Class PostgresOptions

Namespace

Mythosia.VectorDb.Postgres

Assembly

Mythosia.VectorDb.Postgres.dll

Configuration options for PostgresStore.

public class PostgresOptions

Inheritance

object

PostgresOptions

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Properties

ConnectionString

PostgreSQL connection string. Required.

public string ConnectionString { get; set; }

Property Value

string

Dimension

Embedding vector dimension. Must match the dimension used by the embedding provider. Required (must be greater than 0).

public int Dimension { get; set; }

Property Value

int

DistanceStrategy

Distance function for similarity search. Default: Cosine.

public DistanceStrategy DistanceStrategy { get; set; }

Property Value

DistanceStrategy

EnsureSchema

When true, automatically creates the pgvector extension, table, and indexes if they do not exist. Recommended for development/testing only. When false (default), the schema must already exist or an exception is thrown.

public bool EnsureSchema { get; set; }

Property Value

bool

FailFastOnIndexCreationFailure

When true (default), index creation failures throw and fail fast. When false, index creation failures are downgraded to warnings and startup continues.

public bool FailFastOnIndexCreationFailure { get; set; }

Property Value

bool

Index

Vector index settings. Default: HnswIndexOptions.

public VectorIndexOptions Index { get; set; }

Property Value

VectorIndexOptions

SchemaName

Database schema name. Default: "public".

public string SchemaName { get; set; }

Property Value

string

TableName

Table name for vector storage. Default: "vectors".

public string TableName { get; set; }

Property Value

string

TextSearchConfig

PostgreSQL text search configuration used for to_tsvector / plainto_tsquery in hybrid search. Default: "simple". Only used when TextSearchMode is TsVector.

The simple configuration performs whitespace-based tokenization without morphological analysis. This works well for English but may produce poor recall for agglutinative languages such as Korean, Japanese, or Finnish where particles and inflections are attached to words.

To improve hybrid search quality for such languages, install a language-specific text search configuration (e.g. pg_mecab for Korean/Japanese) and set this property accordingly, or use Trigram instead.

public string TextSearchConfig { get; set; }

Property Value

string

TextSearchMode

Text search mode for the keyword leg of hybrid search. Default: TsVector.

TsVector: PostgreSQL full-text search (tsvector / tsquery). Suitable for European languages with good built-in text search configurations.

Trigram: pg_trgm word_similarity matching. Better for CJK languages (Korean, Japanese, Chinese) where PostgreSQL lacks built-in morphological analysis. Requires the pg_trgm extension (standard PostgreSQL contrib module, available on most managed services).

public TextSearchMode TextSearchMode { get; set; }

Property Value

TextSearchMode

Methods

Validate()

Validates the options and throws ArgumentException if invalid.

public void Validate()