Configuration

Configuration section: vector (in application.yaml) or env prefix LEX_VECTOR__.

The provider’s config_key = "vector". The orchestrator reads the vector: section from YAML and coerces it into VectorConfig.

---

VectorConfig

Full example:

vector:
  enabled: true
  backend: qdrant
  default_distance_metric: cosine
  default_index_type: hnsw
  default_dimension: 1536
  upsert_batch_size: 100
  max_retries: 3
  retry_delay: 0.5
  collection_name: default
  enable_cache: false
  cache_ttl: 86400
  embedding_model: text-embedding-3-small
  pgvector:
    database: primary
    schema: public
  qdrant:
    url: http://localhost:6333
    prefer_grpc: true
  pinecone:
    environment: us-west1-gcp
    index_name: my_index
  backends:
    - name: primary
      primary: true
      backend: qdrant
      qdrant:
        url: http://qdrant:6333
    - name: archive
      backend: pgvector
      pgvector:
        database: archive

Env-var override form:

export LEX_VECTOR__BACKEND=pgvector
export LEX_VECTOR__DEFAULT_DIMENSION=768
export LEX_VECTOR__PGVECTOR__DATABASE=primary

Top-Level Fields

Key	Type	Default	Env Var	Description
enabled	bool	True	LEX_VECTOR__ENABLED	Enable vector subsystem
backend	str	"memory"	LEX_VECTOR__BACKEND	Driver: memory, pgvector, qdrant, pinecone, chroma, weaviate
default_distance_metric	DistanceMetric	"cosine"	LEX_VECTOR__DEFAULT_DISTANCE_METRIC	cosine, euclidean, dot_product
default_index_type	IndexType	"hnsw"	LEX_VECTOR__DEFAULT_INDEX_TYPE	hnsw, ivfflat
default_dimension	int	1536	LEX_VECTOR__DEFAULT_DIMENSION	Default vector dimension
upsert_batch_size	int	100	LEX_VECTOR__UPSERT_BATCH_SIZE	Vectors per upsert batch
max_retries	int	3	LEX_VECTOR__MAX_RETRIES	Operation retries
retry_delay	float	0.5	LEX_VECTOR__RETRY_DELAY	Delay between retries (seconds)
collection_name	str	"default"	LEX_VECTOR__COLLECTION_NAME	Default collection for AI-layer ops
enable_cache	bool	False	LEX_VECTOR__ENABLE_CACHE	Embedding cache (requires CacheBackendProtocol)
cache_ttl	int	86400	LEX_VECTOR__CACHE_TTL	Cache TTL (seconds)
embedding_model	str	"text-embedding-3-small"	LEX_VECTOR__EMBEDDING_MODEL	Model name for AI-layer embedding

Driver-Specific Configs

PgVectorConfig (pgvector.)

Key	Type	Default	Description
database	str	"primary"	Database backend name from db.backends
schema	str	"public"	Database schema for vector tables
default_lists	int	100	Lists for IVFFlat index
default_probes	int	10	Probes for IVFFlat index
default_ef_search	int	64	ef_search for HNSW index
table_prefix	str	"vec_"	Table name prefix
create_extension	bool	True	Auto-create pgvector extension

QdrantConfig (qdrant.)

Key	Type	Default	Description
url	str	"http://localhost:6333"	Server URL
api_key	SecretStr | None	None	API key
grpc_port	int	6334	gRPC port
prefer_grpc	bool	True	Prefer gRPC over HTTP
timeout	float	30.0	Request timeout

PineconeConfig (pinecone.)

Key	Type	Default	Description
api_key	SecretStr	""	Pinecone API key
environment	str	""	Environment (e.g. us-west1-gcp)
index_name	str	""	Index name
namespace	str	""	Default namespace
timeout	float	30.0	Request timeout
pool_threads	int	4	Connection pool threads

ChromaConfig (chroma.)

Key	Type	Default	Description
host	str	"localhost"	Server host
port	int	8000	Server port
use_http_client	bool	True	False = ephemeral in-memory client
api_key	SecretStr | None	None	API key
collection_name	str	"default"	Default collection
timeout	float	30.0	Request timeout

WeaviateConfig (weaviate.)

Key	Type	Default	Description
url	str	"http://localhost:8080"	Cluster URL
api_key	SecretStr | None	None	API key
grpc_port	int	50051	gRPC port
timeout	float	30.0	Request timeout

MemoryConfig (memory.)

Key	Type	Default	Description
max_collections	int	100	Max in-memory collections
max_vectors_per_collection	int	100000	Max vectors per collection

---

Multi-Backend Config

vector:
  backends:
    - name: primary
      primary: true
      backend: qdrant
      qdrant:
        url: http://qdrant:6333
    - name: archive
      backend: pgvector
      pgvector:
        database: archive_db

Env-var form for named backends uses indexed arrays (not recommended for complex configs — use YAML).

---

Environment Variables

All keys map from env vars using LEX_VECTOR__ prefix with __ as nested delimiter:

export LEX_VECTOR__BACKEND=pgvector
export LEX_VECTOR__DEFAULT_DIMENSION=1536
export LEX_VECTOR__PGVECTOR__DATABASE=primary
export LEX_VECTOR__QDRANT__URL=http://qdrant:6333
export LEX_VECTOR__ENABLE_CACHE=true
export LEX_VECTOR__EMBEDDING_MODEL=text-embedding-3-small

Embedding client config uses a separate prefix: LEX_VECTOR__EMBEDDING__.