soev-indexingReference
Commands
Index a file into a pgvector collection.
soev-index
soev-index [-h] {index} ...soev-index index
Parse, chunk, embed and upsert a document into pgvector
soev-index index [-h] --collection COLLECTION [--subject SUBJECT]
[--config-root CONFIG_ROOT] [--postgres-uri POSTGRES_URI]
[--schema SCHEMA] [--embed-model EMBED_MODEL]
[--pdf-backend {pymupdf,liteparse}]
path| Argument | Default | Description |
|---|---|---|
path | required | the PDF, text or delimited file to index |
--collection | required | the collection to upsert the chunks into |
--subject | key-holder | the subject that owns the collection. soev-server shows it to requests without X-Soev-Subject |
--config-root | the inference config root. Required, or set $SOEV_SERVER_INFERENCE_CONFIG_ROOT | |
--postgres-uri | the URI of the Postgres database. Required, or set $SOEV_SERVER_POSTGRES_URI | |
--schema | soev_indexing | the schema of the chunk table. $SOEV_SERVER_INDEXING_SCHEMA overrides the default |
--embed-model | local--embedding | the inference endpoint that embeds the chunks. $SOEV_SERVER_EMBED_MODEL overrides the default |
--pdf-backend | pymupdf | the parser for PDF files. One of pymupdf, liteparse. |