LangChain & Ollama: Building Complex Systems

If you are building more than just a chatbot—perhaps an agent that researches the web, writes code, and then runs tests—you need an orchestration framework. LangChain is the industry leader for this, and it has "First-Class" support for Ollama.

1. Why use LangChain?

• Interchangeability: You can write your code once. If you start with a local llama3 and later want to upgrade to GPT-4o, you only change one line of code.
• Memory: LangChain handles the complex logic of "sliding windows" for your chat history automatically.
• Connectors: LangChain makes it easy to connect Ollama to PDFs, websites, and databases.

---

2. Basic Setup (Python)

from langchain_community.llms import Ollama

llm = Ollama(model="llama3")

response = llm.invoke("What is the capital of Mars?")
print(response)

---

Visualizing the Process

graph TD
    Start[Input] --> Process[Processing]
    Process --> Decision{Check}
    Decision -->|Success| End[Complete]
    Decision -->|Retry| Process

---

3. The "Chat" Interface

For conversational apps, use ChatOllama:

from langchain_ollama import ChatOllama
from langchain_core.messages import HumanMessage

model = ChatOllama(model="llama3")

messages = [
    HumanMessage(content="Hi! I'm an AI engineer.")
]

response = model.invoke(messages)
print(response.content)

---

4. Building a "Chain"

LangChain allows you to "Pipe" different instructions together.

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_template("Translate this to French: {input}")
chain = prompt | model

result = chain.invoke({"input": "Hello, how are you?"})
print(result.content)

---

5. Local Embeddings

LangChain can also use Ollama to generate Embeddings (vectors) for RAG systems. This ensures that even your document indexing stays local and private.

from langchain_community.embeddings import OllamaEmbeddings

embeddings = OllamaEmbeddings(model="llama3")
vector = embeddings.embed_query("This is a document about privacy.")

---

Key Takeaways

• LangChain provides a standardized way to talk to LLMs.
• It simplifies interchangeability between local and cloud models.
• ChatOllama and OllamaEmbeddings are the two primary classes used in the Ollama ecosystem.
• Chains allow you to build complex logic by piping prompts and results together.