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
use http_req::{
    request::{Method, Request},
    uri::Uri,
};
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use urlencoding::encode;

use crate::Retry;

/// Response struct for the chat completion.
#[derive(Debug, Deserialize)]
pub struct ChatResponse {
    /// The response from ChatGPT.
    pub choice: String,
}

impl Default for ChatResponse {
    fn default() -> ChatResponse {
        ChatResponse {
            choice: String::new(),
        }
    }
}

/// struct for setting the chat options.
#[derive(Debug, Default, Serialize)]
pub struct ChatOptions<'a> {
    /// The ID or name of the model to use for completion.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub model: Option<&'a str>,

    /// The token limit of the model
    pub token_limit: u32,

    /// When true, a new conversation will be created.
    pub restart: bool,

    /// The prompt of the system role.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub system_prompt: Option<&'a str>,

    /// What sampling temperature to use, between 0 and 2.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub temperature: Option<f32>,

    /// An alternative to sampling with temperature
    #[serde(skip_serializing_if = "Option::is_none")]
    pub top_p: Option<f32>,

    /// Up to 4 sequences where the API will stop generating further tokens.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub stop: Option<Vec<String>>,

    /// The maximum number of tokens to generate in the chat completion.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub max_tokens: Option<u16>,

    /// Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub presence_penalty: Option<f32>,

    /// Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub frequency_penalty: Option<f32>,

    /// Modify the likelihood of specified tokens appearing in the completion.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub logit_bias: Option<HashMap<String, i8>>,
}

impl<'a> crate::LLMServiceFlows<'a> {
    /// Create chat completion with the provided sentence.
    /// It uses OpenAI's [GPT-4](https://platform.openai.com/docs/models/gpt-4) model to make a conversation.
    ///
    /// `conversation_id` is the identifier of the conversation.
    /// The history will be fetched and attached to the `sentence` as a whole prompt for ChatGPT.
    ///
    /// `sentence` is a String that reprensents the current utterance of the conversation.
    ///
    ///```rust  
    ///     // Create a conversation_id.
    ///     // Only numbers, letters, underscores, dashes, and pound signs are allowed, up to 50 characters.
    ///     let chat_id = format!("news-summary-N");
    ///     // System_prompt content in text.
    ///     let system = &format!("You're a news editor AI.");
    ///
    ///     // Create ChatOptions.
    ///     let co = ChatOptions {
    ///         model: Some("gpt-4"),
    ///         token_limit: 8192,
    ///         restart: true,
    ///         system_prompt: Some(system),
    ///     // Use .. to extract the default value for the remaining fields.
    ///         ..Default::default()
    ///     };
    ///
    ///     // Create a `sentence`, the concatenation of user prompt and the text to work with.
    ///     let question = format!("Make a concise summary within 200 words on this: {news_body}.");
    ///
    ///     // Chat completion to get the result and handle the failure.
    ///     match llm.chat_completion(&chat_id, &question, &co).await {
    ///         Ok(r) => Ok(r.choice),
    ///         Err(e) =>  Err(e.into()),
    ///     }
    /// ```
    pub async fn chat_completion(
        &self,
        conversation_id: &str,
        sentence: &str,
        options: &ChatOptions<'_>,
    ) -> Result<ChatResponse, String> {
        self.keep_trying(|endpoint, api_key| {
            chat_completion_inner(endpoint, api_key, conversation_id, sentence, options)
        })
    }
}

fn chat_completion_inner(
    endpoint: &str,
    api_key: &str,
    conversation_id: &str,
    sentence: &str,
    options: &ChatOptions,
) -> Retry<ChatResponse> {
    let flows_user = unsafe { crate::_get_flows_user() };
    let flow_id = unsafe { crate::_get_flow_id() };

    let mut writer = Vec::new();
    let uri = format!(
        "{}/{}/{}/chat_completion?endpoint={}&api_key={}&conversation={}",
        crate::LLM_API_PREFIX.as_str(),
        flows_user,
        flow_id,
        encode(endpoint),
        encode(api_key),
        encode(conversation_id),
    );
    let uri = Uri::try_from(uri.as_str()).unwrap();
    let body = serde_json::to_vec(&serde_json::json!({
        "sentence": sentence,
        "params": options
    }))
    .unwrap_or_default();
    match Request::new(&uri)
        .method(Method::POST)
        .header("Content-Type", "application/json")
        .header("Content-Length", &body.len())
        .body(&body)
        .send(&mut writer)
    {
        Ok(res) => {
            match res.status_code().is_success() {
                true => Retry::No(
                    serde_json::from_slice::<ChatResponse>(&writer)
                        .or(Err(String::from("Unexpected error"))),
                ),
                false => {
                    match res.status_code().into() {
                        409 | 429 | 503 => {
                            // 409 TryAgain 429 RateLimitError
                            // 503 ServiceUnavailable
                            Retry::Yes(String::from_utf8_lossy(&writer).into_owned())
                        }
                        _ => Retry::No(Err(String::from_utf8_lossy(&writer).into_owned())),
                    }
                }
            }
        }
        Err(e) => Retry::No(Err(e.to_string())),
    }
}

#[derive(Debug, Deserialize)]
#[serde(rename_all = "camelCase")]
pub enum ChatRole {
    User,
    Assistant,
}

#[derive(Debug, Deserialize)]
pub struct ChatMessage {
    pub role: ChatRole,
    pub content: String,
}

/// Fetch the question history of conversation_id
/// Result will be an array of string whose length is
/// restricted by limit.
/// When limit is 0, all history will be returned.
///
///```rust,no_run
/// // The conversation_id we are interested in.
/// let conversation_id = "unique_conversation_id";
/// // Limit the number of messages returned.
/// let limit: u8 = 10;
/// // Call `chat_history` to fetch the conversation history.
/// let history = chat_history(conversation_id, limit);
///
/// match history {
///     Some(messages) => {
///         println!("Chat history (most recent {} messages):", limit);
///         for message in messages.iter().rev() {
///             let role = match message.role {
///                 ChatRole::User => "User",
///                 ChatRole::Assistant => "Assistant",
///             };
///             println!("{}: {}", role, message.content);
///         }
///     }
///     None => {
///         println!(
///             "Failed to fetch chat history for conversation {}",
///             conversation_id
///         );
///     }
/// }
/// ```

pub fn chat_history(conversation_id: &str, limit: u8) -> Option<Vec<ChatMessage>> {
    let flows_user = unsafe { crate::_get_flows_user() };
    let flow_id = unsafe { crate::_get_flow_id() };

    let mut writer = Vec::new();
    let uri = format!(
        "{}/{}/{}/chat_history?conversation={}&limit={}",
        crate::LLM_API_PREFIX.as_str(),
        flows_user,
        flow_id,
        encode(conversation_id),
        limit
    );
    let uri = Uri::try_from(uri.as_str()).unwrap();
    match Request::new(&uri).method(Method::GET).send(&mut writer) {
        Ok(res) => match res.status_code().is_success() {
            true => serde_json::from_slice::<Vec<ChatMessage>>(&writer).ok(),
            false => None,
        },
        Err(_) => None,
    }
}