Configuration¶

All parameters for FoxNoseRetriever.

Client¶

Parameter	Type	Required	Description
`client`	`FluxClient`	One of `client`/`async_client`	Synchronous Flux client
`async_client`	`AsyncFluxClient`	One of `client`/`async_client`	Async Flux client

Folder¶

Parameter	Type	Required	Default	Description
`folder_path`	`str`	Yes	—	FoxNose folder path (e.g. `"knowledge-base"`)

Content Mapping¶

Exactly one of these must be set:

Parameter	Type	Description
`page_content_field`	`str`	Single `data` field for `page_content`
`page_content_fields`	`list[str]`	Multiple fields concatenated (must be non-empty)
`page_content_mapper`	`Callable`	Custom `(result) -> str` function

Parameter	Type	Default	Description
`page_content_separator`	`str`	`"\n\n"`	Separator for `page_content_fields`

Metadata¶

Parameter	Type	Default	Description
`metadata_fields`	`list[str]`	`None`	Whitelist of `data` fields for metadata
`exclude_metadata_fields`	`list[str]`	`None`	Blacklist of `data` fields
`include_sys_metadata`	`bool`	`True`	Include `_sys` fields (key, folder, created_at, updated_at)

Warning

metadata_fields and exclude_metadata_fields are mutually exclusive. Setting both raises a validation error.

Search¶

Parameter	Type	Default	Description
`search_mode`	`str`	`"hybrid"`	`"text"`, `"vector"`, `"hybrid"`, or `"vector_boosted"`
`top_k`	`int`	`5`	Maximum results to return (must be >= 1)
`search_fields`	`list[str]`	`None`	Fields for text search (`find_text.fields`)
`text_threshold`	`float`	`None`	Typo tolerance (`find_text.threshold`, 0-1)
`vector_fields`	`list[str]`	`None`	Fields for vector search (`vector_search.fields`)
`similarity_threshold`	`float`	`None`	Minimum cosine similarity (0-1)
`where`	`dict`	`None`	Persistent structured filter
`hybrid_config`	`dict`	`None`	Hybrid mode config (`vector_weight`, `text_weight`, `rerank_results`)
`vector_boost_config`	`dict`	`None`	Boost config (`boost_factor`, `similarity_threshold`, `max_boost_results`)
`sort`	`list[str]`	`None`	Sort fields (prefix `-` for descending)
`search_kwargs`	`dict`	`{}`	Extra params passed to SDK methods (see below)

Note

Known keys in search_kwargs like limit and offset are extracted as named SDK method parameters. The rest are passed through **extra_body.

Warning

search_kwargs must not contain keys that conflict with SearchRequest fields: search_mode, vector_search, vector_field_search, hybrid_config, vector_boost_config, find_text, find_phrase.

Note

text_threshold and similarity_threshold are validated to be in the range 0-1.

Custom Embeddings¶

Parameter	Type	Default	Description
`embeddings`	`Embeddings`	`None`	LangChain Embeddings model for converting query text to vectors
`query_vector`	`list[float]`	`None`	Pre-computed static query vector
`vector_field`	`str`	`None`	Field name for `vector_field_search` (required when `embeddings` or `query_vector` is set)

Warning

embeddings and query_vector are mutually exclusive. vector_field and vector_fields are mutually exclusive. Custom embeddings are only supported in vector and vector_boosted modes.

Warning

When using embeddings, the query text may be sent to a third-party service (e.g. OpenAI) depending on the Embeddings implementation.

Search Modes¶

Mode	Text Search	Vector Search	Use Case
`text`	Required	Not allowed	Keyword/phrase search
`vector`	Not allowed	Required	Pure semantic search
`hybrid`	Required	Required	Blended text + vector with weights
`vector_boosted`	Required	Required	Text results boosted by vector similarity