Home Schema Example JSON Schema

Query Engine Settings

Type: object

This contains all the settings needed for the Jarvis v2 query engine.

root enabled

Enabled

Type: boolean Default: true

Whether this query engine is enabled.

root modules

Modules

Type: array

The modules to be used in the query engine.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items

root modules modules items oneOf Standard IR Module Settings

Standard IR Module Settings

Type: object

Settings for our standard IR-based module.

root modules modules items oneOf Standard IR Module Settings answer_finder

Answer Finder

Type: object

Configure how we retrieve, score, and rank answers.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker

Reranker

Type: object

Configure how we rerank answers.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions

Reranker Functions

Type: array

Rerankers are executed one after the other.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker

Min Cutoff Reranker

Type: object

Minimum cutoff reranker function. All resulting candidates will have scores greater or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition operator

Operator

Type: enum (of string)

The comparison operator.

Must be one of:

  • "gt"
  • "gte"
  • "lt"
  • "lte"
  • "eq"
  • "ne"
  • ">"
  • "≥"
  • ">="
  • "<"
  • "≤"
  • "<="
  • "="
  • "=="
  • "!="
  • "<>"
  • "≠"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition score_key

Score Key

Type: string

The key in the score vector to use for the comparison. When the key doesn't exist or is None, we return false.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker type

Type

Type: const Default: "min-cutoff-reranker"
Specific value: "min-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker

Max Cutoff Reranker

Type: object

Maximum cutoff reranker function. All resulting candidates will have scores lesser or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

Same definition as Condition

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker type

Type

Type: const Default: "max-cutoff-reranker"
Specific value: "max-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker

Top K Reranker

Type: object

Reranker keeps the top K candidates based on distinct score value and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker k

K

Type: integer Default: 1

Keep the top K candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker keep_ties

Keep Ties

Type: boolean Default: true

Whether to keep candidates whose score tie for top K. Defaults to true.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker

Load Full Document Reranker

Type: object

A pseudo-reranker that calls the load method on all candidate documents.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker type

Type

Type: const Default: "load-full-document-reranker"
Specific value: "load-full-document-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker

Order By Reranker

Type: object

Reorder the candidates based on scoring keys.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by

Order By

Type: array of string

The scoring fields to order the candidates by. Prefix with a +/- for ascending/descending order. You can also access sv, doc, and source fields using JMESPath syntax, e.g., sv.missed_tokens_count, doc.name, source.phone, etc.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by order_by items
Type: string

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker type

Type

Type: const Default: "order-by-reranker"
Specific value: "order-by-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker

First N Reranker

Type: object

Reranker keeps the first N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker n

N

Type: integer Default: 16

Keep the first N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker type

Type

Type: const Default: "first-n-reranker"
Specific value: "first-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker

Last N Reranker

Type: object

Reranker keeps the last N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker n

N

Type: integer Default: 16

Keep the last N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker type

Type

Type: const Default: "last-n-reranker"
Specific value: "last-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker

Reverse Reranker

Type: object

Reranker reverses the order of the candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker type

Type

Type: const Default: "reverse-reranker"
Specific value: "reverse-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker

Answer Field Reranker

Type: object

Reranker that keeps entries containing matching answer fields.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker match_type

Match Type

Type: enum (of string) Default: "all"

Whether to match all or any of the answer fields.

Must be one of:

  • "all"
  • "any"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker type

Type

Type: const Default: "answer-field-reranker"
Specific value: "answer-field-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker

Discard Or Keep Information Reranker

Type: object

Reranker that shortlists or discards entry based on the condition.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker discard_or_keep

Discard Or Keep

Type: enum (of string)

The field that decides whether to discard or keep the entries.

Must be one of:

  • "discard"
  • "keep"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker type

Type

Type: const Default: "discard-or-keep-information-reranker"
Specific value: "discard-or-keep-information-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever

Retriever

Type: object

Configure how we retrieve answer.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions

Retriever Functions

Type: array

Retrievers are executed concurrently and then concatenated after.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever

Answer Phrase Retriever

Type: object

Settings for retrieving intent phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever size

Size

Type: integer Default: 128

The size parameter to use when retrieving documents from the KB. Note that scan is not used.

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match

Tiered Fuzzy Match

Default: {"fuzzy_transpositions": true, "length_thresholds": [3, 6], "prefix_length": 2, "scores": [0.8, 0.9, 1.0]}

Configure how we perform fuzzy matching when retrieving phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings

TieredFuzzyMatchSettings

Type: object

Settings for tiered fuzzy matching.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings fuzzy_transpositions

Fuzzy Transpositions

Type: boolean Default: true

Whether to allow transpositions in fuzzy matching. See https://opensearch.org/docs/latest/query-dsl/full-text/match/#transpositions

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings length_thresholds

Length Thresholds

Type: array Default: [3, 6]

A tuple [a, b] where string lengths in [0, a] do not have fuzzy matching, lengths in (a, b] allow 1 Levenshtein distance, and lengths (b, ∞) allow 2 Levenshtein distances.

