Documentation
¶
Overview ¶
Package sqlite is a persistent backend: one SQLite database file holding all collections, documents, chunks, and vectors. A single Store exposes the three persistence ports through Collections/Documents/Vectors accessors — Go forbids one type from having several same-named methods (Upsert, Delete), so the ports are separate types sharing the *sql.DB. Vector search is brute-force cosine in Go over stored BLOBs. The driver is pure-Go (modernc.org/sqlite): no cgo, so the static-binary and cross-compile goals hold.
Index ¶
- type AnswerCache
- type CollectionRepository
- func (r *CollectionRepository) Create(ctx context.Context, c *domain.Collection) error
- func (r *CollectionRepository) Delete(ctx context.Context, name string) error
- func (r *CollectionRepository) Get(ctx context.Context, name string) (*domain.Collection, error)
- func (r *CollectionRepository) List(ctx context.Context) ([]*domain.Collection, error)
- func (r *CollectionRepository) RecordSource(ctx context.Context, name, source string) error
- type DocumentRepository
- func (r *DocumentRepository) Delete(ctx context.Context, collection string, id domain.DocumentID) ([]domain.ChunkID, error)
- func (r *DocumentRepository) DeleteChunks(ctx context.Context, collection string, ids []domain.ChunkID) ([]domain.ChunkID, error)
- func (r *DocumentRepository) DeleteCollection(ctx context.Context, collection string) ([]domain.ChunkID, error)
- func (r *DocumentRepository) GetBySource(ctx context.Context, collection, sourceURI string) (*domain.Document, error)
- func (r *DocumentRepository) GetChunks(ctx context.Context, ids []domain.ChunkID) ([]domain.Chunk, error)
- func (r *DocumentRepository) GetChunksByDocument(ctx context.Context, collection string, id domain.DocumentID) ([]domain.Chunk, error)
- func (r *DocumentRepository) GetChunksByIDs(ctx context.Context, collection string, ids []string) ([]domain.Chunk, error)
- func (r *DocumentRepository) GetDocuments(ctx context.Context, ids []domain.DocumentID) ([]*domain.Document, error)
- func (r *DocumentRepository) ListDocuments(ctx context.Context, collection string) ([]*domain.Document, error)
- func (r *DocumentRepository) Upsert(ctx context.Context, doc *domain.Document, chunks []domain.Chunk) error
- type LexicalIndex
- func (x *LexicalIndex) Delete(ctx context.Context, collection string, ids []domain.ChunkID) error
- func (x *LexicalIndex) Search(ctx context.Context, collection, query string, k int, filter domain.Predicate) ([]domain.ChunkID, error)
- func (x *LexicalIndex) Upsert(ctx context.Context, collection string, docs []app.LexicalDoc) error
- type Store
- type VectorIndex
- func (x *VectorIndex) Delete(ctx context.Context, collection string, ids []domain.ChunkID) error
- func (x *VectorIndex) Entries(ctx context.Context, collection string) ([]app.VectorEntry, error)
- func (x *VectorIndex) Search(ctx context.Context, collection string, query []float32, k int, ...) ([]domain.VectorMatch, error)
- func (x *VectorIndex) Upsert(ctx context.Context, collection string, entries []app.VectorEntry) error
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type AnswerCache ¶
type AnswerCache struct {
// contains filtered or unexported fields
}
AnswerCache is the SQLite-backed app.AnswerCache: synthesized answers keyed by an opaque content hash, surviving across processes (the point of caching for a CLI). The answer is stored as JSON; timestamps are RFC3339Nano text so the stored_at index orders correctly.
func (*AnswerCache) Get ¶
func (c *AnswerCache) Get(ctx context.Context, key string, notBefore time.Time) (app.Answer, bool, error)
Get returns the cached answer for key when present and stored at or after notBefore; otherwise ok is false.
