This video is a step-by-step tutorial to create an agentic application with Langchain and Ollama locally with function calling.
Code:
conda create -n agentic python=3.11 -y && conda activate agentic
pip install langchain-experimental
from langchain_experimental.llms.ollama_functions import OllamaFunctions
model = OllamaFunctions(model="llama3:8b", format="json")
model = model.bind_tools(
tools=[
{
"name": "get_current_weather",
"description": "Get the current weather in a given location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, " "e.g. San Francisco, CA",
},
"unit": {
"type": "string",
"enum": ["celsius", "fahrenheit"],
},
},
"required": ["location"],
},
}
],
function_call={"name": "get_current_weather"},
)
from langchain_core.messages import HumanMessage
model.invoke("what is the weather in Boston?")