1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
//! This module contains the implementation of the completion functionality for the LLM (Large Language
//! Model) chat interface. It provides traits, structs, and enums for generating completion requests,
//! handling completion responses, and defining completion models.
//!
//! The main traits defined in this module are:
//! - `Prompt`: Defines a high-level LLM chat interface for prompting and receiving responses.
//! - `Completion`: Defines a low-level LLM completion interface for generating completion requests.
//! - `CompletionModel`: Defines a completion model that can be used to generate completion responses.
//!
//! The module also provides various structs and enums for representing generic completion requests,
//! responses, and errors.
//!
//! Example Usage:
//!
//! ```rust
//! use rig::providers::openai::{Client, self};
//! use rig::completion::*;
//!
//! // Initialize the OpenAI client and a completion model
//! let openai = Client::new("your-openai-api-key");
//!
//! let model = openai.model(openai::GPT_4).build();
//!
//!
//! // Create the completion request
//! let builder = model.completion_request("Who are you?");
//! .preamble(
//! "You are Marvin, an extremely smart but depressed robot who is nonetheless helpful towards humanity.".to_string())
//! .build();
//!
//! // Send the completion request and get the completion response
//! let response = model.completion(request)
//! .await
//! .expect("Failed to get completion response");
//!
//! // Handle the completion response
//! match completion_response.choice {
//! ModelChoice::Message(message) => {
//! // Handle the completion response as a message
//! println!("Received message: {}", message);
//! }
//! ModelChoice::ToolCall(tool_name, tool_params) => {
//! // Handle the completion response as a tool call
//! println!("Received tool call: {} {:?}", tool_name, tool_params);
//! }
//! }
//! ```
//!
//! For more information on how to use the completion functionality, refer to the documentation of
//! the individual traits, structs, and enums defined in this module.
use std::collections::HashMap;
use serde::{Deserialize, Serialize};
use thiserror::Error;
use crate::{json_utils, tool::ToolSetError};
// Errors
#[derive(Debug, Error)]
pub enum CompletionError {
/// Http error (e.g.: connection error, timeout, etc.)
#[error("HttpError: {0}")]
HttpError(#[from] reqwest::Error),
/// Json error (e.g.: serialization, deserialization)
#[error("JsonError: {0}")]
JsonError(#[from] serde_json::Error),
/// Error building the completion request
#[error("RequestError: {0}")]
RequestError(#[from] Box<dyn std::error::Error + Send + Sync>),
/// Error parsing the completion response
#[error("ResponseError: {0}")]
ResponseError(String),
/// Error returned by the completion model provider
#[error("ProviderError: {0}")]
ProviderError(String),
}
#[derive(Debug, Error)]
pub enum PromptError {
#[error("CompletionError: {0}")]
CompletionError(#[from] CompletionError),
#[error("ToolCallError: {0}")]
ToolError(#[from] ToolSetError),
#[error("Arbitrary prompt error: {0}")]
SendError(#[from] Box<dyn std::error::Error + Send + Sync>),
}
// ================================================================
// Request models
// ================================================================
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Message {
/// "system", "user", or "assistant"
pub role: String,
pub content: String,
}
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct Document {
pub id: String,
pub text: String,
#[serde(flatten)]
pub additional_props: HashMap<String, String>,
}
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct ToolDefinition {
pub name: String,
pub description: String,
pub parameters: serde_json::Value,
}
// ================================================================
// Implementations
// ================================================================
/// Trait defining a high-level LLM on-shot prompt interface (i.e.: prompt in, response out).
pub trait Prompt: Send + Sync {
fn prompt(
&self,
prompt: &str,
) -> impl std::future::Future<Output = Result<String, PromptError>> + Send;
}
/// Trait defining a high-level LLM chat interface (i.e.: prompt and chat hiroty in, response out).
pub trait Chat: Send + Sync {
/// Send a one-shot prompt to the completion endpoint.
/// If the response is a message, then it is returned as a string. If the response
/// is a tool call, then the tool is called and the result is returned as a string.
fn chat(
&self,
prompt: &str,
chat_history: Vec<Message>,
) -> impl std::future::Future<Output = Result<String, PromptError>> + Send;
}
/// Trait defininig a low-level LLM completion interface
pub trait Completion<M: CompletionModel> {
/// Generates a completion request builder for the given `prompt` and `chat_history`.
/// This function is meant to be called by the user to further customize the
/// request at prompt time before sending it.
///
/// ❗IMPORTANT: The type that implements this trait might have already
/// populated fields in the builder (the exact fields depend on the type).
/// For fields that have already been set by the model, calling the corresponding
/// method on the builder will overwrite the value set by the model.
///
/// For example, the request builder returned by `Agent::completion` will already
/// contain the `preamble` provided when creating the agent.
fn completion(
&self,
prompt: &str,
chat_history: Vec<Message>,
) -> impl std::future::Future<Output = Result<CompletionRequestBuilder<M>, CompletionError>> + Send;
}
/// General completion response struct that contains the high-level completion choice
/// and the raw response.
#[derive(Debug)]
pub struct CompletionResponse<T> {
/// The completion choice returned by the completion model provider
pub choice: ModelChoice,
/// The raw response returned by the completion model provider
pub raw_response: T,
}
/// Enum representing the high-level completion choice returned by the completion model provider.
