Documentation
¶
Index ¶
- func CombineQueryWithContext(query string, retrievedDocs []Document) string
- func ConvertEmbeddingToPGVector(embedding []float32) string
- func ConvertMetadata(metadata map[string]string) map[string]interface{}
- type Document
- type PGVector
- func (pg *PGVector) Close()
- func (pg *PGVector) InsertDocument(ctx context.Context, content string, embedding []float32) error
- func (pg *PGVector) QueryRelevantDocuments(ctx context.Context, embedding []float32, backend string) ([]Document, error)
- func (pg *PGVector) SaveEmbeddings(ctx context.Context, docID string, embedding []float32, ...) error
- type QdrantVector
- func (qv *QdrantVector) Close()
- func (qv *QdrantVector) CreateCollection(ctx context.Context, collectionName string, vectorSize uint64, distance string) error
- func (qv *QdrantVector) InsertDocument(ctx context.Context, content string, embedding []float32, collection string) error
- func (qv *QdrantVector) QueryRelevantDocuments(ctx context.Context, embedding []float32, collection string, ...) ([]Document, error)
- func (qv *QdrantVector) SaveEmbeddings(ctx context.Context, docID string, embedding []float32, ...) error
- type QueryOpt
- type VectorDatabase
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func CombineQueryWithContext ¶
CombineQueryWithContext combines the user's query with the relevant retrieved documents' content
func ConvertEmbeddingToPGVector ¶
ConvertEmbeddingToPGVector converts a slice of float32 values representing an embedding into a string format compatible with PostgreSQL's vector type. The resulting string is a comma-separated list of values enclosed in curly braces, with each value formatted to 6 decimal places of precision.
func ConvertMetadata ¶
ConvertMetadata converts a map of string keys and string values to a map of string keys and interface{} values. This is useful when working with metadata that needs to be stored in a more flexible format.
Types ¶
type Document ¶
Document represents a single document in the vector database. It contains a unique identifier and associated metadata.
type PGVector ¶
type PGVector struct {
// contains filtered or unexported fields
}
PGVector represents a connection to a PostgreSQL database with pgvector extension. It provides methods for storing and querying vector embeddings.
func NewPGVector ¶
NewPGVector creates a new PGVector instance with a connection to the PostgreSQL database.
Parameters:
- connString: A string containing the connection details for the PostgreSQL database.
Returns:
- A pointer to a new PGVector instance.
- An error if the connection fails, nil otherwise.
func (*PGVector) InsertDocument ¶
InsertDocument inserts a document into the PGVector store, implementing the VectorDatabase interface.
func (*PGVector) QueryRelevantDocuments ¶
func (pg *PGVector) QueryRelevantDocuments(ctx context.Context, embedding []float32, backend string) ([]Document, error)
QueryRelevantDocuments retrieves the most relevant documents from the database based on the given embedding. It uses cosine similarity to find the closest matches and returns a slice of Document structs.
Parameters:
- ctx: The context for the database query.
- embedding: A slice of float32 values representing the query embedding.
Returns:
- A slice of Document structs containing the most relevant documents.
- An error if the query fails or if there's an issue scanning the results.
func (*PGVector) SaveEmbeddings ¶
func (pg *PGVector) SaveEmbeddings(ctx context.Context, docID string, embedding []float32, metadata map[string]interface{}) error
SaveEmbedding stores a document embedding and associated metadata in the database.
Parameters:
- ctx: The context for the database operation.
- docID: A unique identifier for the document.
- embedding: A slice of float32 values representing the document's embedding.
- metadata: A map of additional information associated with the document.
Returns:
- An error if the saving operation fails, nil otherwise.
SaveEmbeddings stores a document embedding and associated metadata in the PostgreSQL database, implementing the VectorDatabase interface.
type QdrantVector ¶
type QdrantVector struct {
// contains filtered or unexported fields
}
QdrantVector represents a connection to Qdrant.
func NewQdrantVector ¶
func NewQdrantVector(address string, port int) (*QdrantVector, error)
NewQdrantVector initializes a connection to Qdrant.
Parameters:
- address: The Qdrant server address (e.g., "localhost").
- port: The port Qdrant is running on (e.g., 6333).
Returns:
- A pointer to a new QdrantVector instance.
- An error if the connection fails, nil otherwise.
func (*QdrantVector) Close ¶
func (qv *QdrantVector) Close()
Close closes the Qdrant client connection.
func (*QdrantVector) CreateCollection ¶
func (qv *QdrantVector) CreateCollection(ctx context.Context, collectionName string, vectorSize uint64, distance string) error
CreateCollection creates a new collection in Qdrant
func (*QdrantVector) InsertDocument ¶
func (qv *QdrantVector) InsertDocument(ctx context.Context, content string, embedding []float32, collection string) error
InsertDocument inserts a document into the Qdrant vector store.
Parameters:
- ctx: Context for the operation.
- vectorDB: A QdrantVector instance.
- content: The document content to be inserted.
- embedding: The embedding vector for the document.
Returns:
- An error if the operation fails, nil otherwise.
QdrantVector should implement the InsertDocument method as defined in VectorDatabase
func (*QdrantVector) QueryRelevantDocuments ¶
func (qv *QdrantVector) QueryRelevantDocuments( ctx context.Context, embedding []float32, collection string, queryOpts ...QueryOpt, ) ([]Document, error)
QueryRelevantDocuments retrieves the most relevant documents based on a given embedding.
Parameters:
- ctx: The context for the query.
- embedding: The query embedding.
- limit: The number of documents to return.
- collection: The collection name to query.
Returns:
- A slice of QDrantDocument structs containing the most relevant documents.
- An error if the query fails.
func (*QdrantVector) SaveEmbeddings ¶
func (qv *QdrantVector) SaveEmbeddings(ctx context.Context, docID string, embedding []float32, metadata map[string]interface{}, collection string) error
SaveEmbedding stores an embedding and metadata in Qdrant.
Parameters:
- ctx: Context for the operation.
- docID: A unique identifier for the document.
- embedding: A slice of float32 values representing the document's embedding.
- metadata: A map of additional information associated with the document.
Returns:
- An error if the saving operation fails, nil otherwise.
SaveEmbeddings stores an embedding and metadata in Qdrant, implementing the VectorDatabase interface.
type QueryOpt ¶
type QueryOpt func(*qdrant.QueryPoints)
QueryOpt represents an option for a query. This is the type that should be returned from query options functions.
func WithScoreThreshold ¶
WithScoreThreshold sets the score threshold for a query. The higher the threshold, the more relevant the results.
type VectorDatabase ¶
type VectorDatabase interface {
InsertDocument(ctx context.Context, content string, embedding []float32) error
QueryRelevantDocuments(ctx context.Context, embedding []float32, backend string) ([]Document, error)
SaveEmbeddings(ctx context.Context, docID string, embedding []float32, metadata map[string]interface{}) error
}
VectorDatabase is the interface that both QdrantVector and PGVector implement
Source Files