cognis-rag 0.3.1

RAG primitives for Cognis: embeddings, vector stores (in-memory, FAISS, Chroma, Qdrant, Pinecone, Weaviate), retrievers, text splitters, document loaders, and incremental indexing pipelines.
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

//! Embeddings trait + implementations.

use async_trait::async_trait;

use cognis_core::{CognisError, Result};

mod fake;
pub use fake::FakeEmbeddings;

mod batched;
mod cached;
mod router;
pub use batched::BatchedEmbeddings;
pub use cached::CachedEmbeddings;
pub use router::{EmbeddingRouter, EmbeddingsRouter, FnRouter, LengthRouter};

#[cfg(feature = "openai")]
pub mod openai;
#[cfg(feature = "openai")]
pub use openai::OpenAIEmbeddings;

#[cfg(feature = "ollama")]
pub mod ollama;
#[cfg(feature = "ollama")]
pub use ollama::OllamaEmbeddings;

#[cfg(feature = "google")]
pub mod google;
#[cfg(feature = "google")]
pub use google::{GoogleEmbeddings, GoogleEmbeddingsBuilder};

#[cfg(feature = "voyage")]
pub mod voyage;
#[cfg(feature = "voyage")]
pub use voyage::{VoyageEmbeddings, VoyageEmbeddingsBuilder};

/// A model that converts text into vector embeddings. Implementations
/// MUST return vectors of consistent dimensionality across calls — the
/// `dimensions()` accessor (when known) reports it.
#[async_trait]
pub trait Embeddings: Send + Sync {
    /// Embed a batch of documents. Order of returned vectors matches
    /// the input order.
    async fn embed_documents(&self, texts: Vec<String>) -> Result<Vec<Vec<f32>>>;

    /// Embed a single query string. Default impl wraps `embed_documents`
    /// with a single-element batch.
    async fn embed_query(&self, text: String) -> Result<Vec<f32>> {
        let mut vecs = self.embed_documents(vec![text]).await?;
        vecs.pop()
            .ok_or_else(|| CognisError::Internal("embed_documents returned empty vec".into()))
    }

    /// Dimensionality of the embedding vectors, if known statically.
    fn dimensions(&self) -> Option<usize> {
        None
    }

    /// Model identifier (e.g. "text-embedding-3-small", "nomic-embed-text").
    fn model(&self) -> &str;
}