llmprism 0.4.0

A unified, fluent Rust interface for integrating Large Language Model providers -- a Rust port of the PHP Prism library.
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

//! Wire format and mapping for VoyageAI's Embeddings API
//! (`POST /v1/embeddings`), one of this provider's two endpoints (the other
//! being its Rerank API) -- frequently used alongside Anthropic, which has
//! no embeddings endpoint of its own.

use serde::{Deserialize, Serialize};

use crate::embeddings::{Embedding, EmbeddingsRequest, EmbeddingsResponse};
use crate::value_objects::{EmbeddingsUsage, Meta};

#[derive(Debug, Serialize)]
pub struct ApiRequest {
    pub input: Vec<String>,
    pub model: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub output_dimension: Option<u32>,
}

#[derive(Debug, Deserialize)]
pub struct ApiResponse {
    pub data: Vec<ApiEmbedding>,
    pub model: String,
    pub usage: ApiUsage,
}

#[derive(Debug, Deserialize)]
pub struct ApiEmbedding {
    /// Which input this embedding corresponds to -- used to put `data` back
    /// into input order in [`parse_response`], the same defensive choice
    /// OpenAI's embeddings mapping makes rather than assuming array order.
    pub index: usize,
    pub embedding: Vec<f32>,
}

#[derive(Debug, Deserialize)]
pub struct ApiUsage {
    /// VoyageAI reports only a single token count for an embeddings call (no
    /// prompt/completion split, since there's no completion) -- this maps
    /// directly onto [`EmbeddingsUsage::prompt_tokens`].
    pub total_tokens: u32,
}

pub fn build_request(request: &EmbeddingsRequest) -> ApiRequest {
    ApiRequest {
        input: request.input.clone(),
        model: request.model.clone(),
        output_dimension: request.dimensions,
    }
}

pub fn parse_response(mut response: ApiResponse) -> EmbeddingsResponse {
    response.data.sort_by_key(|embedding| embedding.index);

    EmbeddingsResponse {
        embeddings: response
            .data
            .into_iter()
            .map(|embedding| Embedding {
                vector: embedding.embedding,
            })
            .collect(),
        usage: EmbeddingsUsage {
            prompt_tokens: response.usage.total_tokens,
        },
        meta: Meta {
            id: None,
            model: Some(response.model),
            rate_limits: Vec::new(),
        },
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn build_request_passes_dimensions_through_as_output_dimension() {
        let mut request = EmbeddingsRequest::new("voyage-3-large");
        request.dimensions = Some(512);
        let wire_request = build_request(&request);
        assert_eq!(wire_request.output_dimension, Some(512));
    }
}