AI

Model calls, classification, safety, embeddings and vector search. Each function lists its parameters, defaults and return value.

LLM Namespace

LLM chat and completion functions.

Scope: System

Functions

Function	Description
llm.chat()	LLM chat completion with multi-turn messages
llm.complete()	LLM text completion with a single prompt

llm.chat()

Perform LLM chat completion with multi-turn message history.

Note: This function requires a pipeline environment with ProxySelector for model routing. Use the YAML interface (chat task with type: chat) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
messages	array	—	Chat messages with role and content fields (required)
selection	map	{}	Model selection criteria
temperature	float	1.0	Sampling temperature
maxtokens	int	200	Maximum tokens to generate

Returns

Field	Type	Description
success	bool	Whether the completion succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = llm.chat({
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "What is machine learning?" }
  ],
  selection: { model: "gpt-4" },
  maxtokens: 500
})
// Use the chat YAML task with type: chat instead

llm.complete()

Perform LLM text completion with a single prompt.

Note: This function requires a pipeline environment with ProxySelector for model routing. Use the YAML interface (chat task with type: completion) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
prompt	string	—	Prompt text for completion (required)
selection	map	{}	Model selection criteria
temperature	float	1.0	Sampling temperature
maxtokens	int	200	Maximum tokens to generate

Returns

Field	Type	Description
success	bool	Whether the completion succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = llm.complete({
  prompt: "Summarize the following text: ...",
  selection: { accuracy: "high" },
  temperature: 0.7,
  maxtokens: 300
})
// Use the chat YAML task with type: completion instead

Intent Namespace

Intent classification functions.

Scope: System

Functions

Function	Description
intent.classify()	Classify the intent of a user prompt

intent.classify()

Classify the intent of user prompts by sending them to an intent classification model endpoint.

Note: This function requires a pipeline environment with ProxySelector for model routing. Use the YAML interface (getintent task) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
prompt	string	—	Prompt to classify (required)
selection	map	{}	Model selection criteria

Returns

Field	Type	Description
success	bool	Whether the classification succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = intent.classify({
  prompt: "I want to book a flight to New York"
})
// Use the getintent YAML task instead

Guard Namespace

LLM safety classification functions.

Scope: System

Functions

Function	Description
guard.check()	Classify content safety using Llama Guard

guard.check()

Classify user prompts or agent responses against unsafe content categories (S1–S11).

Note: This function requires a pipeline environment with ProxySelector for model routing. Use the YAML interface (pre-guard/post-guard tasks) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
prompt	string	—	Content to classify (required)

Returns

Field	Type	Description
success	bool	Whether the classification completed
error	string	Error message (stub always returns error)

Unsafe Content Categories

Category	Description
S1	Violent Crimes
S2	Non-Violent Crimes
S3	Sex Crimes
S4	Child Exploitation
S5	Specialized Advice
S6	Privacy
S7	Intellectual Property
S8	Indiscriminate Weapons
S9	Hate
S10	Self-Harm
S11	Sexual Content

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = guard.check({
  prompt: "Is this message safe?"
})
// Use the pre-guard/post-guard YAML tasks instead

Embed Namespace

Embedding generation and text splitting functions.

Scope: System

Functions

Function	Description
embed.generate()	Generate vector embeddings (requires pipeline)
embed.split()	Split text data by delimiter or by length

embed.generate()

Generate vector embeddings from text data.

Note: This function requires a pipeline environment with ProxySelector for model routing. Use the YAML interface (embed or embeddata tasks) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
data	array	—	Array of text strings to embed (required)

Returns

Field	Type	Description
success	bool	Whether embedding generation succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = embed.generate({
  data: ["Hello world", "Machine learning is great"]
})
// Use the embed or embeddata YAML tasks instead

embed.split()

Split text data into chunks by delimiter or by character length. Works independently without requiring a pipeline environment.

Parameters

Parameter	Type	Default	Description
data	array	—	Array of text strings to split (required)
delimiter	string	—	Delimiter string to split by
length	int	—	Character length to split into chunks

One of delimiter or length is required.

Returns

Field	Type	Description
success	bool	Whether the split succeeded
result	array	Array of arrays — one split-result array per input string
error	string	Error message if failed

Examples

// Split by delimiter
let result = embed.split({
  data: ["Hello world. This is a test.", "Another document here."],
  delimiter: ". "
})
// result.result = [["Hello world", "This is a test."], ["Another document here."]]

