Skip to content

Latest commit

 

History

History
27 lines (22 loc) · 3.07 KB

File metadata and controls

27 lines (22 loc) · 3.07 KB

This page documents all node configuration properties related to this extension.

Embedding models

A node can be configured to use multiple embedding models by configuring multiple <model> properties. The <model> is the name to be used on the blockchain, and is used for the node to lookup the model API configuration. This also helps us use different services for different nodes since the model name often is named slightly different.

A blockchain is then using one of the supported models:

vector_db_extension:
  embedding_compute:
    model: "qwen3_embedding_0_6b"
Name Description Type Required Default Environment Variable
extension.vector_db.embedding.<model>.api_type Type of service API, either openai or gcp. String No "openai" POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_API_TYPE
extension.vector_db.embedding.<model>.basic_auth_user Basic http auth username. String No POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_BASIC_AUTH_USER
extension.vector_db.embedding.<model>.basic_auth_password Basic http auth password. String No POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_BASIC_AUTH_PASSWORD
extension.vector_db.embedding.<model>.auth_bearer Token for Authorization: Bearer <token> format. String No POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_AUTH_BEARER
extension.vector_db.embedding.<model>.x_api_key Value for a X-API-Key header (used for GCP). String No POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_X_API_KEY
extension.vector_db.embedding.<model>.model The name to use in request against the model service API. String Yes POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_MODEL
extension.vector_db.embedding.<model>.url The URL of the model service API. String Yes POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_URL
extension.vector_db.embedding.<model>.retry_count How many times to retry embedding HTTP request if it fails . Integer No 2 POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_RETRY_COUNT
extension.vector_db.embedding.<model>.retry_delay How long to wayt between each ettempt in milliseconds. Integer No 1000 POSTCHAIN_EXTENSION_VECTOR_DB_EMBEDDING_<model>_RETRY_DELAY