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

use std::pin::Pin;

use async_trait::async_trait;
use futures::Stream;
use serde_json::Value;

use crate::{
    chain::{
        load_stuff_qa, options::ChainCallOptions, Chain, ChainError, LLMChain, StuffQAPromptBuilder,
    },
    language_models::{llm::LLM, GenerateResult},
    prompt::PromptArgs,
    schemas::{Document, StreamData},
};

const COMBINE_DOCUMENTS_DEFAULT_INPUT_KEY: &str = "input_documents";
const COMBINE_DOCUMENTS_DEFAULT_OUTPUT_KEY: &str = "text";
const COMBINE_DOCUMENTS_DEFAULT_DOCUMENT_VARIABLE_NAME: &str = "context";
const STUFF_DOCUMENTS_DEFAULT_SEPARATOR: &str = "\n\n";

pub struct StuffDocument {
    llm_chain: LLMChain,
    input_key: String,
    document_variable_name: String,
    separator: String,
}

impl StuffDocument {
    pub fn new(llm_chain: LLMChain) -> Self {
        Self {
            llm_chain,
            input_key: COMBINE_DOCUMENTS_DEFAULT_INPUT_KEY.to_string(),
            document_variable_name: COMBINE_DOCUMENTS_DEFAULT_DOCUMENT_VARIABLE_NAME.to_string(),
            separator: STUFF_DOCUMENTS_DEFAULT_SEPARATOR.to_string(),
        }
    }

    fn join_documents(&self, docs: Vec<Document>) -> String {
        docs.iter()
            .map(|doc| doc.page_content.clone())
            .collect::<Vec<_>>()
            .join(&self.separator)
    }

    pub fn qa_prompt_builder<'a>(&self) -> StuffQAPromptBuilder<'a> {
        StuffQAPromptBuilder::new()
    }

    /// load_stuff_qa return an instance of StuffDocument
    /// with a prompt desiged for question ansering
    ///
    /// # Example
    /// ```rust,ignore
    ///
    /// let llm = OpenAI::default();
    /// let chain = StuffDocument::load_stuff_qa(llm);
    ///
    /// let input = chain
    /// .qa_prompt_builder()
    /// .documents(&[
    /// Document::new(format!(
    /// "\nQuestion: {}\nAnswer: {}\n",
    /// "Which is the favorite text editor of luis", "Nvim"
    /// )),
    /// Document::new(format!(
    /// "\nQuestion: {}\nAnswer: {}\n",
    /// "How old is Luis", "24"
    /// )),
    /// ])
    /// .question("How old is luis and whats his favorite text editor")
    /// .build();
    ///
    /// let ouput = chain.invoke(input).await.unwrap();
    ///
    /// println!("{}", ouput);
    /// ```
    ///
    pub fn load_stuff_qa<L: LLM + 'static>(llm: L) -> Self {
        load_stuff_qa(llm, None)
    }

    /// load_stuff_qa_with_options return an instance of StuffDocument
    /// with a prompt desiged for question ansering
    ///
    /// # Example
    /// ```rust,ignore
    ///
    /// let llm = OpenAI::default();
    /// let chain = StuffDocument::load_stuff_qa_with_options(llm,ChainCallOptions::default());
    ///
    /// let input = chain
    /// .qa_prompt_builder()
    /// .documents(&[
    /// Document::new(format!(
    /// "\nQuestion: {}\nAnswer: {}\n",
    /// "Which is the favorite text editor of luis", "Nvim"
    /// )),
    /// Document::new(format!(
    /// "\nQuestion: {}\nAnswer: {}\n",
    /// "How old is Luis", "24"
    /// )),
    /// ])
    /// .question("How old is luis and whats his favorite text editor")
    /// .build();
    ///
    /// let ouput = chain.invoke(input).await.unwrap();
    ///
    /// println!("{}", ouput);
    /// ```
    ///
    pub fn load_stuff_qa_with_options<L: LLM + 'static>(llm: L, opt: ChainCallOptions) -> Self {
        load_stuff_qa(llm, Some(opt))
    }
}

#[async_trait]
impl Chain for StuffDocument {
    async fn call(&self, input_variables: PromptArgs) -> Result<GenerateResult, ChainError> {
        let docs = input_variables
            .get(&self.input_key)
            .ok_or_else(|| ChainError::MissingInputVariable(self.input_key.clone()))?;

        let documents: Vec<Document> = serde_json::from_value(docs.clone()).map_err(|e| {
            ChainError::IncorrectInputVariable {
                source: e,
                expected_type: "Vec<Document>".to_string(),
            }
        })?;

        let mut input_values = input_variables.clone();
        input_values.insert(
            self.document_variable_name.clone(),
            Value::String(self.join_documents(documents)),
        );

        self.llm_chain.call(input_values).await
    }

    async fn stream(
        &self,
        input_variables: PromptArgs,
    ) -> Result<Pin<Box<dyn Stream<Item = Result<StreamData, ChainError>> + Send>>, ChainError>
    {
        self.llm_chain.stream(input_variables).await
    }

    fn get_input_keys(&self) -> Vec<String> {
        vec![self.input_key.clone()]
    }
}