Functions

Functions available in the query language.

abs

abs(expr)

Returns the absolute value of the input.

add

add(expr1, expr2)

Adds the two inputs.

and

and(expr1, expr2)

Logical and operation of two boolean columns.

character_count

character_count(text)

Returns the number of characters in a text column.

• Aliases

  • num_chars

coalesce

coalesce(expr)

Return the first expression that evaluates to a non-null value.

concat

concat(expr)

Concatenates multiple text columns into one.

contains

contains(text, text)

Returns true if the input string contains the substring.

count

count(expr)

Computes the number of rows in a column.

count_distinct

count_distinct(expr)

Computes the number of distinct non-null values in a column.

count_if

count_if(expr)

Computes the number of rows in a column that satisfy a condition.

date_trunc

date_trunc(expr1, expr2)

Truncates a timestamp to the specified unit.

deterministic_sample

deterministic_sample(expr)

Returns a deterministic sample value in [0, 1) based on the input value.

divide

divide(expr1, expr2)

Divides the two inputs.

embed

embed(text)

Returns the embedding of a text column. Embedding model: all-mpnet-base-v2.

equal_to

equal_to(expr1, expr2)

Computes the element-wise equal to comparison of two columns.

• Aliases

  • eq

filter

filter(expr1, expr2)

Filters a column using another column as a mask.

greater_than

greater_than(expr1, expr2)

Computes the element-wise greater than comparison of two columns. input1 > input2

• Aliases

  • gt

greater_than_or_equal_to

greater_than_or_equal_to(expr1, expr2)

Computes the element-wise greater than or equal to comparison of two columns. input1 >= input2

• Aliases

  • gte

icontains

icontains(text, text)

Returns true if the input string contains the substring, ignoring case.

is_valid_json

is_valid_json(text)

Returns true if the input string is valid json.

less_than

less_than(expr1, expr2)

Computes the element-wise less than comparison of two columns. input1 < input2

• Aliases

  • lt

less_than_or_equal_to

less_than_or_equal_to(expr1, expr2)

Computes the element-wise less than or equal to comparison of two columns. input1 <= input2

• Aliases

  • lte

levenshtein

levenshtein(output, reference)

Returns Damerau-Levenshtein distance between two strings.

list_contains

list_contains(list, value)

Returns True if the list contains the value.

list_extract

list_extract(list_expr, index_expr)

Extracts the item at the given index from a list.

list_has_duplicate

list_has_duplicate(expr)

Returns True if the list has duplicated items.

list_length

list_length(expr)

Returns the length of lists in a list column.

list_most_common

list_most_common(expr)

Most common item in list.

list_starts_with

list_starts_with(list, prefix)

Returns True if the list starts with the value.

list_zip

list_zip(expr)

Zips multiple lists into a list of structs.

llm_answer_groundedness

llm_answer_groundedness(model_name, prompt_version, answer, context)

Classifies whether the generated answer is grounded in and supported by the provided context.

llm_answer_groundedness_with_justification

llm_answer_groundedness_with_justification(model_name, prompt_version, answer, context)

Classifies whether the generated answer is grounded in and supported by the provided context.

llm_answer_refusal

llm_answer_refusal(model_name, prompt_version, answer)

Classifies whether the model refused to answer the user's question.

llm_answer_refusal_with_justification

llm_answer_refusal_with_justification(model_name, prompt_version, answer)

Classifies whether the model refused to answer the user's question.

llm_answer_relevancy

llm_answer_relevancy(model_name, prompt_version, question, answer)

Classifies whether the generated answer is relevant and responsive to the user's question.

• Aliases

  • rag_answer_relevancy

llm_answer_relevancy_with_justification

llm_answer_relevancy_with_justification(model_name, prompt_version, question, answer)

Classifies whether the generated answer is relevant and responsive to the user's question.

llm_classify

llm_classify(model_name, prompt, classes)

Classifies text into custom categories you define, using your own prompt and labels.

llm_classify_with_justification

llm_classify_with_justification(model_name, prompt, classes)

Classifies text into custom categories you define, using your own prompt and labels.

llm_context_relevancy

llm_context_relevancy(model_name, prompt_version, question, context)

Classifies whether the retrieved context is relevant to the user's question.

llm_context_relevancy_with_justification

llm_context_relevancy_with_justification(model_name, prompt_version, question, context)

Classifies whether the retrieved context is relevant to the user's question.

