Struct rust_bert::pipelines::conversation::Conversation

pub struct Conversation {
    pub past_user_inputs: Vec<String>,
    pub generated_responses: Vec<String>,
    pub new_user_input: Option<String>,
    pub history: Vec<Vec<i64>>,
}

Data structure keeping track of a conversation in the system. It contains past user inputs and generated answers, a history of the tokens generated and a placeholder for new user inputs to be processed by the system if submitted for prediction

Fields

past_user_inputs: Vec<String>

Past user inputs that have already been processed

generated_responses: Vec<String>

Past system generated responses

new_user_input: Option<String>

New user input that needs to be processed

history: Vec<Vec<i64>>

History of the tokens passed as an input and generated so far used as context for next turn generation

Implementations

impl Conversation

pub fn new(text: &str) -> Conversation

Build a new Conversation with an initial user input

Arguments

• text - String with the initial user input to start a conversation

Example

use rust_bert::pipelines::conversation::Conversation;

let conversation = Conversation::new("Hi there!");

pub fn new_empty() -> Conversation

Build a new Conversation placeholder without user input

Example

use rust_bert::pipelines::conversation::Conversation;

let conversation = Conversation::new_empty();

pub fn add_user_input(&mut self, text: &str) -> Result<(), RustBertError>

Adds a new user input to the conversation. This method returns an error if an unprocessed user input already exists

Arguments

• text - &str with the additional user input to continue a conversation

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new_empty();
conversation.add_user_input("Hi there!").unwrap();

pub fn add_user_input_with_overwrite(&mut self, text: &str) -> Option<String>

Adds a new user input to the conversation. If an unprocessed user input already exists, its contents are overwritten by the new value provided.

Arguments

• text - &str with the additional user input to continue a conversation

Returns

• Option<String> containing overwritten string if applicable

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new_empty();
conversation
    .add_user_input("This input will not be used")
    .unwrap();
let unused_string = conversation.add_user_input_with_overwrite("Hi there!");

pub fn contains_new_input(&self) -> bool

Returns true if the conversation contains new user inputs to process

Returns

• bool flag indicating if the conversation contains new inputs to process

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new_empty();
let false_value = conversation.contains_new_input();
conversation
    .add_user_input("This input will not be used")
    .unwrap();
let true_value = conversation.contains_new_input();

pub fn mark_processed(&mut self)

Marks the conversation as processed and moves the user input that was up for processing to the past user inputs.

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new_empty();
let false_value = conversation.contains_new_input();
conversation
    .add_user_input("This input will not be used")
    .unwrap();
let true_value = conversation.contains_new_input();
conversation.mark_processed();
let false_value = conversation.contains_new_input();
assert_eq!(conversation.past_user_inputs.len(), 1usize);

pub fn get_last_input(&self) -> Option<&str>

Returns the last user input provided (including non-processed inputs).

Returns

• Option<&str> representation of the last user input provided

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new_empty();
let none_value = conversation.get_last_input();
conversation
    .add_user_input("This input will not be used")
    .unwrap();
let last_provided_input = conversation.get_last_input();
assert_eq!(last_provided_input, Some("This input will not be used"));

pub fn get_last_response(&self) -> Option<&str>

Returns the last response generated by the system.

Returns

• Option<&str> representation of the last response generated by the system.

Example

use rust_bert::pipelines::conversation::Conversation;

let mut conversation = Conversation::new("Hi There");
let non_value = conversation.get_last_response();

pub fn load_from_history<S, SI>(&mut self, texts: &[S], ids: &[SI])where
    S: AsRef<str>,
    SI: AsRef<[i64]>,

Initializes a conversation form a prior state. It is assumed that a conversation always start from a user interaction.

Arguments

• texts: sequence of strings, alternating between past user inputs and past generated responses.
• ids: sequence of sequence of ids, alternating between past user inputs and past generated responses. These can be generated via a ConversationModel’s encode_prompts.

Example:

use rust_bert::pipelines::conversation::{ConversationManager, ConversationModel};
use rust_bert::pipelines::generation_utils::LanguageGenerator;
let model = ConversationModel::new(Default::default())?;

let mut conversation_manager = ConversationManager::new();
let history = [
    "Going to the movies tonight - any suggestions?",
    "The Big Lebowski",
    "Is it an action movie?",
];
let encoded_history = model.encode_prompts(&history);

let conversation_1_id = conversation_manager.create_empty();
let _ = conversation_manager
    .get(&conversation_1_id)
    .unwrap()
    .load_from_history(&history, &encoded_history);

Trait Implementations

impl Clone for Conversation

fn clone(&self) -> Conversation

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for Conversation

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

Formats the value using the given formatter.