Struct Chat

pub struct Chat {
    pub system_prompt: String,
    pub max_output_tokens: usize,
    pub history: Vec<Message>,
    pub tools: Option<Vec<LlmToolInfo>>,
    pub tool_choice: Option<ToolChoice>,
    /* private fields */
}

The main Chat client that users will interact with. All methods return a new instance rather than mutating the existing one, following the immutable builder pattern.

Fields

system_prompt: String

max_output_tokens: usize

history: Vec<Message>

tools: Option<Vec<LlmToolInfo>>

tool_choice: Option<ToolChoice>

Implementations

impl Chat

pub fn with_system_prompt(self, prompt: impl Into<String>) -> Self

Sets system prompt and returns a new instance

pub fn with_max_output_tokens(self, n: usize) -> Self

Sets max output tokens and returns a new instance

pub fn with_history(self, history: Vec<Message>) -> Self

Sets history and returns a new instance

pub fn add_message(self, msg: Message) -> Self

Adds a message to the conversation history and returns a new instance

pub fn push_message(self, msg: Message) -> Self

Alias for add_message for backward compatibility

pub fn tokens_used(&self) -> usize

Gets the current token count

pub fn with_tool(self, tool: impl ToolDefinition) -> Result<Self>

Add a tool and returns a new instance with the tool added

pub fn with_tools(self, tools: Vec<LlmToolInfo>) -> Self

Add multiple tools at once and return a new instance with the tools added

pub fn with_tool_choice(self, choice: ToolChoice) -> Self

Sets the tool choice strategy and returns a new instance

This method allows configuring how the model should choose tools:

• ToolChoice::Auto - Model can choose whether to use a tool (default)
• ToolChoice::Any - Model must use one of the available tools
• ToolChoice::None - Model must not use any tools
• ToolChoice::Specific(name) - Model must use the specified tool

Different providers implement this with slightly different terminology:

• OpenAI/Mistral use “auto”, “required”, “none”
• Anthropic uses “auto”, “any”, “none”
• Gemini uses function_calling_config with modes

The library transparently handles these differences, providing a consistent API regardless of which provider you’re using.

Examples

use language_barrier_core::{Chat, tool::ToolChoice};

// Require using a tool
let chat = Chat::default()
    .with_tool_choice(ToolChoice::Any);

// Specify a tool by name
let chat = Chat::default()
    .with_tool_choice(ToolChoice::Specific("weather_tool".to_string()));

// Disable tools for this conversation
let chat = Chat::default()
    .with_tool_choice(ToolChoice::None);

pub fn without_tool_choice(self) -> Self

Removes tool choice configuration and returns a new instance

This resets to the default behavior, where the model can choose whether to use tools.

pub fn most_recent_message(&self) -> Option<&Message>

Return the most recent message in the chat.

Trait Implementations

impl Clone for Chat

fn clone(&self) -> Chat

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Chat

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Chat

fn default() -> Self

Returns the “default value” for a type.