Must contain a minimum of 2 items

Must contain a maximum of 2 items

No Additional Items

Tuple Validation

Item at 1 must be:
root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings length_thresholds length_thresholds item 0
Type: integer

Value must be greater or equal to 0

Item at 2 must be:
root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings length_thresholds length_thresholds item 1
Type: integer

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf TieredFuzzyMatchSettings prefix_length

Prefix Length

Type: integer Default: 2

The length of the prefix to use for fuzzy matching. See https://opensearch.org/docs/latest/query-dsl/full-text/match/#prefix-length

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever tiered_fuzzy_match anyOf item 1
Type: null

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Phrase Retriever type

Type

Type: const Default: "answer-phrase-retriever"
Specific value: "answer-phrase-retriever"

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Intent Phrase Retriever

Intent Phrase Retriever

Type: object

Settings for retrieving intent phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Intent Phrase Retriever size

Size

Type: integer Default: 128

The size parameter to use when retrieving documents from the KB. Note that scan is not used.

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Intent Phrase Retriever type

Type

Type: const Default: "intent-phrase-retriever"
Specific value: "intent-phrase-retriever"

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever

Answer Retriever

Type: object

Settings for retrieving answers.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever max_clause_count

Max Clause Count

Type: integer Default: 1000

The maximum number of answer phrase clauses to use in a query. When there are more clauses, multiple queries will be executed concurrently.

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever size

Size

Type: integer Default: 128

The size parameter to use when retrieving documents from the KB. Note that scan is not used.

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights

Retrieval Weights Settings

Type: array

The weights to use for each field in the retrieval function. By default, each field is weighted 1.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights Retrieval Weights

Retrieval Weights

Type: object

Settings for retrieval weights.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights Retrieval Weights answer_type

Answer Type

Type: enum (of string)

The answer type where the object is used.

Must be one of:

  • "clinical-trial"
  • "dialogue"
  • "directory"
  • "document"
  • "document-snippet"
  • "monograph"
  • "roster"
  • "test"

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights Retrieval Weights weight

Weight

Default: "matched_tokens_precision"

The weight to assign to the field. It can also be one of the supported phrase scoring keys or a constant weight value.

root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights Retrieval Weights weight anyOf item 0
Type: number
root modules modules items oneOf Standard IR Module Settings answer_finder retriever retriever_functions retriever_functions items oneOf Answer Retriever weights Retrieval Weights weight anyOf item 1
Type: enum (of string)

Must be one of:

  • "phrase_tokens_count"
  • "matched_tokens_score"
  • "missed_tokens_score"
  • "matched_tokens_count"
  • "missed_tokens_count"
  • "matched_tokens_precision"
  • "missed_tokens_precision"

root modules modules items oneOf Standard IR Module Settings answer_finder retriever verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer

Scorer

Type: object

Configure how we score answers.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions

Scorer Functions

Type: array

Scorer functions are executed concurrently and then merged together after.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer

Phrase Scorer

Type: object

Settings for a phrase scorer.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer type

Type

Type: const Default: "phrase-scorer"
Specific value: "phrase-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities

Institution Priority Order

Type: array

The list of institution affinities.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity

InstitutionAffinity

Type: object

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities

Affinities

Type: array of string

The list of institutions starting from most to least affiniated. This field is case insensitive.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities affinities items
Type: string

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer type

Type

Type: const Default: "directory-answer-scorer"
Specific value: "directory-answer-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders

Answer Phrase Finders

Type: array

Configure how we retrieve, score, and rank answer phrases. This is done per answer type/field. Multiple finders for each answer type/field is supported.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings

Answer Phrase Finder Settings

Type: object

Configure how we retrieve, score, and rank answer phrases in the module.

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings answer_field

Answer Field

Type: string

The field in the answer where the object is used.

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings answer_type

Answer Type

Type: enum (of string)

The answer type where the object is used.

Must be one of:

  • "clinical-trial"
  • "dialogue"
  • "directory"
  • "document"
  • "document-snippet"
  • "monograph"
  • "roster"
  • "test"

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings reranker

Reranker

Type: object

Configure how we rerank phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions

Reranker Functions

Type: array

Rerankers are executed one after the other.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker

Min Cutoff Reranker

Type: object

Minimum cutoff reranker function. All resulting candidates will have scores greater or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition operator

Operator

Type: enum (of string)

The comparison operator.

Must be one of:

  • "gt"
  • "gte"
  • "lt"
  • "lte"
  • "eq"
  • "ne"
  • ">"
  • "≥"
  • ">="
  • "<"
  • "≤"
  • "<="
  • "="
  • "=="
  • "!="
  • "<>"
  • "≠"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition score_key

Score Key

Type: string

The key in the score vector to use for the comparison. When the key doesn't exist or is None, we return false.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker type

Type

Type: const Default: "min-cutoff-reranker"
Specific value: "min-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker

Max Cutoff Reranker

Type: object

Maximum cutoff reranker function. All resulting candidates will have scores lesser or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

Same definition as Condition

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker type

Type

Type: const Default: "max-cutoff-reranker"
Specific value: "max-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker

Top K Reranker

Type: object

Reranker keeps the top K candidates based on distinct score value and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker k

