langchain_rs 0.0.2

Rust version of Langchain
Documentation 
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

use anyhow::Result;
use async_openai::config::OpenAIConfig;
use async_openai::types::ChatCompletionRequestMessageArgs;
use async_openai::types::CreateChatCompletionRequestArgs;
use async_openai::types::Role;
use async_openai::Client;
use tiktoken_rs::r50k_base;

use crate::schema::GenerateResult;
use crate::schema::TokenUsage;
use crate::schema::LLM;
pub enum OpenAIModel {
    Gpt35,
    Gpt4,
}

impl ToString for OpenAIModel {
    fn to_string(&self) -> String {
        match self {
            OpenAIModel::Gpt35 => "gpt-3.5-turbo".to_string(),
            OpenAIModel::Gpt4 => "gpt-4".to_string(),
        }
    }
}

pub struct OpenAI {
    config: OpenAIConfig,
    model: OpenAIModel,
    max_tokens: u16,
    temperature: f32,
    top_p: f32,
    frequency_penalty: f32,
    presence_penalty: f32,
}

impl Default for OpenAI {
    fn default() -> Self {
        Self {
            config: OpenAIConfig::default(),
            model: OpenAIModel::Gpt35,
            max_tokens: 256,
            temperature: 0.7,
            top_p: 1.0,
            frequency_penalty: 0.0,
            presence_penalty: 0.0,
        }
    }
}

impl OpenAI {
    pub fn new() -> Self {
        Default::default()
    }

    pub fn with_config(config: OpenAIConfig) -> Self {
        Self {
            config,
            ..Default::default()
        }
    }

    pub fn with_model(mut self, model: OpenAIModel) -> Self {
        self.model = model;
        self
    }

    pub fn with_max_tokens(mut self, max_tokens: u16) -> Self {
        self.max_tokens = max_tokens;
        self
    }

    pub fn with_temperature(mut self, temperature: f32) -> Self {
        self.temperature = temperature;
        self
    }

    pub fn with_top_p(mut self, top_p: f32) -> Self {
        self.top_p = top_p;
        self
    }

    pub fn with_frequency_penalty(mut self, frequency_penalty: f32) -> Self {
        self.frequency_penalty = frequency_penalty;
        self
    }

    pub fn with_presence_penalty(mut self, presence_penalty: f32) -> Self {
        self.presence_penalty = presence_penalty;
        self
    }

    pub fn get_client(&self) -> Client<OpenAIConfig> {
        Client::with_config(self.config.clone())
    }
}

#[async_trait::async_trait]
impl LLM for OpenAI {
    async fn generate(&self, text: &str) -> Result<GenerateResult> {
        let request = CreateChatCompletionRequestArgs::default()
            .model(self.model.to_string())
            .max_tokens(self.max_tokens - text.len() as u16)
            .temperature(self.temperature)
            .top_p(self.top_p)
            .frequency_penalty(self.frequency_penalty)
            .presence_penalty(self.presence_penalty)
            .messages([ChatCompletionRequestMessageArgs::default()
                .role(Role::Assistant)
                .content(text)
                .build()?])
            .build()?;

        let client = self.get_client();
        let response = client.chat().create(request).await?;

        let mut generate_result = GenerateResult::default();

        if let Some(usage) = response.usage {
            generate_result.tokens = Some(TokenUsage {
                prompt_tokens: usage.prompt_tokens,
                completion_tokens: usage.completion_tokens,
                total_tokens: usage.total_tokens,
            });
        }

        if let Some(choice) = response.choices.first() {
            generate_result.generation = choice.message.content.clone().unwrap_or_default();
        } else {
            generate_result.generation = "".to_string();
        }

        Ok(generate_result)
    }

    fn tokenize(&self, text: &str) -> Result<Vec<String>> {
        // TODO: select appropriate tokenizer based on model
        let rke = r50k_base()?;
        let splitted_text = rke.split_by_token(text, true)?;
        Ok(splitted_text)
    }
}