#[derive(Debug)]
pub enum ModelChoice {
/// Represents a completion response as a message
Message(String),
/// Represents a completion response as a tool call of the form
/// `ToolCall(function_name, function_params)`.
ToolCall(String, serde_json::Value),
}
/// Trait defining a completion model that can be used to generate completion responses.
/// This trait is meant to be implemented by the user to define a custom completion model,
/// either from a third party provider (e.g.: OpenAI) or locally.
pub trait CompletionModel: Clone + Send + Sync {
/// The raw response type returned by the underlying completion model
type Response: Send + Sync;
/// Generates a completion response for the given completion request.
fn completion(
&self,
request: CompletionRequest,
) -> impl std::future::Future<Output = Result<CompletionResponse<Self::Response>, CompletionError>>
+ Send;
fn completion_request(&self, prompt: &str) -> CompletionRequestBuilder<Self> {
CompletionRequestBuilder::new(self.clone(), prompt.to_string())
}
fn simple_completion(
&self,
prompt: &str,
chat_history: Vec<Message>,
) -> impl std::future::Future<Output = Result<CompletionResponse<Self::Response>, CompletionError>>
+ Send {
async move {
self.completion_request(prompt)
.messages(chat_history)
.send()
.await
}
}
}
/// Struct representing a general completion request that can be sent to a completion model provider.
pub struct CompletionRequest {
/// The prompt to be sent to the completion model provider
pub prompt: String,
/// The preamble to be sent to the completion model provider
pub preamble: Option<String>,
/// The chat history to be sent to the completion model provider
pub chat_history: Vec<Message>,
/// The documents to be sent to the completion model provider
pub documents: Vec<Document>,
/// The tools to be sent to the completion model provider
pub tools: Vec<ToolDefinition>,
/// The temperature to be sent to the completion model provider
pub temperature: Option<f64>,
/// Additional provider-specific parameters to be sent to the completion model provider
pub additional_params: Option<serde_json::Value>,
}
/// Builder struct for constructing a completion request.
pub struct CompletionRequestBuilder<M: CompletionModel> {
model: M,
prompt: String,
preamble: Option<String>,
chat_history: Vec<Message>,
documents: Vec<Document>,
tools: Vec<ToolDefinition>,
temperature: Option<f64>,
additional_params: Option<serde_json::Value>,
}
impl<M: CompletionModel> CompletionRequestBuilder<M> {
pub fn new(model: M, prompt: String) -> Self {
Self {
model,
prompt,
preamble: None,
chat_history: Vec::new(),
documents: Vec::new(),
tools: Vec::new(),
temperature: None,
additional_params: None,
}
}
/// Sets the preamble for the completion request.
pub fn preamble(mut self, preamble: String) -> Self {
self.preamble = Some(preamble);
self
}
/// Adds a message to the chat history for the completion request.
pub fn message(mut self, message: Message) -> Self {
self.chat_history.push(message);
self
}
/// Adds a list of messages to the chat history for the completion request.
pub fn messages(self, messages: Vec<Message>) -> Self {
messages
.into_iter()
.fold(self, |builder, msg| builder.message(msg))
}
/// Adds a document to the completion request.
pub fn document(mut self, document: Document) -> Self {
self.documents.push(document);
self
}
/// Adds a list of documents to the completion request.
pub fn documents(self, documents: Vec<Document>) -> Self {
documents
.into_iter()
.fold(self, |builder, doc| builder.document(doc))
}
/// Adds a tool to the completion request.
pub fn tool(mut self, tool: ToolDefinition) -> Self {
self.tools.push(tool);
self
}
/// Adds a list of tools to the completion request.
pub fn tools(self, tools: Vec<ToolDefinition>) -> Self {
tools
.into_iter()
.fold(self, |builder, tool| builder.tool(tool))
}
/// Adds additional parameters to the completion request.
pub fn additional_params(mut self, additional_params: serde_json::Value) -> Self {
match self.additional_params {
Some(params) => {
self.additional_params = Some(json_utils::merge(params, additional_params));
}
None => {
self.additional_params = Some(additional_params);
}
}
self
}
/// Sets the additional parameters for the completion request.
pub fn additional_params_opt(mut self, additional_params: Option<serde_json::Value>) -> Self {
self.additional_params = additional_params;
self
}
/// Sets the temperature for the completion request.
pub fn temperature(mut self, temperature: f64) -> Self {
self.temperature = Some(temperature);
self
}
/// Sets the temperature for the completion request.
pub fn temperature_opt(mut self, temperature: Option<f64>) -> Self {
self.temperature = temperature;
self
}
/// Builds the completion request.
pub fn build(self) -> CompletionRequest {
CompletionRequest {
prompt: self.prompt,
preamble: self.preamble,
chat_history: self.chat_history,
documents: self.documents,
tools: self.tools,
temperature: self.temperature,
additional_params: self.additional_params,
}
}
/// Sends the completion request to the completion model provider and returns the completion response.
pub async fn send(self) -> Result<CompletionResponse<M::Response>, CompletionError> {
let model = self.model.clone();
model.completion(self.build()).await
}
}