// Split by character length
let result = embed.split({
  data: ["A long text that needs to be split into smaller chunks for embedding"],
  length: 20
})
// result.result = [["A long text that nee", "ds to be split into ", "smaller chunks for e", "mbedding"]]

// Split paragraphs for embedding prep
let result = embed.split({
  data: [
    "First paragraph content.\n\nSecond paragraph content.\n\nThird paragraph.",
    "Another document.\n\nWith two paragraphs."
  ],
  delimiter: "\n\n"
})
for (let i = 0; i < result.result.length; i++) {
  log("Document " + i + " has " + result.result[i].length + " chunks")
}

AI Knowledge Namespace

AI Knowledge vector store functions.

Scope: System

Functions

Function	Description
ai-knowledge.embed()	Send embeddings to a AI Knowledge vector store REST endpoint

ai-knowledge.embed()

Send embeddings and associated data to a AI Knowledge vector store REST endpoint.

Note: This function requires a pipeline environment with REST service connection. Use the YAML interface (ai-knowledge task) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
url	string	—	AI Knowledge REST endpoint URL (required)
embeddingskey	string	—	Environment key for embeddings data (required)

Returns

Field	Type	Description
success	bool	Whether the operation succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = ai-knowledge.embed({
  url: "https://ai-knowledge.example.com/v1/documents",
  embeddingskey: "embeddings"
})
// Use the ai-knowledge YAML task instead

Milvus Namespace

Milvus vector database functions. Provides insert, search, query, list, and delete operations for vector collections with tenant-based database isolation.

Scope: System

Functions

Function	Description
milvus.insert()	Insert vector embeddings into a collection
milvus.search()	Vector similarity search in a collection
milvus.query()	Query records by field filters
milvus.list()	List unique topic names for a user
milvus.delete()	Delete records by user/topic/URL

Note: All Milvus functions require a pipeline environment with Milvus client connection and tenant context. Use the YAML interface (topics tasks) for full functionality. The script namespace stubs return errors directing to the YAML interface.

milvus.insert()

Insert vector embeddings into a Milvus collection.

Parameters

Parameter	Type	Default	Description
url	string	—	Milvus gRPC endpoint URL (required)
collection	string	—	Collection name (required)
embeddings	string	—	Environment key for embeddings data (required)
userid	string	—	User ID for inserted records (required)
topic	string	—	Topic name for inserted records (required)

Returns

Field	Type	Description
success	bool	Whether the insert succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = milvus.insert({
  url: "localhost:19530",
  collection: "knowledge_base",
  embeddings: "embeddingresult",
  userid: "user123",
  topic: "machine-learning"
})
// Use the topicsinsert YAML task instead

milvus.search()

Perform vector similarity search using L2 distance.

Parameters

Parameter	Type	Default	Description
url	string	—	Milvus gRPC endpoint URL (required)
collection	string	—	Collection name (required)
embeddings	string	—	Environment key for query embedding (required)
topic	string	—	Topic name to filter by (required)
userid	string	—	User ID to filter by
limit	int	3	Maximum results (top-k)

Returns

Field	Type	Description
success	bool	Whether the search succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = milvus.search({
  url: "localhost:19530",
  collection: "knowledge_base",
  embeddings: "queryembedding",
  topic: "machine-learning",
  limit: 5
})
// Use the topicssearch YAML task instead

milvus.query()

Query records from a collection using field-based filters.

Parameters

Parameter	Type	Default	Description
url	string	—	Milvus gRPC endpoint URL (required)
collection	string	—	Collection name (required)
topic	string	—	Topic name to query by (required)
userid	string	—	User ID to filter by
limit	int	3	Maximum results

Returns

Field	Type	Description
success	bool	Whether the query succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = milvus.query({
  url: "localhost:19530",
  collection: "knowledge_base",
  topic: "machine-learning",
  limit: 10
})
// Use the topicsquery YAML task instead

milvus.list()

List unique topic names for a user from a collection.

Parameters

Parameter	Type	Default	Description
url	string	—	Milvus gRPC endpoint URL (required)
collection	string	—	Collection name (required)
userid	string	—	User ID to filter by (required)

Returns

Field	Type	Description
success	bool	Whether the listing succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = milvus.list({
  url: "localhost:19530",
  collection: "knowledge_base",
  userid: "user123"
})
// Use the topicslist YAML task instead

milvus.delete()

Delete records from a collection by user, topic, and optional URL.