K

Type: integer Default: 1

Keep the top K candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker keep_ties

Keep Ties

Type: boolean Default: true

Whether to keep candidates whose score tie for top K. Defaults to true.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker

Load Full Document Reranker

Type: object

A pseudo-reranker that calls the load method on all candidate documents.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker type

Type

Type: const Default: "load-full-document-reranker"
Specific value: "load-full-document-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker

Order By Reranker

Type: object

Reorder the candidates based on scoring keys.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by

Order By

Type: array of string

The scoring fields to order the candidates by. Prefix with a +/- for ascending/descending order. You can also access sv, doc, and source fields using JMESPath syntax, e.g., sv.missed_tokens_count, doc.name, source.phone, etc.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by order_by items
Type: string

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker type

Type

Type: const Default: "order-by-reranker"
Specific value: "order-by-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker

First N Reranker

Type: object

Reranker keeps the first N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker n

N

Type: integer Default: 16

Keep the first N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker type

Type

Type: const Default: "first-n-reranker"
Specific value: "first-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker

Last N Reranker

Type: object

Reranker keeps the last N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker n

N

Type: integer Default: 16

Keep the last N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker type

Type

Type: const Default: "last-n-reranker"
Specific value: "last-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker

Reverse Reranker

Type: object

Reranker reverses the order of the candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker type

Type

Type: const Default: "reverse-reranker"
Specific value: "reverse-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker

Answer Field Reranker

Type: object

Reranker that keeps entries containing matching answer fields.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker match_type

Match Type

Type: enum (of string) Default: "all"

Whether to match all or any of the answer fields.

Must be one of:

  • "all"
  • "any"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker type

Type

Type: const Default: "answer-field-reranker"
Specific value: "answer-field-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker

Discard Or Keep Information Reranker

Type: object

Reranker that shortlists or discards entry based on the condition.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker discard_or_keep

Discard Or Keep

Type: enum (of string)

The field that decides whether to discard or keep the entries.

Must be one of:

  • "discard"
  • "keep"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker type

Type

Type: const Default: "discard-or-keep-information-reranker"
Specific value: "discard-or-keep-information-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings scorer

Scorer

Type: object

Configure how we score phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions

Scorer Functions

Type: array

Scorer functions are executed concurrently and then merged together after.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer

Phrase Scorer

Type: object

Settings for a phrase scorer.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer type

Type

Type: const Default: "phrase-scorer"
Specific value: "phrase-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities

Institution Priority Order

Type: array

The list of institution affinities.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity

InstitutionAffinity

Type: object

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities

Affinities

Type: array of string

The list of institutions starting from most to least affiniated. This field is case insensitive.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities affinities items
Type: string

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer type

Type

Type: const Default: "directory-answer-scorer"
Specific value: "directory-answer-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_phrase_finders Answer Phrase Finder Settings type

Type

Type: const Default: "answer"
Specific value: "answer"

root modules modules items oneOf Standard IR Module Settings intent_phrase_finder

Intent Phrase Finder

Type: object

Configure how we retrieve, score, and rank intent phrases.

root modules modules items oneOf Standard IR Module Settings intent_phrase_finder reranker

Reranker

Type: object

Configure how we rerank phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions

Reranker Functions

Type: array

Rerankers are executed one after the other.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker

Min Cutoff Reranker

Type: object

Minimum cutoff reranker function. All resulting candidates will have scores greater or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition operator

Operator

Type: enum (of string)

The comparison operator.

Must be one of:

  • "gt"
  • "gte"
  • "lt"
  • "lte"
  • "eq"
  • "ne"
  • ">"
  • "≥"
  • ">="
  • "<"
  • "≤"
  • "<="
  • "="
  • "=="
  • "!="
  • "<>"
  • "≠"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker conditions Condition score_key

Score Key

Type: string

The key in the score vector to use for the comparison. When the key doesn't exist or is None, we return false.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker type

Type

Type: const Default: "min-cutoff-reranker"
Specific value: "min-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Min Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker

Max Cutoff Reranker

Type: object

Maximum cutoff reranker function. All resulting candidates will have scores lesser or equal to the cutoff.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions

Conditions

Type: array Default: []

Conditions for documents before we apply the reranker Documents that fail the condition will be passed through to the next reranker.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker conditions Condition

Condition

Type: object

A condition to check before applying a reranker.