llm_conversation_summary

llm_conversation_summary(model_name, prompt_version, conversation)

Generates a concise summary of a full conversation session between an AI assistant and a user.

llm_question_clarity

llm_question_clarity(model_name, prompt_version, question)

Scores how clear and well-formed a question is, from 1 (ambiguous or incoherent) to 5 (perfectly clear).

llm_question_clarity_with_justification

llm_question_clarity_with_justification(model_name, prompt_version, question)

Scores how clear and well-formed a question is, from 1 (ambiguous or incoherent) to 5 (perfectly clear).

llm_score

llm_score(model_name, prompt)

Scores text on a 1–5 scale using your own custom evaluation prompt.

llm_score_with_justification

llm_score_with_justification(model_name, prompt)

Scores text on a 1–5 scale using your own custom evaluation prompt.

llm_summarization

llm_summarization(model_name, prompt_version, input, output)

Generates a concise summary of a single conversational exchange (input and output).

llm_text_frustration

llm_text_frustration(model_name, prompt_version, text)

Scores the level of user frustration expressed in a text, from 1 (not frustrated) to 5 (extremely frustrated).

llm_text_frustration_with_justification

llm_text_frustration_with_justification(model_name, prompt_version, text)

Scores the level of user frustration expressed in a text, from 1 (not frustrated) to 5 (extremely frustrated).

llm_text_sentiment

llm_text_sentiment(model_name, prompt_version, text)

Classifies the overall sentiment of a text as positive, negative, or neutral.

• Aliases

  • text_sentiment

llm_text_sentiment_with_justification

llm_text_sentiment_with_justification(model_name, prompt_version, text)

Classifies the overall sentiment of a text as positive, negative, or neutral.

llm_text_similarity

llm_text_similarity(model_name, prompt_version, output, reference)

Scores how semantically similar an output is to a target reference, from 1 (completely different) to 5 (equivalent).

• Aliases

  • text_similarity

llm_text_similarity_with_justification

llm_text_similarity_with_justification(model_name, prompt_version, output, reference)

Scores how semantically similar an output is to a target reference, from 1 (completely different) to 5 (equivalent).

llm_text_toxicity

llm_text_toxicity(model_name, prompt_version, text)

Scores how toxic or harmful a piece of text is, from 1 (not toxic) to 5 (highly toxic).

llm_text_toxicity_with_justification

llm_text_toxicity_with_justification(model_name, prompt_version, text)

Scores how toxic or harmful a piece of text is, from 1 (not toxic) to 5 (highly toxic).

llm_user_frustration

llm_user_frustration(model_name, prompt_version, conversation)

Scores the overall user frustration across a conversation session, from 1 (satisfied) to 5 (extremely frustrated).

llm_user_frustration_with_justification

llm_user_frustration_with_justification(model_name, prompt_version, conversation)

Scores the overall user frustration across a conversation session, from 1 (satisfied) to 5 (extremely frustrated).

map_extract

map_extract(map_expr, key_expr)

Extracts the value for a given key from a map, returning null if the key is not in the map.

max

max(expr)

Computes the max of a column.

mean

mean(expr)

Computes the mean of a column.

median

median(expr)

Computes the median of a column.

min

min(expr)

Computes the min of a column.

mode

mode(expr)

Computes the mode of a column.

multiply

multiply(expr1, expr2)

Multiplies the two inputs.

negate

negate(expr)

Returns the negation of the input.

not

not(expr)

Logical not operation of a boolean column.

not_equal_to

not_equal_to(expr1, expr2)

Computes the element-wise not equal to comparison of two columns.

• Aliases

  • neq

or

or(expr1, expr2)

Logical or operation of two boolean columns.

percentile

percentile(expr1, expr2)

Computes the nth percentile of a column.

rouge1

rouge1(output, reference)

Returns the rouge1 score between two columns.

rouge2

rouge2(output, reference)

Returns the rouge2 score between two columns.

rougeL

rougeL(output, reference)

Returns the rougeL score between two columns.

rougeLsum

rougeLsum(output, reference)

Returns the rougeLsum score between two columns.

stddev

stddev(expr)

Computes the sample standard deviation of a column.

struct_extract

struct_extract(struct_expr, field_name)

Extracts a field from a struct expression.

subtract

subtract(expr1, expr2)

Subtracts the two inputs.

sum

sum(expr)

Computes the sum of a column.