langchain-chatbot

README

LangChain-Go Chatbot with Zerfoo

This example demonstrates using Zerfoo as an LLM provider through the integrations/langchain adapter. The adapter exposes the same method signatures as LangChain-Go's schema.LLM interface, so it works as a drop-in replacement pointing at Zerfoo's OpenAI-compatible API.

Prerequisites

• A compiled zerfoo binary (or go run ./cmd/zerfoo)
• A GGUF model file (e.g. Llama 3 1B)

Setup

1. Start the Zerfoo server:

zerfoo serve --model path/to/model.gguf --port 8080

1. Run the chatbot:

go run ./examples/langchain-chatbot/ \
  --server http://localhost:8080 \
  --model llama3 \
  --temperature 0.7 \
  --max-tokens 512

1. Type a message and press Enter. Type quit to exit.

How It Works

The adapter (integrations/langchain.Adapter) wraps Zerfoo's /v1/chat/completions endpoint. It implements Call, Generate, and Type methods matching LangChain-Go's LLM interface without importing the langchaingo module, keeping the dependency footprint minimal.

llm := langchain.NewAdapter(
    "http://localhost:8080",
    "llama3",
    langchain.WithTemperature(0.7),
    langchain.WithMaxTokens(512),
)

reply, err := llm.Call(ctx, "What is the capital of France?")

Flags

Flag	Default	Description
--server	http://localhost:8080	Zerfoo server URL
--model	llama3	Model name
--temperature	0.7	Sampling temperature (0-1)
--max-tokens	512	Maximum tokens to generate

Documentation

Overview

Command langchain-chatbot demonstrates using the Zerfoo LangChain adapter as a drop-in LLM for a simple interactive chatbot loop.

Start a Zerfoo server first:

zerfoo serve --model path/to/model.gguf --port 8080

Then run this example:

go run ./examples/langchain-chatbot/ --server http://localhost:8080 --model llama3