markhor_core 0.1.0-alpha.0.2

Core library for Markhor, a project connecting AI models, documents, and workflows for knowledge work
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

mod error;
mod embedder;
mod vector_store;

pub use error::{EmbeddingError};
pub use embedder::{Embedder, EmbeddingUseCase};
pub use vector_store::{VectorStore, ChunkDataResult};

use serde::{Deserialize, Serialize};

/// Represents an embedding vector.
///
/// This struct simply wraps a `Vec<f32>`.
#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
pub struct Embedding(pub Vec<f32>);

// Allow easy conversion from the raw Vec<Vec<f32>> for implementers.
impl From<Vec<f32>> for Embedding {
    fn from(vec: Vec<f32>) -> Self {
        Embedding(vec)
    }
}

impl Embedding {
    /// Computes the cosine similarity between two embeddings.
    ///
    /// Returns an error if the vectors are empty or have mismatched lengths.
    pub fn similarity(&self, other: &Embedding) -> Result<f32, EmbeddingError> {
        if self.0.is_empty() || other.0.is_empty() {
            // TODO: fix errors
            //return Err(EmbeddingError::ZeroLength);
            panic!("zero length vector");
        }

        if self.0.len() != other.0.len() {
            //return Err(EmbeddingError::MismatchedLengths);
            panic!("mismatched vector lengths");
        }

        let dot_product: f32 = self.0.iter().zip(&other.0).map(|(a, b)| a * b).sum();
        let norm_self: f32 = self.0.iter().map(|x| x * x).sum::<f32>().sqrt();
        let norm_other: f32 = other.0.iter().map(|x| x * x).sum::<f32>().sqrt();

        if norm_self == 0.0 || norm_other == 0.0 {
            //return Err(EmbeddingError::ZeroLength);
            panic!("zero length vector");
        }

        Ok(dot_product / (norm_self * norm_other))
    }
}