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

/// Models for Chat
#[derive(Debug, Clone, Copy)]
pub enum ChatModel {
    ChatBison,
}

impl Serialize for ChatModel {
    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        match self {
            ChatModel::ChatBison => serializer.serialize_str("chat-bison"),
        }
    }
}

impl Default for ChatModel {
    fn default() -> ChatModel {
        ChatModel::ChatBison
    }
}

/// 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.
    pub model: ChatModel,

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

    /// Context shapes how the model responds throughout the conversation
    #[serde(skip_serializing_if = "Option::is_none")]
    pub context: Option<&'a str>,

    /// List of structured messages to the model to learn how to respond to the conversation.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub examples: Option<Vec<(String, String)>>,

    /// The temperature is used for sampling during the response generation, which occurs when topP and topK are applied.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub temperature: Option<f32>,

    /// Top-p changes how the model selects tokens for output.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub top_p: Option<f32>,

    /// Top-k changes how the model selects tokens for output.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub top_k: Option<u8>,

    /// Maximum number of tokens that can be generated in the response.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub max_output_tokens: Option<u16>,
}

/// Create chat completion with the provided sentence.
/// It uses Vertex's [chat-bison](https://cloud.google.com/vertex-ai/docs/generative-ai/learn/models?hl=en) 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 Vertex.
///
/// `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: ChatModel::ChatBison,
///         restart: true,
///         context: 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 to get the result and handle the failure.
///     match chat(&chat_id, &question, &co).await {
///         Ok(r) => Ok(r),
///         Err(e) =>  Err(e.into()),
///     }
/// ```
pub async fn chat(
    conversation_id: &str,
    sentence: &str,
    options: &ChatOptions<'_>,
) -> Result<String, String> {
    let flows_user = unsafe { crate::_get_flows_user() };
    let flow_id = unsafe { crate::_get_flow_id() };

    let mut writer = Vec::new();
    let uri = format!(
        "{}/{}/{}/vertex/chat?&conversation={}",
        crate::GOOGLE_CLOUD_SERVICE_API_PREFIX.as_str(),
        flows_user,
        flow_id,
        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 => String::from_utf8(writer).or(Err(String::from("Unexpected error"))),
            false => Err(String::from_utf8_lossy(&writer).into_owned()),
        },
        Err(e) => Err(e.to_string()),
    }
}

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

#[derive(Debug, Deserialize)]
pub struct ChatMessage {
    pub author: ChatAuthor,
    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 author = match message.author {
///                 ChatAuthor::User => "user",
///                 ChatAuthor::Assistant => "assistant",
///             };
///             println!("{}: {}", author, 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!(
        "{}/{}/{}/vertex/chat_history?conversation={}&limit={}",
        crate::GOOGLE_CLOUD_SERVICE_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,
    }
}