Vector Retrieval¶

Embeddings, vector stores, and hybrid graph+vector retrieval.

Auto-generated docs

When trails is installed, run ENABLE_MKDOCSTRINGS=true ./scripts/docs-build for full docstring-extracted reference.

Retrieval¶

Symbol	Signature	Description
`retrieve`	`retrieve(query: str, *, ctx=None, mode: "graph" \\| "vector" \\| "hybrid" = "hybrid", k: int = 10, sparql_filter: str \\| None = None, vector_store=None, embedder=None) -> list[RetrievalHit]`	Top-level entry point for graph, vector, or hybrid retrieval
`RetrievalHit`	`RetrievalHit(iri: str, score: float, snippet: str \\| None, metadata: dict)`	One retrieval result

Symbol	Signature	Description
`SqliteVecStore`	`SqliteVecStore(dim: int, path: str = ":memory:")`	Zero-ops default store using `sqlite-vec`. Supports `add`, `search`, `delete`, `count`
`QdrantStore`	`QdrantStore(dim: int, collection: str, ...)`	Scale adapter over `qdrant-client`. Same surface as `SqliteVecStore`
`VectorStore`	Protocol	Minimal contract: `.dim`, `.add()`, `.search()`, `.delete()`, `.count()`

Symbol	Signature	Description
`SentenceTransformerEmbedder`	`SentenceTransformerEmbedder(model: str = ...)`	Local CPU-friendly embedder (default). Lazy-imports `sentence-transformers`
`OpenAIEmbedder`	`OpenAIEmbedder(model: str = "text-embedding-3-small", api_key: str \\| None = None)`	Remote embedder for OpenAI `text-embedding-3-*` family
`MockEmbedder`	`MockEmbedder(dim: int = 384, seed: int = 42)`	Deterministic pseudo-random embedder for tests
`EmbeddingProvider`	Protocol	Minimal contract: `.dim`, `.embed(text)`, `.embed_batch(texts)`