LangChain helps build language-model applications without wiring every prompt, model call, parser, tool, and workflow step by hand. Its main abstraction, LangChain Expression Language (LCEL), composes basic components into reusable chains. section 7.3.1
A simple chain turns a prompt template into a model request and parses the result. That abstraction is excessive for a single f-string, but useful when a pipeline grows into retrieval, SQL, API calls, synthetic-data generation, or asynchronous branching. section 7.3.1
Listing 7.1 Example of creating a basic LangChain chain
import os
from langchain.chat_models import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
from langchain.schema.output_parser import StrOutputParser
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
prompt = ChatPromptTemplate.from_template("Tell me a story about {topic}")
model = ChatOpenAI(model="gpt-3.5-turbo", openai_api_key=OPENAI_API_KEY)
output_parser = StrOutputParser()
chain = prompt | model | output_parser
chain.invoke("the printing press")
A minimal Python chain connects a chat prompt, an OpenAI model, and a string parser with the LCEL pipe operator. section 7.3.1
Use LangChain when the surrounding workflow matters as much as the model call:
- Combine external data retrieval tools, such as web and YouTube search, with model-generated summaries. section 8.3
- Add LangServe to host a chain as an API and LangSmith to trace how results change across its links. section 7.3.1
- Treat chains as language-model data pipelines or DAGs rather than assuming every workflow is linear. section 7.3.1
- Approach agents cautiously: Python and CSV agents can execute tools and inspect data, yet examples show parsing, import, and reasoning failures. section 8.3
From search tools to agents
A basic setup calls DuckDuckGo and YouTube search directly. A larger agent setup combines those tools with Python execution, mathematical tools, conversation memory, and a local Mistral model; the resulting behavior still requires skepticism and inspection. section 8.3