Parameters

Parameter	Type	Default	Description
url	string	—	Milvus gRPC endpoint URL (required)
collection	string	—	Collection name (required)
userid	string	—	User ID to delete by (required)
topic	string	—	Topic name to delete by (required)

Returns

Field	Type	Description
success	bool	Whether the delete succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = milvus.delete({
  url: "localhost:19530",
  collection: "knowledge_base",
  userid: "user123",
  topic: "old-topic"
})
// Use the topicsdelete YAML task instead

Text Namespace

Text transformation functions.

Scope: System

Functions

Function	Description
text.tosentence()	Convert entity YAML to natural language sentences (requires pipeline)
text.split()	Split text by delimiter or by length

text.tosentence()

Convert entity YAML definitions into natural language sentences.

Note: This function requires entity YAML context from the pipeline environment. Use the YAML interface (tosentence task) for full functionality. The script namespace stub returns an error directing to the YAML interface.

Parameters

Parameter	Type	Default	Description
entityyaml	string	—	Entity YAML string to convert (required)

Returns

Field	Type	Description
success	bool	Whether the conversion succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = text.tosentence({
  entityyaml: "name: User\nfields:\n  - name: email\n    type: string"
})
// Use the tosentence YAML task instead

text.split()

Split text by a delimiter string or by character length.

Parameters

Parameter	Type	Default	Description
text	string	—	Text to split (required)
delimiter	string	—	Delimiter string to split by
length	int	—	Character length to split into chunks

One of delimiter or length is required.

Returns

Field	Type	Description
success	bool	Whether the split succeeded
result	array	Array of text chunks
error	string	Error message if failed

Examples

// Split by delimiter
let result = text.split({
  text: "hello world. this is a test. another sentence.",
  delimiter: ". "
})
// result.result = ["hello world", "this is a test", "another sentence."]

// Split by length
let result = text.split({
  text: "The quick brown fox jumps over the lazy dog",
  length: 10
})
// result.result = ["The quick ", "brown fox ", "jumps over", " the lazy ", "dog"]

// Split paragraphs
let result = text.split({
  text: "First paragraph.\n\nSecond paragraph.\n\nThird paragraph.",
  delimiter: "\n\n"
})
for (let part of result.result) {
  log("Paragraph: " + part)
}

Validate Namespace

JSON and YAML schema validation functions.

Scope: System

Functions

Function	Description
validate.json()	Validate a JSON value against a JSON schema
validate.yaml()	Validate a YAML value against a JSON schema

validate.json()

Validate a JSON value against a JSON schema using the JSON Schema specification.

Parameters

Parameter	Type	Default	Description
schema	string	—	JSON schema string (required)
value	string	—	JSON value string to validate (required)

Returns

Field	Type	Description
success	bool	Whether the validation completed
valid	bool	Whether the value matches the schema
error	string	Validation error details if invalid, or error message if failed

Examples

// Validate a JSON object
let result = validate.json({
  schema: '{"type": "object", "properties": {"name": {"type": "string"}, "age": {"type": "integer"}}, "required": ["name"]}',
  value: '{"name": "Alice", "age": 30}'
})
if (result.success && result.valid) {
  log("JSON is valid")
}

// Check for validation errors
let result = validate.json({
  schema: '{"type": "object", "required": ["email"]}',
  value: '{"name": "Bob"}'
})
if (result.success && !result.valid) {
  log("Validation error: " + result.error)
}

// Validate an array
let result = validate.json({
  schema: '{"type": "array", "items": {"type": "number"}}',
  value: '[1, 2, 3]'
})
log("Valid: " + result.valid)

validate.yaml()

Validate a YAML value against a JSON schema. The YAML is parsed and then validated using the JSON Schema specification.

Parameters

Parameter	Type	Default	Description
schema	string	—	JSON schema string (required)
value	string	—	YAML value string to validate (required)

Returns

Field	Type	Description
success	bool	Whether the validation completed
valid	bool	Whether the value matches the schema
error	string	Validation error details if invalid, or error message if failed

Examples

// Validate a YAML config
let result = validate.yaml({
  schema: '{"type": "object", "properties": {"name": {"type": "string"}, "version": {"type": "string"}}, "required": ["name"]}',
  value: "name: my-app\nversion: 1.0.0"
})
if (result.success && result.valid) {
  log("YAML config is valid")
}

// Validate YAML against a schema
let result = validate.yaml({
  schema: '{"type": "object", "properties": {"port": {"type": "integer"}}, "required": ["port"]}',
  value: "host: localhost\nport: 8080"
})
log("Valid: " + result.valid)

// Check for missing required fields
let result = validate.yaml({
  schema: '{"type": "object", "required": ["name", "email"]}',
  value: "name: Alice"
})
if (!result.valid) {
  log("Missing fields: " + result.error)
}

Scrape Namespace

Web and document scraping functions.