Same definition as Condition

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker type

Type

Type: const Default: "max-cutoff-reranker"
Specific value: "max-cutoff-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Max Cutoff Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker

Top K Reranker

Type: object

Reranker keeps the top K candidates based on distinct score value and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker k

K

Type: integer Default: 1

Keep the top K candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker keep_ties

Keep Ties

Type: boolean Default: true

Whether to keep candidates whose score tie for top K. Defaults to true.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker score_key

Score Key

Type: string

The key in the score vector to use for the cutoff.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Top K Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker

Load Full Document Reranker

Type: object

A pseudo-reranker that calls the load method on all candidate documents.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Load Full Document Reranker type

Type

Type: const Default: "load-full-document-reranker"
Specific value: "load-full-document-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker

Order By Reranker

Type: object

Reorder the candidates based on scoring keys.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by

Order By

Type: array of string

The scoring fields to order the candidates by. Prefix with a +/- for ascending/descending order. You can also access sv, doc, and source fields using JMESPath syntax, e.g., sv.missed_tokens_count, doc.name, source.phone, etc.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker order_by order_by items
Type: string

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker type

Type

Type: const Default: "order-by-reranker"
Specific value: "order-by-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Order By Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker

First N Reranker

Type: object

Reranker keeps the first N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker n

N

Type: integer Default: 16

Keep the first N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker type

Type

Type: const Default: "first-n-reranker"
Specific value: "first-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf First N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker

Last N Reranker

Type: object

Reranker keeps the last N candidates and discards the rest.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker n

N

Type: integer Default: 16

Keep the last N candidates.

Value must be strictly greater than 0

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker type

Type

Type: const Default: "last-n-reranker"
Specific value: "last-n-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Last N Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker

Reverse Reranker

Type: object

Reranker reverses the order of the candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker type

Type

Type: const Default: "reverse-reranker"
Specific value: "reverse-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Reverse Reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker

Answer Field Reranker

Type: object

Reranker that keeps entries containing matching answer fields.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker keep_original_if_all_discarded

Keep Original If All Discarded

Type: boolean Default: false

If all documents are discarded, keep the original candidates.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker match_type

Match Type

Type: enum (of string) Default: "all"

Whether to match all or any of the answer fields.

Must be one of:

  • "all"
  • "any"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Answer Field Reranker type

Type

Type: const Default: "answer-field-reranker"
Specific value: "answer-field-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker

Discard Or Keep Information Reranker

Type: object

Reranker that shortlists or discards entry based on the condition.

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker discard_or_keep

Discard Or Keep

Type: enum (of string)

The field that decides whether to discard or keep the entries.

Must be one of:

  • "discard"
  • "keep"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker reranker_functions reranker_functions items oneOf Discard Or Keep Information Reranker type

Type

Type: const Default: "discard-or-keep-information-reranker"
Specific value: "discard-or-keep-information-reranker"

root modules modules items oneOf Standard IR Module Settings answer_finder reranker verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings intent_phrase_finder retriever

Retriever Settings

Type: object

Configure how we retrieve documents.

Same definition as retriever

root modules modules items oneOf Standard IR Module Settings intent_phrase_finder scorer

Scorer

Type: object

Configure how we score phrases.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions

Scorer Functions

Type: array

Scorer functions are executed concurrently and then merged together after.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer

Phrase Scorer

Type: object

Settings for a phrase scorer.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer type

Type

Type: const Default: "phrase-scorer"
Specific value: "phrase-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Phrase Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.
root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer

Basic Answer Scorer

Type: object

Score an answer based on standard token level P/R/F metrics.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities

Institution Priority Order

Type: array

The list of institution affinities.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity

InstitutionAffinity

Type: object

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities

Affinities

Type: array of string

The list of institutions starting from most to least affiniated. This field is case insensitive.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer institution_affinities InstitutionAffinity affinities affinities items
Type: string

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer type

Type

Type: const Default: "directory-answer-scorer"
Specific value: "directory-answer-scorer"

root modules modules items oneOf Standard IR Module Settings answer_finder scorer scorer_functions scorer_functions items oneOf Basic Answer Scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings answer_finder scorer verbose

Verbose

Type: boolean Default: false

Whether to log verbose debugging information.

root modules modules items oneOf Standard IR Module Settings intent_phrase_finder type

Type

Type: const Default: "intent"
Specific value: "intent"

root modules modules items oneOf Standard IR Module Settings intents

Intents

Type: array of string

The intents will likely trigger this module.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings intents Intent

Intent

Type: string

The intent phrase.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings module_responder

Module Responder


The responder to use for this module. Module responders evaluate all the candidate answers to form a cohesive response for a module.

root modules modules items oneOf Standard IR Module Settings module_responder anyOf Directory Module Responder Settings

Directory Module Responder Settings

Type: object

Settings for our directory module responder.

root modules modules items oneOf Standard IR Module Settings module_responder anyOf Directory Module Responder Settings max_answer_candidates_to_load

Max Answer Candidates To Load

Default: null

The maximum number of answer candidate bare answers that we will load in order to sort and display to the user. Default to None which means to load all answer candidates.

root modules modules items oneOf Standard IR Module Settings module_responder anyOf Directory Module Responder Settings type

Type

Type: const Default: "directory-ir"
Specific value: "directory-ir"
root modules modules items oneOf Standard IR Module Settings module_responder anyOf Clinical Trial Module Responder Settings

Clinical Trial Module Responder Settings

Type: object

Settings for our clinical trial module responder.

root modules modules items oneOf Standard IR Module Settings module_responder anyOf Clinical Trial Module Responder Settings max_answer_candidates_to_load

Max Answer Candidates To Load

Default: null

The maximum number of answer candidate bare answers that we will load in order to sort and display to the user. Default to None which means to load all answer candidates.

root modules modules items oneOf Standard IR Module Settings module_responder anyOf Clinical Trial Module Responder Settings type

Type

Type: const Default: "clinical-trial-ir"
Specific value: "clinical-trial-ir"

root modules modules items oneOf Standard IR Module Settings module_stopwords

Module Stopwords

Type: array

Stopwords used in this module.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings module_stopwords Stopword Settings

Stopword Settings

Type: object

Configure stopwords on a per answer type/field level.

root modules modules items oneOf Standard IR Module Settings module_stopwords Stopword Settings answer_field

Answer Field

Type: string

The field in the answer where the object is used.

root modules modules items oneOf Standard IR Module Settings module_stopwords Stopword Settings answer_type

Answer Type

Type: enum (of string)

The answer type where the object is used.

