LLMs are trained on a “fixed” data set (i.e., they have a “knowledge cutoff”).
I don’t have the ability to check real-time weather data.
LLMs are trained on a “fixed” data set – providers typically refer to this as a “knowledge cutoff”.
Three main techniques:
RAG is a popular term, but the core idea is outdated. Now you would supply a tool that can retrieve relevant information, and let the model call it.
Note that:
x tool with y argumentsexercises/chat-tools.py.eval() it.⚠️ This is a illustrative example — never execute arbitrary code in production without proper sandboxing and security measures.
Three main strategies to mitigate risk:
chatlas comes with some “built-in” tools that use the provider’s native capabilities.
These currently include:
tool_web_search() — search the web for current informationtool_web_fetch() — read a specific URLMCP tools can be served through either:
MCP tools are async, which means you must also use .chat_async() to call them.
MCP tools can be served through either:
exercises/chat-mcp-tools.py. Tweak it to use the MCP server you picked.raghilda for the store. Learn more here.