Scope: System

Functions

Function	Description
scrape.web()	Scrape content from web URLs
scrape.pdf()	Scrape text from PDF files
scrape.excel()	Scrape content from Excel files
scrape.docx()	Scrape text from DOCX files
scrape.pptx()	Scrape text from PPTX files

Note: All scrape functions require a pipeline environment with ProxySelector for service routing (except scrape.excel which processes files directly). Use the YAML interface (scrape tasks) for full functionality. The script namespace stubs return errors directing to the YAML interface.

scrape.web()

Scrape content from web URLs with configurable crawl depth and page limits.

Parameters

Parameter	Type	Default	Description
urls	array	—	URLs to scrape (required)

Returns

Field	Type	Description
success	bool	Whether the scrape succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = scrape.web({
  urls: ["https://docs.example.com/guide"]
})
// Use the scrapedata YAML task instead

scrape.pdf()

Scrape text content from PDF files via a remote scraping service.

Parameters

Parameter	Type	Default	Description
url	string	—	URL of the PDF file (required)

Returns

Field	Type	Description
success	bool	Whether the scrape succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = scrape.pdf({
  url: "https://example.com/document.pdf"
})
// Use the scrape-pdf YAML task instead

scrape.excel()

Scrape content from Excel files.

Parameters

Parameter	Type	Default	Description
file	string	—	URL of the Excel file (required)

Returns

Field	Type	Description
success	bool	Whether the scrape succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = scrape.excel({
  file: "https://example.com/data.xlsx"
})
// Use the scrape-excel YAML task instead

scrape.docx()

Scrape text content from DOCX files via a remote scraping service.

Parameters

Parameter	Type	Default	Description
url	string	—	URL of the DOCX file (required)

Returns

Field	Type	Description
success	bool	Whether the scrape succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = scrape.docx({
  url: "https://example.com/document.docx"
})
// Use the scrape-docx YAML task instead

scrape.pptx()

Scrape text content from PPTX (PowerPoint) files via a remote scraping service.

Parameters

Parameter	Type	Default	Description
url	string	—	URL of the PPTX file (required)

Returns

Field	Type	Description
success	bool	Whether the scrape succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = scrape.pptx({
  url: "https://example.com/presentation.pptx"
})
// Use the scrape-pptx YAML task instead

Pipeline Namespace

ML pipeline chat, request, and response functions for streaming multi-turn conversations within data pipelines.

Scope: System

Functions

Function	Description
pipeline.chat()	Stream pipeline ML chat with conversation memory
pipeline.request()	Stream pipeline request ingestion
pipeline.response()	Stream pipeline response delivery

Note: All pipeline functions require a pipeline environment with ProxySelector, pipe, and record context. Use the YAML interface (mlpipeline tasks) for full functionality. The script namespace stubs return errors directing to the YAML interface.

pipeline.chat()

Stream pipeline ML chat. Reads incoming data records, builds chat messages from conversation memory, and delegates to the chat task.

Parameters

Parameter	Type	Default	Description
selection	map	{}	Model selection criteria
promptkey	string	—	Record field name containing the user prompt

Returns

Field	Type	Description
success	bool	Whether the chat succeeded
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = pipeline.chat({
  selection: { model: "gpt-4" },
  promptkey: "message"
})
// Use the mlpipelinechat/mlpipechat YAML task instead

pipeline.request()

Stream pipeline request ingestion. Reads incoming data records and forwards them with conversation context.

Parameters

No additional parameters required beyond pipeline wiring.

Returns

Field	Type	Description
success	bool	Whether the request was processed
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = pipeline.request({})
// Use the mlpipelinerequest/mlpiperequest YAML task instead

pipeline.response()

Stream pipeline response delivery. Reads processed records and submits them to an external URL via HTTP POST.

Parameters

Parameter	Type	Default	Description
url	string	—	URL to submit the response to (required)

Returns

Field	Type	Description
success	bool	Whether the response was delivered
error	string	Error message (stub always returns error)

Examples

// Note: Requires pipeline environment. This stub returns an error.
let result = pipeline.response({
  url: "https://api.example.com/v1/conversation/response"
})
// Use the mlpipelineresponse/mlpiperesponse YAML task instead