Must be one of:

  • "clinical-trial"
  • "dialogue"
  • "directory"
  • "document"
  • "document-snippet"
  • "monograph"
  • "roster"
  • "test"

root modules modules items oneOf Standard IR Module Settings module_stopwords Stopword Settings stopwords

Stopwords


The stopwords for this answer type/field. It can be a list of phrases or a stopword set name.

root modules modules items oneOf Standard IR Module Settings module_synonyms

Module Synonyms

Type: array

Synonyms used in this module.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings

Synonym Settings

Type: object

Configure synonyms on a per answer type/field level.

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings answer_field

Answer Field

Type: string

The field in the answer where the object is used.

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings answer_type

Answer Type

Type: enum (of string)

The answer type where the object is used.

Must be one of:

  • "clinical-trial"
  • "dialogue"
  • "directory"
  • "document"
  • "document-snippet"
  • "monograph"
  • "roster"
  • "test"

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonym_expansion_rounds

Synonym Expansion Rounds

Type: integer Default: 3

The number of rounds to expand the synonyms. Each round will generate more candidate synonyms.

Value must be greater or equal to 0

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms

Synonyms

Type: array

The synonyms for this answer type/field as a list of synonym groups.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group

Synonym Group


Describes how synonyms are generated

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group anyOf One Way Synonym Group

One Way Synonym Group

Type: object

In a one-way synonym group, all phrases are synonyms of the first phrase but not the other way round.

Each additional property must conform to the following schema

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group anyOf One Way Synonym Group additionalProperties
Type: array of string

Synonyms of the first phrase.

Must contain a minimum of 1 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group anyOf One Way Synonym Group additionalProperties additionalProperties items
Type: string
root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group anyOf Two Way Synonym Group

Two Way Synonym Group

Type: array of string

In a two-way synonym group, all pairs of phrases are synonyms of each other and can be used interchangeably.

Must contain a minimum of 2 items

No Additional Items

Each item of this array must be:

root modules modules items oneOf Standard IR Module Settings module_synonyms Synonym Settings synonyms Synonym Group anyOf Two Way Synonym Group Synonym

Synonym

Type: string

A Single synonym phrase.

Must be at least 1 characters long

root modules modules items oneOf Standard IR Module Settings name

Name

Type: string

The name of the module. This should be unique within the query engine.

root modules modules items oneOf Standard IR Module Settings type

Type

Type: const Default: "standard-ir"
Specific value: "standard-ir"
root modules modules items oneOf LLM Document Module Settings

LLM Document Module Settings

Type: object

Settings for our document LLM module.

root modules modules items oneOf LLM Document Module Settings early_terminate

Allow early termination

Type: boolean Default: false

Defines whether to allow early termination (before LLM call). True when stock phrase functionality is requested. Default is False.

root modules modules items oneOf LLM Document Module Settings emb_model_to_use

Model to use

Type: string Default: "cohere.embed-multilingual-v3"

Name of model to be used, it should correspond with one of the values in embedding_model.

root modules modules items oneOf LLM Document Module Settings embedding_model

Embedding

Type: array

Defines settings for the embedding model during data and query flow.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings

Embedding Model Settings

Type: object

Defines settings for the embedding model during data and query flow.

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings context_size

Context window size of the embedding model

Type: integer Default: 512

Context size

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings default_prompt_name

Prompt Name

Type: enum (of string) Default: "query"

Prompt for the embedding.

Must be one of:

  • "query"
  • "text"

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings dimension

Dimension

Type: integer Default: 1024

Dimension of the embedding

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings emb_model_source

Model Source

Type: enum (of string) Default: "bedrock"

Where is the model from?

Must be one of:

  • "huggingface"
  • "bedrock"

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings hnsw_kwargs

HNSW settings

Default: null

HNSW settings

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings hnsw_kwargs anyOf HNSW Graph Settings

HNSW Graph Settings

Type: object

Defines settings for HNSW graph. See
https://github.com/run-llama/llamaindex/blob/977d60a058c691957dae3eb3c66c1894faea24ac/llama-index-integrations/vectorstores/llama-index-vector-stores-postgres/llamaindex/vectorstores/postgres/base.py#L570

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings hnsw_kwargs anyOf HNSW Graph Settings hnsw_dist_method

dist_method

Type: enum (of string) Default: "vector_cosine_ops"

Distance metric to use. Note that by default PGVectorStore.buildquery calls cosine_distance

Must be one of:

  • "vector_l2_ops"
  • "vector_ip_ops"
  • "vector_cosine_ops"
  • "vector_l1_ops"
  • "bit_hamming_ops"
  • "bit_jaccard_ops"

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings hnsw_kwargs anyOf HNSW Graph Settings hnsw_ef_construction

ef_construction

Type: integer Default: 64

Size of the dynamic candidate list for constructing the graph. Higher value provides better recall at the cost of speed
root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings hnsw_kwargs anyOf item 1
Type: null

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings model

Model

Type: string Default: "cohere.embed-multilingual-v3"

Name of embedding model to use.

