Sample code for: LangChain Tutorial: Build Your First Chains and Agents

langchain-tutorial/.env.example

@@ -0,0 +1 @@
+OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>

langchain-tutorial/README.md

@@ -0,0 +1,66 @@
+# LangChain Tutorial: Build Your First Chains and Agents
+
+Sample code for the Real Python tutorial [LangChain Tutorial: Build Your First Chains and Agents](https://realpython.com/langchain-tutorial/).
+
+This is the `langchain_intro` project you build throughout the tutorial: a chat model, reusable prompt templates, an LCEL chain, a ChromaDB-backed review retriever (RAG), and a tool-calling agent that answers questions about patient reviews and hospital wait times.
+
+## Project layout
+
+```
+langchain-tutorial/
+│
+├── data/
+│   └── reviews.csv
+│
+├── langchain_intro/
+│   ├── chatbot.py            # final chat model + prompt templates + RAG chain + agent
+│   ├── create_retriever.py   # builds the ChromaDB vector database from reviews.csv
+│   └── tools.py              # get_current_wait_time() tool
+│
+├── .env.example
+├── requirements.txt
+└── README.md
+```
+
+## Setup
+
+1. Create and activate a virtual environment (Python 3.10 or later), then install the dependencies:
+
+   ```console
+   (venv) $ python -m pip install -r requirements.txt
+   ```
+
+2. Copy `.env.example` to `.env` and add your OpenAI API key:
+
+   ```console
+   (venv) $ cp .env.example .env
+   ```
+
+   ```dotenv
+   OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
+   ```
+
+3. Build the ChromaDB vector database from the reviews. Run this from the project root; it creates a `chroma_data/` directory with the embedded reviews:
+
+   ```console
+   (venv) $ python langchain_intro/create_retriever.py
+   ```
+
+## Try it out
+
+Start a Python REPL **from the project root** so the `langchain_intro` package is importable and `dotenv` finds your `.env`:
+
+```pycon
+>>> from langchain_intro.chatbot import review_chain
+>>> review_chain.invoke("Has anyone complained about communication with the hospital staff?")
+
+>>> from langchain_intro.chatbot import hospital_agent_executor
+>>> response = hospital_agent_executor.invoke(
+...     {
+...         "messages": [
+...             {"role": "user", "content": "What is the current wait time at hospital C?"}
+...         ]
+...     }
+... )
+>>> response["messages"][-1].text
+```

langchain-tutorial/langchain_intro/chatbot.py

@@ -0,0 +1,102 @@
+import dotenv
+from langchain.agents import create_agent
+from langchain_chroma import Chroma
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.prompts import (
+    ChatPromptTemplate,
+    HumanMessagePromptTemplate,
+    PromptTemplate,
+    SystemMessagePromptTemplate,
+)
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.tools import Tool
+from langchain_openai import ChatOpenAI
+from langchain_openai import OpenAIEmbeddings
+
+from langchain_intro.tools import get_current_wait_time
+
+REVIEWS_CHROMA_PATH = "chroma_data/"
+
+dotenv.load_dotenv()
+
+review_template_str = """Your job is to use patient
+reviews to answer questions about their experience at
+a hospital. Use the following context to answer questions.
+Be as detailed as possible, but don't make up any information
+that's not from the context. If you don't know an answer, say
+you don't know.
+
+{context}
+"""
+
+review_system_prompt = SystemMessagePromptTemplate(
+    prompt=PromptTemplate(
+        input_variables=["context"],
+        template=review_template_str,
+    )
+)
+
+review_human_prompt = HumanMessagePromptTemplate(
+    prompt=PromptTemplate(
+        input_variables=["question"],
+        template="{question}",
+    )
+)
+messages = [review_system_prompt, review_human_prompt]
+
+review_prompt_template = ChatPromptTemplate(
+    input_variables=["context", "question"],
+    messages=messages,
+)
+
+chat_model = ChatOpenAI(model="gpt-5.6-luna", output_version="responses/v1")
+
+output_parser = StrOutputParser()
+
+reviews_vector_db = Chroma(
+    persist_directory=REVIEWS_CHROMA_PATH,
+    embedding_function=OpenAIEmbeddings(),
+)
+
+reviews_retriever = reviews_vector_db.as_retriever(search_kwargs={"k": 10})
+
+review_chain = (
+    {"context": reviews_retriever, "question": RunnablePassthrough()}
+    | review_prompt_template
+    | chat_model
+    | output_parser
+)
+
+tools = [
+    Tool(
+        name="Reviews",
+        func=review_chain.invoke,
+        description="""Useful when you need to answer questions
+        about patient reviews or experiences at the hospital.
+        Not useful for answering questions about specific visit
+        details such as payer, billing, treatment, diagnosis,
+        chief complaint, hospital, or physician information.
+        Pass the entire question as input to the tool. For instance,
+        if the question is "What do patients think about the triage system?",
+        the input should be "What do patients think about the triage system?"
+        """,
+    ),
+    Tool(
+        name="Waits",
+        func=get_current_wait_time,
+        description="""Use when asked about current wait times
+        at a specific hospital. This tool can only get the current
+        wait time at a hospital and does not have any information about
+        aggregate or historical wait times. This tool returns wait times in
+        minutes. Do not pass the word "hospital" as input,
+        only the hospital name itself. For instance, if the question is
+        "What is the wait time at hospital A?", the input should be "A".
+        """,
+    ),
+]
+
+hospital_agent_executor = create_agent(
+    model=chat_model,
+    tools=tools,
+    system_prompt="You're a helpful assistant.",
+)

langchain-tutorial/langchain_intro/create_retriever.py

@@ -0,0 +1,32 @@
+import csv
+
+import dotenv
+from langchain_chroma import Chroma
+from langchain_core.documents import Document
+from langchain_openai import OpenAIEmbeddings
+
+REVIEWS_CSV_PATH = "data/reviews.csv"
+REVIEWS_CHROMA_PATH = "chroma_data"
+
+dotenv.load_dotenv()
+
+
+def load_reviews(csv_path):
+    with open(csv_path, newline="", encoding="utf-8") as csv_file:
+        reader = csv.DictReader(csv_file)
+        return [
+            Document(
+                page_content="\n".join(
+                    f"{column}: {value}" for column, value in row.items()
+                ),
+                metadata={"source": row["review"], "row": index},
+            )
+            for index, row in enumerate(reader)
+        ]
+
+
+reviews = load_reviews(REVIEWS_CSV_PATH)
+
+reviews_vector_db = Chroma.from_documents(
+    reviews, OpenAIEmbeddings(), persist_directory=REVIEWS_CHROMA_PATH
+)

langchain-tutorial/langchain_intro/tools.py

@@ -0,0 +1,14 @@
+import random
+import time
+
+
+def get_current_wait_time(hospital: str) -> int | str:
+    """Dummy function to generate fake wait times"""
+
+    if hospital not in ["A", "B", "C", "D"]:
+        return f"Hospital {hospital} does not exist"
+
+    # Simulate API call delay
+    time.sleep(1)
+
+    return random.randint(0, 10000)

langchain-tutorial/requirements.txt

@@ -0,0 +1,5 @@
+langchain==1.3.12
+langchain-openai==1.3.4
+python-dotenv==1.2.2
+chromadb==1.5.9
+langchain-chroma==1.1.0