Settings

Setting

Valid values

Description

ENABLE_DEBUG_RESPONSES

true or false

Determines whether to log original English responses and translated responses for debugging

ES_USE_KEYWORD_FILTERS

true or false

Determines whether to detect keywords from Comprehend when searching for answers

ES_EXPAND_CONTRACTIONS

JSON format

Expand contractions to resolve problems with keyword filters.

ES_KEYWORD_SYNTAX_TYPES

comma separate list
See Analyze Syntax

A list of tokens representing parts of speech identified by Amazon Comprehend for matching questions

ES_SYNTAX_CONFIDENCE_LIMIT

decimal between 0 and 0.99

Amazon Comprehend makes a best effort to determine the parts of speech in a sentence. The keywords will only be used if the confidence limit is greater than this amount

ES_MINIMUM_SHOULD_MATCH

“2<75%”
See query-dsl-minimum-should-match for syntax

Determines how close a question should match to return a hit

ES_NO_HITS_QUESTION

text

The question AgentX should use when it cannot find an answer

ES_USE_FUZZY_MATCH

true or false

Determines whether AgentX should return answers similar to the question asked. See Fuzzy Query for more information

ES_PHRASE_BOOST

integer

If the user’s question is a phrase match to a question in the knowledge then boost the score by this factor.

ES_SCORE_ANSWER_FIELD

true or false

Search the content of the answer field as a 2nd pass query (if there’s no good match from 1st pass query on question fields).

ENABLE_SENTIMENT_SUPPORT

true or false

Enables Amazon Comprehend be used for sentiment analysis

ENABLE_MULTI_LANGUAGE_SUPPORT

true or false

Enable or Disable Amazon Translate support. See list of Automatic Translation

ENABLE_CUSTOM_TERMINOLOGY

true or false

Enable support for installed Custom Terminology files when using Amazon Translate

MINIMUM_CONFIDENCE_SCORE

decimal between 0.0 and 0,99

The minimum confidence before Amazon Comprehend will determine the user’s language

ALT_SEARCH_KENDRA_INDEXES

An Array of comma separated Ids

A list of one or more Amazon Kendra indexes used for Kendra fallback

ALT_SEARCH_KENDRA_FALLBACK_CONFIDENCE_SCORE

VERY_HIGH | HIGH | MEDIUM | LOW

Answers will only be returned that or at or above the specified confidence level when using Kendra Fallback

ALT_SEARCH_KENDRA_FAQ_CONFIDENCE_SCORE:

VERY_HIGH | HIGH | MEDIUM | LOW

Minimum Kendra confidence level threshold for Kendra FAQ

ALT_SEARCH_KENDRA_S3_SIGNED_URLS

true or false

Enables signed S3 Urls for Amazon Kendra results. If enabled, allows support for Kendra documents which are not publicly accessible.

ALT_SEARCH_KENDRA_S3_SIGNED_URL_EXPIRE_SECS

integer

Determines length of time in seconds for the validity of signed S3 Urls in Kendra fallback.

ALT_SEARCH_KENDRA_MAX_DOCUMENT_COUNT

integer

Number of documents returned by Amazon Kendra fallback

ALT_SEARCH_KENDRA_FAQ_MESSAGE

string

Heading when a Frequently Asked Question is found by Amazon Kendra.

ALT_SEARCH_KENDRA_TOP_ANSWER_MESSAGE

string

Heading when the top answer is found by Amazon Kendra

ALT_SEARCH_KENDRA_ANSWER_MESSAGE

string

Heading when a Document is returned by Amazon Kendra

ALT_SEARCH_KENDRA_RESPONSE_TYPES

Comma separated list. One or more valid Amazon Kendra response type

Kendra fallback will only return responses of the listed types

ALT_SEARCH_KENDRA_ABBREVIATE_MESSAGE_FOR_SSML

boolean

If a set to “true”, an abbreviate Amazon Kendra response will be sent via voice. If set to “false”, the full text of the Kendra fallback response will be sent when using voice.

KENDRA_FAQ_INDEX

Kendra Index Id

Amazon Kendra Index to use sync Elastic Search questions and answers

KENDRA_FAQ_CONFIG_MAX_RETRIES

integer

Number of times to retry syncing FAQ’s when a throttling error occurs

KENDRA_FAQ_CONFIG_RETRY_DELAY

integer

Amount of time to wait in seconds between attempts to retry syncing

KENDRA_FAQ_ES_FALLBACK

true or false

When Kendra FAQ is enabled, but does not return an answer then query ElasticSearch

ENABLE_KENDRA_WEB_INDEXER

true or false

Enables the web indexer

KENDRA_INDEXER_URLS

comma separated list

List of web addresses AgentX should crawl and index with Kendra

KENDRA_INDEXER_SCHEDULE

CloudWatch Rate Syntax

Interval Indexer should crawl

KENDRA_INDEXER_CRAWL_DEPTH

number

Sets the depth to the number of levels in a website from the seed level that you want to crawl

KENDRA_INDEXER_CRAWL_MODE

HOST_ONLY | SUBDOMAINS | EVERYTHING

Determines which addresses should be crawled

KENDRA_WEB_PAGE_INDEX

Kendra Index Id

The index to use for the web crawler, a custom data source will automatically be added to the specified index.

