Skip to main content

graphrt_rag/
lib.rs

1//! # graphrt-rag
2//!
3//! Retrieval building blocks for graphrt workflows and the `knowledge-retrieval` node.
4//!
5//! - [`Embedder`], [`VectorStore`], [`Reranker`], [`Retriever`]: pluggable traits.
6//! - [`MemoryStore`]: in-process cosine + BM25 store (optionally file-backed).
7//! - [`SqliteStore`] (feature `sqlite`): the same search over one SQLite file,
8//!   with incremental writes and several processes sharing the database.
9//! - [`OpenAiEmbedder`], [`HttpReranker`], [`HttpRetriever`] (feature `http`) and
10//!   [`QdrantStore`] (feature `qdrant`).
11//! - [`LocalEmbedder`] (feature `local-embed`): sentence embeddings computed in
12//!   this process with candle, no embedding server or API key.
13//! - [`chunk_text`] and [`Ingestor`] for indexing documents.
14//! - [`KnowledgeBase`]: the service the node uses, built in code or from a
15//!   [`KnowledgeConfig`] file.
16//!
17//! ```
18//! use std::sync::Arc;
19//! use graphrt_core::{Registry, Services};
20//! use graphrt_rag::{Document, HashEmbedder, KnowledgeBase, MemoryStore};
21//!
22//! # futures::executor::block_on(async {
23//! let kb = KnowledgeBase::new()
24//!     .with_store_dataset("faq", Arc::new(MemoryStore::new()), Arc::new(HashEmbedder::new(256)), None);
25//! kb.ingestor("faq").unwrap().ingest(&Document::new("faq", "returns", "Items can be returned within 30 days.")).await.unwrap();
26//!
27//! let mut registry = Registry::with_builtins();
28//! graphrt_rag::register(&mut registry);
29//! let services = Services::new().with(kb);
30//! # let _ = services;
31//! # });
32//! ```
33
34mod chunker;
35mod error;
36mod filter;
37mod hash;
38#[cfg(feature = "http")]
39mod http;
40mod ingest;
41mod kb;
42#[cfg(feature = "local-embed")]
43mod local_embed;
44mod memory;
45mod node;
46mod pipeline;
47#[cfg(feature = "qdrant")]
48mod qdrant;
49#[cfg(feature = "sqlite")]
50mod sqlite;
51mod types;
52
53#[cfg(feature = "doc")]
54mod doc_extractor;
55pub use chunker::{ChunkerConfig, TextChunk, chunk_text};
56#[cfg(feature = "doc")]
57pub use doc_extractor::DocExtractor;
58pub use error::RagError;
59pub use filter::{Logical, MetadataCondition, MetadataFilter, UnknownOperator};
60pub use hash::HashEmbedder;
61#[cfg(feature = "http")]
62pub use http::{
63    HttpReranker, HttpRerankerConfig, HttpRetriever, HttpRetrieverConfig, HttpSettings, OpenAiEmbedder,
64    OpenAiEmbedderConfig, Secret,
65};
66pub use ingest::{Document, IngestReport, Ingestor};
67pub use kb::{
68    DatasetConfig, DatasetDetails, DatasetInfo, DatasetSpec, EnvLookup, KnowledgeBase, KnowledgeConfig, ResourceNames,
69};
70#[cfg(feature = "local-embed")]
71pub use local_embed::{DEFAULT_LOCAL_MODEL, LocalEmbedder, LocalEmbedderConfig, Pooling};
72pub use memory::{MemoryStore, cosine, dot, tokenize};
73pub use node::{KIND, KnowledgeRetrievalNode, descriptor, register};
74pub use pipeline::{StoreRetriever, finalize, merge_weighted, rerank};
75#[cfg(feature = "qdrant")]
76pub use qdrant::{QdrantConfig, QdrantStore};
77#[cfg(feature = "sqlite")]
78pub use sqlite::SqliteStore;
79pub use types::{
80    DEFAULT_DOCUMENT_PAGE, DISABLED_KEY, DocumentSummary, Embedder, Extractor, Hit, HybridWeights, KeywordQuery,
81    MAX_DOCUMENT_PAGE, Metadata, PlainTextExtractor, Reranker, RetrievalRequest, Retriever, StoredChunk, VectorQuery,
82    VectorStore, check_vector, content_hash, is_disabled,
83};