Model operators

The Model family contains 5 operators. Direct LLM calls — responses, classification, extraction, embeddings, reranking.

Implementation status

The model family handler is in flight — some operators may run as pass-throughs and emit an operator-not-implemented warning until their handler lands. See The operator contract.

Operators in this family

• model.classify
• model.embed
• model.extract
• model.rerank
• model.respond

model.classify

Classifies text into predefined categories using an LLM.

Version 1.0.0 · Semantics passthrough · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
labels	string[]	yes	—	Classification labels.
max_tokens	integer	no	4096
model	string	yes	—	Model identifier (e.g. ‘deepseek/deepseek-r1’).
multi_label	boolean	no	false
stop_sequences	string[]	no	—
temperature	number	no	0.7
top_p	number	no	—

Input ports

Port	Type	Required	Description
input	string	yes	Input

Output ports

Port	Type	Required	Description
output	any	no	Classification
usage	object	no	Usage

model.embed

Generates vector embeddings from text.

Version 1.0.0 · Semantics passthrough · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
dimensions	integer	no	—
model	string	yes	—	Embedding model identifier.

Input ports

Port	Type	Required	Description
input	string	yes	Input

Output ports

Port	Type	Required	Description
output	number[]	no	Embedding
usage	object	no	Usage

model.extract

Extracts structured data from text using an LLM with a JSON schema.

Version 1.0.0 · Semantics passthrough · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
max_tokens	integer	no	4096
model	string	yes	—	Model identifier (e.g. ‘deepseek/deepseek-r1’).
output_schema	object	yes	—	JSON Schema defining the extraction target.
stop_sequences	string[]	no	—
temperature	number	no	0.7
top_p	number	no	—

Input ports

Port	Type	Required	Description
input	string	yes	Input

Output ports

Port	Type	Required	Description
output	any	no	Extracted
usage	object	no	Usage

model.rerank

Reranks documents by relevance to a query.

Version 1.0.0 · Semantics passthrough · Visibility advanced · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
model	string	yes	—
top_k	integer	no	10

Input ports

Port	Type	Required	Description
documents	array	yes	Documents
query	string	yes	Query

Output ports

Port	Type	Required	Description
output	array	no	Ranked

model.respond

Sends a prompt to an LLM and returns generated text.

Version 1.0.0 · Semantics passthrough · Secrets OPENROUTER_API_KEY

Config

Field	Type	Required	Default	Description
max_tokens	integer	no	4096
model	string	yes	—	Model identifier (e.g. ‘deepseek/deepseek-r1’).
response_format	”text” | “json_object”	no	"text"
stop_sequences	string[]	no	—
system_prompt	string	no	—
temperature	number	no	0.7
top_p	number	no	—

Input ports

Port	Type	Required	Description
prompt	any	yes	Prompt

Output ports

Port	Type	Required	Description
output	string	no	Text
usage	object	no	Usage

Source schemas

Every operator above is rendered from its JSON schema in /opt/glyph/schemas/operators/:

• schemas/operators/model.classify.json
• schemas/operators/model.embed.json
• schemas/operators/model.extract.json
• schemas/operators/model.rerank.json
• schemas/operators/model.respond.json