Must be at least 1 characters long

root modules modules items oneOf LLM Document Module Settings embedding_model Embedding Model Settings query_instruction

Query Instruction

Type: string Default: "Represent this sentence for searching relevant passages:"

Instruction for the query.

root modules modules items oneOf LLM Document Module Settings llm_model

LLM Settings

Type: object

Defines settings for the LLM model during query flow.

root modules modules items oneOf LLM Document Module Settings llm_model context_template

Context Template

Type: string Default: "You are a friendly and warm question-answer hospital assistant, employed as an employee of the hospital.\n The only information you know is the context provided. You can use only that information as your hidden knowledge base.\n When you answer, the context provided will be known as \"my training.\"\n If the context provided provides links, show it to the user."

Template for the context

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_identifier

Guardrail Identifier

Type: string Default: "arn:aws:bedrock:us-west-2:394252546268:guardrail/f83z6kx1d6hl"

Name of AWS guardrail if applicable. Should start with arn:aws:bedrock:us-west-2:.

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_version

Guardrail Version

Type: string Default: "1"

AWS Guardrail version (string) if applicable

root modules modules items oneOf LLM Document Module Settings llm_model max_tokens

Max Tokens

Type: integer Default: 1000

Max tokens returned by LLM

root modules modules items oneOf LLM Document Module Settings llm_model model

Model

Type: string

Name of LLM as per HuggingFace

Must be at least 1 characters long

root modules modules items oneOf LLM Document Module Settings llm_model reranker_similarity_cutoff

Reranker Similarity Cutoff

Type: number Default: 0

Similarity score cutoff for node postprocessing based on reranker score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_cutoff

Similarity Cutoff

Type: number Default: 0.38

Similarity score cutoff for node postprocessing based on embedding score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_top_k

Similarity top K

Type: integer Default: 6

Number of nodes to return after retrieval

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes

Supported language codes

Type: array of string

Restricts bot to only answer only these languages. Provide list of 2-letter language codes, and double-check that Amazon Comprehend / Translate supports them.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf LLM Document Module Settings llm_model system_prompt

System Prompt

Type: string Default: "Answer the QUERY below only using the DOCUMENTs below as context, and not your trained knowledge."

Prompt for the LLM

root modules modules items oneOf LLM Document Module Settings llm_model temperature

Temperature

Type: number Default: 0

Temperature

root modules modules items oneOf LLM Document Module Settings name

Name

Type: string

The name of the module. This should be unique within the query engine.

root modules modules items oneOf LLM Document Module Settings postprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions

Postprocessor Functions

Type: array

Postprocessor functions to be applied one after another in the model.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Language Remover

Language Remover

Type: object

Language Remover LLM Postprocessor function.

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Amazon Translator

Amazon Translator

Type: object

Amazon Translator LLM Postprocessor function.

root modules modules items oneOf LLM Document Module Settings preprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions

Preprocessor Functions

Type: array

Preprocessor functions to be applied one after another.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Truncate

Truncate

Type: object

Truncate Preprocessor function.
root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter

Language Filter

Type: object

Language Filter Preprocessor function.

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes

Supported language codes

Type: array of string

The language codes that are supported for this client. By default, if this is empty, all languages are supported. If a language is not supported, query will be nullified

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf LLM Document Module Settings query_feature_settings

Query Feature settings

Type: object

Configure which query feature functions to use

root modules modules items oneOf LLM Document Module Settings query_feature_settings query_feature_functions

Query Feature Functions

Type: array

Query Feature functions to be applied one after another.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings query_feature_settings query_feature_functions Language Detector

Language Detector

Type: object

Language Detector Query Feature function.

root modules modules items oneOf LLM Document Module Settings query_feature_settings query_feature_functions Language Detector type

Type

Type: const Default: "language-detector"
Specific value: "language-detector"

root modules modules items oneOf LLM Document Module Settings storage_prefix

Storage Prefix

Type: const Default: "document-llm"
Specific value: "document-llm"

root modules modules items oneOf LLM Document Module Settings type

Type

Type: const Default: "document-llm"
Specific value: "document-llm"
root modules modules items oneOf Cleo Dialogue LLM Module Settings

Cleo Dialogue LLM Module Settings

Type: object

Settings for our Cleo dialogue LLM module.

root modules modules items oneOf Cleo Dialogue LLM Module Settings early_terminate

Allow early termination

Type: boolean Default: false

Defines whether to allow early termination (before LLM call). True when stock phrase functionality is requested. Default is False.

root modules modules items oneOf Cleo Dialogue LLM Module Settings emb_model_to_use

Model to use

Type: string Default: "cohere.embed-multilingual-v3"

Name of model to be used, it should correspond with one of the values in embedding_model.

root modules modules items oneOf Cleo Dialogue LLM Module Settings embedding_model

Embedding

Type: array

Defines settings for the embedding model during data and query flow.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Cleo Dialogue LLM Module Settings embedding_model Embedding Model Settings

Embedding Model Settings

Type: object

Defines settings for the embedding model during data and query flow.