KENDRA_INDEXED_DOCUMENTS_LANGUAGES

comma separate list

Should be one of supported Kendra languages mentioned in documentation

ERRORMESSAGE

text

Response to the user when a processing error occurs

EMPTYMESSAGE

text

Response to the user when an answer could not be found

DEFAULT_ALEXA_LAUNCH_MESSAGE

text

Initial greeting when using Alexa

DEFAULT_ALEXA_REPROMPT

text

Default text used for Alexa reprompt capability

DEFAULT_ALEXA_STOP_MESSAGE

text

User response to end session with Alexa

SMS_HINT_REMINDER_ENABLE

true or false

Enables SMS_HINT_REMINDER

SMS_HINT_REMINDER

text

Reminds user how to use the bot on first use after SMS_HINT_REMINDER_INTERVAL_HRS

SMS_HINT_REMINDER_INTERVAL_HRS

integer

The amount of time in hours when to send SMS_HINT_REMINDER

IDENTITY_PROVIDER_JWKS_URLS

array of urls

User can override this empty list to add trusted IdPs (eg from Lex-Web-UI CognitoUserPoolPubKey)

ENFORCE_VERIFIED_IDENTITY

true or false

Set to true to make AgentX require verified identity from client

NO_VERIFIED_IDENTITY_QUESTION

text

If user identity cannot be verified, replace question string with this.

ELICIT_RESPONSE_MAX_RETRIES

integer

Number of times an elicitResponse LexBot can be called before giving up when the Bot returns Failed

ELICIT_RESPONSE_RETRY_MESSAGE

text

Default retry message when working with LexBot

ELICIT_RESPONSE_BOT_FAILURE_MESSAGE

text

Message used when maximum number of retries is exceeded

ELICIT_RESPONSE_DEFAULT_MSG

text

CONNECT_IGNORE_WORDS

comma separated list

Throw an error if the transcript provided by connect only contains the words in this list (case insensitive). This is useful if you find many missed utterances due to the use of filler words before a proper utterance (e.g. “a”, “e”, “umm”, “like”, etc.). This setting can not be used as a transcript filter (see LAMBDA_PREPROCESS_HOOK or LAMBDA_POSTPROCESS_HOOK if you wish to apply custom processing to questions/answers)

CONNECT_ENABLE_VOICE_RESPONSE_INTERRUPT

true or false

Return bot response in session attribute to enable contact flow to use response as an interruptible prompt.

CONNECT_NEXT_PROMPT_VARNAME

text

Name of session var to use for next prompt

ENABLE_REDACTING

true or false

Enable the system to redact log output

ENABLE_REDACTING_WITH_COMPREHEND

true or false

Enables Amazon Comprehend based PII Redacting

COMPREHEND_REDACTING_ENTITY_TYPES

comma separated list

A list of PII Entity Types

COMPREHEND_REDACTING_CONFIDENCE_SCORE

number (0 to .99)

Only redact PII where Amazon Comprehend’s confidence score is greater than this number

REDACTING_REGEX

regex expression

Redacts expressions matching regex from logs

PII_REJECTION_ENABLED

true or false

Enables PII Rejection

PII_REJECTION_QUESTION

text

If PII is found, the user’s request (question) will change to this phrase

PII_REJECTION_REGEX

regex expression

Used to find PII based on a regex

PII_REJECTION_ENTITY_TYPES

comma separated list of PII Entity Categories

Only recognize PII entity types in the list

PII_REJECTION_CONFIDENCE_SCORE

number (0 to 0.99)

Only reject PII where Amazon Comprehend’s confidence score is greater than this number

DISABLE_CLOUDWATCH_LOGGING

true or false

Disable all logging in fulfillment es query handler lambda. does not disable logging from Lambda Hooks or Conditional Chaining Lambda functions

MINIMAL_ES_LOGGING

true or false

do not log utterances or session attributes to ElasticSearch for Kibana logging

S3_PUT_REQUEST_ENCRYPTION

text

enable header x-amz-server-side-encryption header and set with this value

BOT_ROUTER_WELCOME_BACK_MSG

text

The text used by AgentX when ending communication from a specialty bot

BOT_ROUTER_EXIT_MSGS

comma separated list

The exit phrases in comma separated list available for the a user to end communication with a specialty bot

RUN_LAMBDAHOOK_FROM_QUERY_STEP

true or false

Controls timing of execution for Lambda hooks

LAMBDA_PREPROCESS_HOOK

string

name of AWS Lambda to run before each question is processed. The name of the Lambda must start with “qna-” or “QNA-” to comply with the permissions of the role attached to the Fulfillment Lambda .

