent_with_context(test_sentence)
+print(f"{result['response']}")
+for message in result["messages"]:
+    print(message)

intent.py

@@ -0,0 +1,45 @@
+import csv
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_ollama import OllamaEmbeddings
+from langchain_chroma import Chroma
+from langchain.schema import Document
+import os
+
+def ingest_csv_to_chroma(csv_path):
+    messages = []
+    local_embeddings = OllamaEmbeddings(model="llama3.1:8b")
+    persist_directory = "db-intents"
+    
+    if os.path.exists(persist_directory):
+        vectorstore = Chroma(persist_directory=persist_directory, embedding_function=local_embeddings)
+        messages.append("Loaded the existing Chroma DB")
+    else:
+        vectorstore = Chroma(persist_directory=persist_directory, embedding_function=local_embeddings)
+        messages.append("Created the Chroma DB")
+    
+    # Load data from CSV
+    documents = []
+    with open(csv_path, 'r') as csv_file:
+        reader = csv.DictReader(csv_file)
+        for row in reader:
+            sentence = row['sentence']
+            intent = row['intent']
+            # Create a Document object with metadata for intent data
+            doc = Document(page_content=sentence, metadata={"intent": intent, "type": "intent_data"})
+            documents.append(doc)
+    
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
+    all_splits = text_splitter.split_documents(documents)
+    
+    vectorstore.add_documents(documents=all_splits)
+    messages.append("Added intent data to Chroma DB")
+    
+    return messages
+
+# Example usage
+csv_path = "data/intent.csv"
+
+# Ingest CSV data
+csv_messages = ingest_csv_to_chroma(csv_path)
+for message in csv_messages:
+    print(message)

scrape.py

@@ -27,5 +27,3 @@ def fetch_and_persist_article(url):
     messages.append(f"Added to Chroma DB")
 
     return messages
-    
-