Same definition as Embedding Model Settings

root modules modules items oneOf Cleo Dialogue LLM Module Settings llm_model

LLM Settings

Type: object

Defines settings for the LLM model during query flow.

root modules modules items oneOf LLM Document Module Settings llm_model context_template

Context Template

Type: string Default: "You are a friendly and warm question-answer hospital assistant, employed as an employee of the hospital.\n The only information you know is the context provided. You can use only that information as your hidden knowledge base.\n When you answer, the context provided will be known as \"my training.\"\n If the context provided provides links, show it to the user."

Template for the context

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_identifier

Guardrail Identifier

Type: string Default: "arn:aws:bedrock:us-west-2:394252546268:guardrail/f83z6kx1d6hl"

Name of AWS guardrail if applicable. Should start with arn:aws:bedrock:us-west-2:.

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_version

Guardrail Version

Type: string Default: "1"

AWS Guardrail version (string) if applicable

root modules modules items oneOf LLM Document Module Settings llm_model max_tokens

Max Tokens

Type: integer Default: 1000

Max tokens returned by LLM

root modules modules items oneOf LLM Document Module Settings llm_model model

Model

Type: string

Name of LLM as per HuggingFace

Must be at least 1 characters long

root modules modules items oneOf LLM Document Module Settings llm_model reranker_similarity_cutoff

Reranker Similarity Cutoff

Type: number Default: 0

Similarity score cutoff for node postprocessing based on reranker score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_cutoff

Similarity Cutoff

Type: number Default: 0.38

Similarity score cutoff for node postprocessing based on embedding score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_top_k

Similarity top K

Type: integer Default: 6

Number of nodes to return after retrieval

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes

Supported language codes

Type: array of string

Restricts bot to only answer only these languages. Provide list of 2-letter language codes, and double-check that Amazon Comprehend / Translate supports them.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf LLM Document Module Settings llm_model system_prompt

System Prompt

Type: string Default: "Answer the QUERY below only using the DOCUMENTs below as context, and not your trained knowledge."

Prompt for the LLM

root modules modules items oneOf LLM Document Module Settings llm_model temperature

Temperature

Type: number Default: 0

Temperature

root modules modules items oneOf Cleo Dialogue LLM Module Settings metadata_filter_keys

Metadata Filter keys

Type: array of string

Configure valid keys that will be used to filter metadata. They should by default be at least any of the types of QueryFeatureSettings.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Cleo Dialogue LLM Module Settings metadata_filter_keys metadata_filter_keys items
Type: string

root modules modules items oneOf Cleo Dialogue LLM Module Settings name

Name

Type: string

The name of the module. This should be unique within the query engine.

root modules modules items oneOf Cleo Dialogue LLM Module Settings postprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions

Postprocessor Functions

Type: array

Postprocessor functions to be applied one after another in the model.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Language Remover

Language Remover

Type: object

Language Remover LLM Postprocessor function.

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Amazon Translator

Amazon Translator

Type: object

Amazon Translator LLM Postprocessor function.

root modules modules items oneOf Cleo Dialogue LLM Module Settings preprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions

Preprocessor Functions

Type: array

Preprocessor functions to be applied one after another.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Truncate

Truncate

Type: object

Truncate Preprocessor function.
root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter

Language Filter

Type: object

Language Filter Preprocessor function.

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes

Supported language codes

Type: array of string

The language codes that are supported for this client. By default, if this is empty, all languages are supported. If a language is not supported, query will be nullified

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf Cleo Dialogue LLM Module Settings query_feature_settings

Scorer Settings

Type: object

Configure how we generate query features.

Same definition as query_feature_settings

root modules modules items oneOf Cleo Dialogue LLM Module Settings storage_prefix

Storage Prefix

Type: const Default: "cleo-dialogue-llm"
Specific value: "cleo-dialogue-llm"

root modules modules items oneOf Cleo Dialogue LLM Module Settings type

Type

Type: const Default: "dialogue-llm"
Specific value: "dialogue-llm"
root modules modules items oneOf Agentic Module Settings

Agentic Module Settings

Type: object

Defines settings for Agentic modules

# schema: jarvisv2/QueryEngineSettings.schema

name: agentic-module-query-engine
tenants: [all]
modules:
- type: agentic-module
name: agentic-module-v0
storage_prefix: agentic-module-v0-
#end agentic-module-v0

end modules

root modules modules items oneOf Agentic Module Settings early_terminate

Allow early termination

Type: boolean Default: false

Defines whether to allow early termination (before LLM call). True when stock phrase functionality is requested. Default is False.

root modules modules items oneOf Agentic Module Settings emb_model_to_use

Model to use

Type: string Default: "cohere.embed-multilingual-v3"

Name of model to be used, it should correspond with one of the values in embedding_model.

root modules modules items oneOf Agentic Module Settings embedding_model

Embedding

Type: array

Defines settings for the embedding model during data and query flow.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Agentic Module Settings embedding_model Embedding Model Settings

Embedding Model Settings

Type: object

Defines settings for the embedding model during data and query flow.