LAMBDA_POSTPROCESS_HOOK

string

name of AWS Lambda to run after the question is processed. But before user profile information is saved. The name of the Lambda must start with “qna-” or “QNA-” to comply with the permissions of the role attached to the Fulfillment Lambda .

SEARCH_REPLACE_QUESTION_SUBSTRINGS

string

replace words or phrases in user questions by defining search/replace pairs in a JSON object like: {“searchString”:”replaceString”}. Add additional pairs separated by commas, like: {“searchString”:”replaceString”, “searchString2”:”replaceString2”}.

EMBEDDINGS_ENABLE

true or false

Disable use of semantic search using embeddings. Set to TRUE only if AgentX stack was deployed with embeddings enabled.

EMBEDDINGS_SCORE_THRESHOLD

0-1

Unlike regular elasticsearch queries, embeddings queries always return scores between 0 and 1, so we can apply a threshold to separate good from bad results. If embedding similarity score is under threshold the match it’s rejected and AgentX reverts to try to find a match on the answer field (only if ES_SCORE_ANSWER_FIELD is set to TRUE), Text item passage query, Kendra fallback or no_hits. Use the Content Designer TEST tab to see the hits ranked by score for your query results.

EMBEDDINGS_SCORE_ANSWER_THRESHOLD

0-1

Used only when ES_SCORE_ANSWER_FIELD is TRUE. If embedding similarity score for answer field query is under threshold the match is rejected and AgentX reverts to Text item passage query, Kendra fallback or no_hits. Use the Content Designer TEST tab to see the hits ranked by score for your answer field query results. For Match on, choose qna item answer to see answer field scores.

EMBEDDINGS_TEXT_PASSAGE_SCORE_THRESHOLD

0-1

If embedding similarity score for text item passage field query is under threshold the match it’s rejected and AgentX reverts to Kendra fallback or no_hits. Use the Content Designer TEST tab to see the hits ranked by score for your answer field query results. For Match on, choose text item passage to see passage field scores.

LLM_API

SAGEMAKER, LAMBDA

Specifies the LLM (Language Model) API used by the AgentX Stack based on the chosen value during deployment or update.

LLM_GENERATE_QUERY_ENABLE

true or false

Enables or disables question disambiguation using the LLM model.

LLM_GENERATE_QUERY_PROMPT_TEMPLATE

Prompt template with placeholders: {history}, {input}

The template used to construct a prompt for LLM to disambiguate follow-up questions. It can use placeholders to provide conversational context and the current user utterance/question.

LLM_GENERATE_QUERY_MODEL_PARAMS

JSON string, e.g. {"temperature":0}

Parameters sent to the LLM model when disambiguating follow-up questions.

LLM_QA_ENABLE

true or false

Enables or disables generative answers from passages retrieved via embeddings or Kendra fallback when no FAQ match is found. Applied only to passages and Kendra results - does not apply when an FAQ/QID matches the question.

LLM_QA_USE_KENDRA_RETRIEVAL_API

true or false

Enables or disables use of Kenda’s new retrieval API. When enabled, AgentX uses Kendra Retrieve to retrieve semantically relevant passages of up to 200 token words from the documents (not FAQs) in your index. When disabled, AgentX uses Kendra Query to retrieve shorter passages or answers. Takes effect only when LLM_QA_ENABLE is true. The default is true (recommended) when LLM QA is enabled. See https://docs.aws.amazon.com/kendra/latest/APIReference/API_Retrieve.html

LLM_QA_PROMPT_TEMPLATE

Prompt template with placeholders: {context}, {history}, {input}, {query}

The template used to construct a prompt for LLM to generate an answer from the context of retrieved passages (from Kendra or Text Item passages).

LLM_QA_NO_HITS_REGEX

Regular expression pattern

If the LLM response matches the specified pattern (e.g., “Sorry, I don’t know”), the response is treated as no_hits, and the default EMPTYMESSAGE or a custom ‘no_hits’ item is returned instead. Disabled by default, since enabling it prevents easy debugging of LLM don’t know responses.

LLM_QA_MODEL_PARAMS

JSON string, e.g. {"temperature":0}

Parameters sent to the LLM model when generating answers to questions.

LLM_QA_PREFIX_MESSAGE

Message used to prefix LLM-generated answer

May be empty.

LLM_QA_SHOW_CONTEXT_TEXT

true or false

Enables or disables inclusion of the passages used as context for LLM-generated answers.

LLM_QA_SHOW_SOURCE_LINKS

true or false

Enables or disables Kendra Source Links or passage refMarkdown links (document references) in markdown answers.

LLM_CHAT_HISTORY_MAX_MESSAGES

Positive integer

Specifies the maximum number of previous messages maintained in the AgentX DynamoDB UserTable for conversational context and follow-up question disambiguation.

LLM_PROMPT_MAX_TOKEN_LIMIT

Positive integer

Specifies the maximum number of tokens in the prompt message that can be sent to the LLM. AgentX will selectively truncate the prompt by history and context to shorten the total length.