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

use serde::{Deserialize, Serialize};

use crate::{Prompt, Task};

/// Input for a [crate::Client::explanation] request.
pub struct TaskExplanation<'a> {
    /// The prompt that typically was the input of a previous completion request
    pub prompt: Prompt<'a>,
    /// The target string that should be explained. The influence of individual parts
    /// of the prompt for generating this target string will be indicated in the response.
    pub target: &'a str,
    /// Granularity paramaters for the explanation
    pub granularity: Granularity,
}

/// Granularity parameters for the [TaskExplanation]
#[derive(Default)]
pub struct Granularity {
    /// The granularity of the parts of the prompt for which a single
    /// score is computed.
    prompt: PromptGranularity,
}

impl Granularity {
    /// Returns a new [Granularity] based on the given one with the [Granularity::prompt]
    /// being set to `prompt_granularity`.
    pub fn with_prompt_granularity(self, prompt_granularity: PromptGranularity) -> Self {
        Self {
            prompt: prompt_granularity,
        }
    }
}

/// At which granularity should the target be explained in terms of the prompt.
/// If you choose, for example, [PromptGranularity::Sentence] then we report the importance score of each
/// sentence in the prompt towards generating the target output.
/// The default is [PromptGranularity::Auto] which means we will try to find the granularity that
/// brings you closest to around 30 explanations. For large prompts, this would likely
/// be sentences. For short prompts this might be individual words or even tokens.
#[derive(Serialize, Copy, Clone, PartialEq, Default)]
#[serde(tag = "type", rename_all = "snake_case")]
pub enum PromptGranularity {
    /// Let the system decide which granularity is most suitable for the given input.
    /// This will result
    #[default]
    Auto,
    Word,
    Sentence,
    Paragraph,
}

impl PromptGranularity {
    fn is_auto(&self) -> bool {
        self == &PromptGranularity::Auto
    }
}

/// Body sent to the Aleph Alpha API for an explanation request
#[derive(Serialize)]
struct BodyExplanation<'a> {
    prompt: Prompt<'a>,
    target: &'a str,
    #[serde(skip_serializing_if = "PromptGranularity::is_auto")]
    prompt_granularity: PromptGranularity,
    model: &'a str,
}

/// Body received by the Aleph Alpha API from an explanation request
#[derive(Deserialize, Debug, PartialEq)]
pub struct ResponseExplanation {
    /// The Body contains an array of [Explanation]s one for each
    /// part of the target being explained.
    explanations: Vec<Explanation>,
}

/// The result of an explanation request.
#[derive(Debug, PartialEq)]
pub struct ExplanationOutput {
    /// Explanation scores for different parts of the prompt or target.
    pub items: Vec<ItemExplanation>,
}

impl ExplanationOutput {
    fn from(mut response: ResponseExplanation) -> ExplanationOutput {
        ExplanationOutput {
            items: response.explanations.pop().unwrap().items,
        }
    }
}

/// The explanation for the target.
#[derive(Debug, Deserialize, PartialEq)]
pub struct Explanation {
    /// Explanation scores for different parts of the prompt or target.
    pub items: Vec<ItemExplanation>,
}

/// Explanation scores for a [crate::prompt::Modality] or the target.
/// There is one score
/// for each part of a `modality` respectively the target with the parts being choosen according to
/// the [PromptGranularity]
#[derive(PartialEq, Deserialize, Debug)]
#[serde(tag = "type", rename_all = "snake_case")]
pub enum ItemExplanation {
    Text { scores: Vec<TextScore> },
    Image { scores: Vec<ImageScore> },
    Target { scores: Vec<TextScore> },
}

/// Score for the part of a text-modality
#[derive(Debug, PartialEq, Deserialize, Clone)]
pub struct TextScore {
    pub start: u32,
    pub length: u32,
    pub score: f32,
}

/// Resembles the actual response from the API for deserialization
#[derive(Deserialize)]
struct BoundingBox {
    top: f32,
    left: f32,
    width: f32,
    height: f32,
}

/// Resembles the actual response from the API for deserialization
#[derive(Deserialize)]
struct ImageScoreWithRect {
    rect: BoundingBox,
    score: f32,
}

/// Score for a part of an image.
#[derive(Debug, PartialEq, Deserialize, Clone)]
#[serde(from = "ImageScoreWithRect")]
pub struct ImageScore {
    pub left: f32,
    pub top: f32,
    pub width: f32,
    pub height: f32,
    pub score: f32,
}

impl From<ImageScoreWithRect> for ImageScore {
    fn from(value: ImageScoreWithRect) -> Self {
        Self {
            left: value.rect.left,
            top: value.rect.top,
            width: value.rect.width,
            height: value.rect.height,
            score: value.score,
        }
    }
}

impl Task for TaskExplanation<'_> {
    type Output = ExplanationOutput;

    type ResponseBody = ResponseExplanation;

    fn build_request(
        &self,
        client: &reqwest::Client,
        base: &str,
        model: &str,
    ) -> reqwest::RequestBuilder {
        let body = BodyExplanation {
            model,
            prompt: self.prompt.borrow(),
            target: self.target,
            prompt_granularity: self.granularity.prompt,
        };
        client.post(format!("{base}/explain")).json(&body)
    }

    fn body_to_output(&self, response: Self::ResponseBody) -> Self::Output {
        ExplanationOutput::from(response)
    }
}