Same definition as Embedding Model Settings

root modules modules items oneOf Agentic Module Settings llm_model

LLM Settings

Type: object

Defines settings for the LLM model during query flow.

root modules modules items oneOf LLM Document Module Settings llm_model context_template

Context Template

Type: string Default: "You are a friendly and warm question-answer hospital assistant, employed as an employee of the hospital.\n The only information you know is the context provided. You can use only that information as your hidden knowledge base.\n When you answer, the context provided will be known as \"my training.\"\n If the context provided provides links, show it to the user."

Template for the context

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_identifier

Guardrail Identifier

Type: string Default: "arn:aws:bedrock:us-west-2:394252546268:guardrail/f83z6kx1d6hl"

Name of AWS guardrail if applicable. Should start with arn:aws:bedrock:us-west-2:.

root modules modules items oneOf LLM Document Module Settings llm_model guardrail_version

Guardrail Version

Type: string Default: "1"

AWS Guardrail version (string) if applicable

root modules modules items oneOf LLM Document Module Settings llm_model max_tokens

Max Tokens

Type: integer Default: 1000

Max tokens returned by LLM

root modules modules items oneOf LLM Document Module Settings llm_model model

Model

Type: string

Name of LLM as per HuggingFace

Must be at least 1 characters long

root modules modules items oneOf LLM Document Module Settings llm_model reranker_similarity_cutoff

Reranker Similarity Cutoff

Type: number Default: 0

Similarity score cutoff for node postprocessing based on reranker score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_cutoff

Similarity Cutoff

Type: number Default: 0.38

Similarity score cutoff for node postprocessing based on embedding score

root modules modules items oneOf LLM Document Module Settings llm_model similarity_top_k

Similarity top K

Type: integer Default: 6

Number of nodes to return after retrieval

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes

Supported language codes

Type: array of string

Restricts bot to only answer only these languages. Provide list of 2-letter language codes, and double-check that Amazon Comprehend / Translate supports them.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings llm_model supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf LLM Document Module Settings llm_model system_prompt

System Prompt

Type: string Default: "Answer the QUERY below only using the DOCUMENTs below as context, and not your trained knowledge."

Prompt for the LLM

root modules modules items oneOf LLM Document Module Settings llm_model temperature

Temperature

Type: number Default: 0

Temperature

root modules modules items oneOf Agentic Module Settings metadata_filter_keys

Metadata Filter keys

Type: array of string

Configure valid keys that will be used to filter metadata. They should by default be at least any of the types of QueryFeatureSettings.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf Agentic Module Settings metadata_filter_keys metadata_filter_keys items
Type: string

root modules modules items oneOf Agentic Module Settings name

Name

Type: string

The name of the module. This should be unique within the query engine.

root modules modules items oneOf Agentic Module Settings postprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions

Postprocessor Functions

Type: array

Postprocessor functions to be applied one after another in the model.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Language Remover

Language Remover

Type: object

Language Remover LLM Postprocessor function.

root modules modules items oneOf LLM Document Module Settings postprocessor_settings postprocessor_functions postprocessor_functions items oneOf Amazon Translator

Amazon Translator

Type: object

Amazon Translator LLM Postprocessor function.

root modules modules items oneOf Agentic Module Settings preprocessor_settings

Postprocessor settings

Type: object

Configure which postprocessor functions to use

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions

Preprocessor Functions

Type: array

Preprocessor functions to be applied one after another.

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Truncate

Truncate

Type: object

Truncate Preprocessor function.
root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter

Language Filter

Type: object

Language Filter Preprocessor function.

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes

Supported language codes

Type: array of string

The language codes that are supported for this client. By default, if this is empty, all languages are supported. If a language is not supported, query will be nullified

 No Additional Items

Each item of this array must be:

root modules modules items oneOf LLM Document Module Settings preprocessor_settings preprocessor_functions preprocessor_functions items oneOf Language Filter supported_language_codes supported_language_codes items
Type: string

root modules modules items oneOf Agentic Module Settings query_feature_settings

Scorer Settings

Type: object

Configure how we generate query features.

Same definition as query_feature_settings

root modules modules items oneOf Agentic Module Settings storage_prefix

Storage Prefix

Type: const Default: "agentic-module-v0-"
Specific value: "agentic-module-v0-"

root modules modules items oneOf Agentic Module Settings type

Type

Type: const Default: "agentic-module"
Specific value: "agentic-module"
root modules modules items oneOf Noop Module Settings

Noop Module Settings

Type: object

Noop module settings.

root modules modules items oneOf Noop Module Settings name

Name

Type: string Default: "noop"

The name of the module. This should be unique within the query engine.

root modules modules items oneOf Noop Module Settings type

Type

Type: const Default: "noop"
Specific value: "noop"

root name

Name

Type: string

The name of the query engine. This should be globally unique.

root version

Version

Type: string Default: "unknown"

Version of this query engine based on the Kondo resource version.