🔗 LangChain · LangChain · LangChain

1 📦 Install it, and meet the pipe · ติดตั้ง แล้วรู้จักท่อ · 安装，认识那根管子

terminal

pip install langchain langchain-ollama

langchain_demo.py — the setup

"""Four small LangChain examples, all against a free model on this machine."""
import time

from langchain.agents import create_agent
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.tools import tool
from langchain_ollama import ChatOllama
from pydantic import BaseModel, Field

model = ChatOllama(model="qwen3", base_url="http://localhost:11434",
                   num_ctx=8192, reasoning=False)      # reasoning=False: no <think> block

2 🔗 One chain: prompt, model, parser · ท่อแรก พรอมป์ต โมเดล ตัวแปลง · 第一条链：提示词、模型、解析器

the chain

# ── 1. the pipe: prompt | model | parser ──────────────────────────────────────
explain = ChatPromptTemplate.from_template(
    "Explain {thing} to a Grade 9 student in two short sentences.")
chain = explain | model | StrOutputParser()
timed("1 CHAIN", lambda: chain.invoke({"thing": "why the sky is blue"}))

what it printed (12s)

The sky appears blue because sunlight passes through the atmosphere and shorter blue wavelengths are scattered in all directions by tiny gas molecules. This scattered blue light reaches our eyes, making the sky look blue on a clear day.

3 🧱 Structured output: an object, not a paragraph · ผลลัพธ์มีโครงสร้าง ได้ออบเจ็กต์ ไม่ใช่ย่อหน้า · 结构化输出：拿到对象，不是段落

the shape, and the call

# ── 2. structured output: you get an object, not a paragraph ──────────────────
class Word(BaseModel):
    word: str = Field(description="an English word a market seller would use")
    thai: str = Field(description="the Thai translation")
    example: str = Field(description="one short example sentence")


timed("2 STRUCTURED",
      lambda: model.with_structured_output(Word).invoke(
          "Give one useful English word for a Chiang Mai market."))

what it printed (7s)

word='Warorot' thai='วอร์ร็อต' example='Warorot Market is a popular spot for local crafts and street food.'

4 ⛓️ Two steps, joined · ต่อสองขั้นเข้าด้วยกัน · 把两步接起来

two chains, one pipe

# ── 3. two steps, joined ──────────────────────────────────────────────────────
shorten = ChatPromptTemplate.from_template(
    "Rewrite this in one sentence a beginner can read:\n\n{text}")
two_step = {"text": chain} | shorten | model | StrOutputParser()
timed("3 TWO STEPS", lambda: two_step.invoke({"thing": "why rice fields need water"}))

what it printed (13s)

Rice plants need water to grow because their roots use it to get nutrients and oxygen, and water also helps them stand upright and develop their grains.

5 🧰 An agent with a tool · agent ที่มีเครื่องมือ · 带工具的智能体

the tool, and the agent

# ── 4. an agent with a tool ───────────────────────────────────────────────────
@tool
def lessons_in_a_day(minutes: int) -> str:
    """Says how many lessons fit in a seven hour school day.

    Args:
        minutes: how long one lesson lasts.
    """
    return f"{(7 * 60) // minutes} lessons of {minutes} minutes fit in a seven hour school day."


agent = create_agent(model, tools=[lessons_in_a_day],
                     system_prompt="You are a school planner. Use the tool before you answer.")
timed("4 AGENT", lambda: agent.invoke(
    {"messages": [{"role": "user", "content": "How many 45 minute lessons fit in a school day?"}]}
)["messages"][-1].content)

what it printed (16s)

There are 9 lessons of 45 minutes each that fit in a seven-hour school day.

6 👉 LangChain, or one of the others? · เลือก LangChain หรือตัวอื่น · 选 LangChain，还是别的？