type CollectionRepository ¶
type CollectionRepository struct {
// contains filtered or unexported fields
}
CollectionRepository is the SQLite-backed CollectionRepository.
func (*CollectionRepository) Create ¶
func (r *CollectionRepository) Create(ctx context.Context, c *domain.Collection) error
Create inserts the collection, failing with ErrAlreadyExists if the name is taken. The INSERT ... ON CONFLICT DO NOTHING + RowsAffected check is atomic.
func (*CollectionRepository) Delete ¶
func (r *CollectionRepository) Delete(ctx context.Context, name string) error
Delete removes the collection record and its recorded sources, or fails with ErrNotFound. The cross-aggregate cascade (documents and vectors) is the use case's job.
func (*CollectionRepository) Get ¶
func (r *CollectionRepository) Get(ctx context.Context, name string) (*domain.Collection, error)
Get returns the named collection (with its recorded sources), or ErrNotFound.
func (*CollectionRepository) List ¶
func (r *CollectionRepository) List(ctx context.Context) ([]*domain.Collection, error)
List returns every collection (with its recorded sources), in unspecified order.
func (*CollectionRepository) RecordSource ¶
func (r *CollectionRepository) RecordSource(ctx context.Context, name, source string) error
RecordSource adds source to the collection's recorded sources (idempotent via the composite primary key), or fails with ErrNotFound if no such collection.
type DocumentRepository ¶
type DocumentRepository struct {
// contains filtered or unexported fields
}
DocumentRepository is the SQLite-backed DocumentRepository.
func (*DocumentRepository) Delete ¶
func (r *DocumentRepository) Delete(ctx context.Context, collection string, id domain.DocumentID) ([]domain.ChunkID, error)
Delete removes the document and its chunks, returning the removed chunk IDs, or fails with ErrNotFound.
func (*DocumentRepository) DeleteChunks ¶
func (r *DocumentRepository) DeleteChunks(ctx context.Context, collection string, ids []domain.ChunkID) ([]domain.ChunkID, error)
DeleteChunks removes the given chunks that belong to the collection (their documents are left in place), returning the IDs actually removed. IDs absent from the collection are skipped. Resolving which IDs belong to the collection before deleting keeps the returned set exact — the vector cascade needs it.
func (*DocumentRepository) DeleteCollection ¶
func (r *DocumentRepository) DeleteCollection(ctx context.Context, collection string) ([]domain.ChunkID, error)
DeleteCollection removes every document and its chunks in the collection, returning all removed chunk IDs. An empty collection is a no-op.
func (*DocumentRepository) GetBySource ¶
func (r *DocumentRepository) GetBySource(ctx context.Context, collection, sourceURI string) (*domain.Document, error)
GetBySource returns the document for (collection, sourceURI), or ErrNotFound.
func (*DocumentRepository) GetChunks ¶
func (r *DocumentRepository) GetChunks(ctx context.Context, ids []domain.ChunkID) ([]domain.Chunk, error)
GetChunks hydrates chunks by ID, preserving input order and skipping IDs with no stored chunk.
func (*DocumentRepository) GetChunksByDocument ¶
func (r *DocumentRepository) GetChunksByDocument(ctx context.Context, collection string, id domain.DocumentID) ([]domain.Chunk, error)
GetChunksByDocument returns all chunks of one document in seq order. An unknown document (or collection) yields no chunks and no error.
func (*DocumentRepository) GetChunksByIDs ¶
func (r *DocumentRepository) GetChunksByIDs(ctx context.Context, collection string, ids []string) ([]domain.Chunk, error)
GetChunksByIDs returns the chunks with the given IDs that belong to the collection, in input order. IDs absent from the collection (unknown, or owned by another collection) are omitted. An unknown collection yields no chunks.
func (*DocumentRepository) GetDocuments ¶
func (r *DocumentRepository) GetDocuments(ctx context.Context, ids []domain.DocumentID) ([]*domain.Document, error)
GetDocuments hydrates documents by ID, preserving input order and skipping IDs with no stored document.
func (*DocumentRepository) ListDocuments ¶
func (r *DocumentRepository) ListDocuments(ctx context.Context, collection string) ([]*domain.Document, error)
ListDocuments returns every document in the collection, ordered by source URI. An unknown or empty collection yields no documents and no error.
type LexicalIndex ¶ added in v1.0.0
type LexicalIndex struct {
// contains filtered or unexported fields
}
LexicalIndex is the SQLite FTS5-backed LexicalIndex: chunk text is indexed in an FTS5 virtual table and ranked by its built-in BM25. The metadata predicate is evaluated in Go after the MATCH (the same single-source-of-truth choice the vector index makes), so it cannot drift from the memstore reference.
func (*LexicalIndex) Delete ¶ added in v1.0.0
Delete removes the given chunk IDs from the collection; absent IDs are a no-op.
func (*LexicalIndex) Search ¶ added in v1.0.0
func (x *LexicalIndex) Search(ctx context.Context, collection, query string, k int, filter domain.Predicate) ([]domain.ChunkID, error)
Search returns up to k chunk IDs whose content matches any query term, ranked by FTS5 BM25 (best first), filtered to those whose metadata satisfies filter.
func (*LexicalIndex) Upsert ¶ added in v1.0.0
func (x *LexicalIndex) Upsert(ctx context.Context, collection string, docs []app.LexicalDoc) error
Upsert replaces each chunk's lexical content. FTS5 has no UPSERT, so each row is deleted then inserted within one transaction.
type Store ¶
type Store struct {
// contains filtered or unexported fields
}
Store is a SQLite-backed persistence engine.
func Open ¶
Open opens (creating if needed) the database at path and applies the schema. Pass ":memory:" for an ephemeral store. The pool is capped to one connection: SQLite is single-writer, and an in-memory database lives on its connection.
func (*Store) Cache ¶
func (s *Store) Cache() *AnswerCache
Cache returns the AnswerCache view of the store.
func (*Store) Collections ¶
func (s *Store) Collections() *CollectionRepository
Collections returns the CollectionRepository view of the store.
func (*Store) Documents ¶
func (s *Store) Documents() *DocumentRepository
Documents returns the DocumentRepository view of the store.
func (*Store) Lexical ¶ added in v1.0.0
func (s *Store) Lexical() *LexicalIndex
Lexical returns the LexicalIndex (FTS5) view of the store.
func (*Store) Vectors ¶
func (s *Store) Vectors() *VectorIndex
Vectors returns the VectorIndex view of the store.
type VectorIndex ¶
type VectorIndex struct {
// contains filtered or unexported fields
}
VectorIndex is the SQLite-backed VectorIndex: vectors are stored as BLOBs and searched by brute-force cosine in Go.
func (*VectorIndex) Delete ¶
Delete removes the given chunk IDs from the collection; absent IDs are a no-op.
func (*VectorIndex) Entries ¶
func (x *VectorIndex) Entries(ctx context.Context, collection string) ([]app.VectorEntry, error)
Entries returns every stored (chunk_id, vector, metadata) for the collection. An unknown collection yields no entries and no error.
func (*VectorIndex) Search ¶
func (x *VectorIndex) Search(ctx context.Context, collection string, query []float32, k int, filter domain.Predicate) ([]domain.VectorMatch, error)
Search returns up to k matches by cosine similarity, best first, considering only rows whose metadata satisfies filter. The predicate is evaluated in Go (in the brute-force scan the index already runs), not translated to SQL, so the domain evaluator is the single source of truth for filter semantics and memstore and sqlite cannot diverge. The zero predicate skips the metadata decode entirely.
func (*VectorIndex) Upsert ¶
func (x *VectorIndex) Upsert(ctx context.Context, collection string, entries []app.VectorEntry) error
Upsert stores the vectors, replacing entries with the same ChunkID.
Source Files