API Reference

Complete reference for all public classes in SDG Hub. Every block, flow component, and connector is documented with its fields, methods, and type signatures.

BaseBlock

Bases: BaseModel, ABC

from sdg_hub.core.blocks import BaseBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`Optional[str]`	`--`	Block type (e.g., 'llm', 'transform', 'parser', 'filtering')
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict

Methods

BaseBlock.from_configclassmethod

(cls, config: dict[str, Any]) -> 'BaseBlock'

Instantiate block from serialized config.

Parameter	Type	Default
`config`	`dict[str, Any]`	required

Returns BaseBlock

BaseBlock.generateabstract

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Subclass method to implement data generation logic.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

BaseBlock.get_config

(self) -> dict[str, Any]

Return only constructor arguments for serialization.

Returns dict[str, Any]

BaseBlock.get_info

(self) -> dict[str, Any]

Return a high-level summary of block metadata and config.

Returns dict[str, Any]

BaseBlock.normalize_input_colsclassmethod

(cls, v: str | list[str] | dict[str, Any] | None) -> list[str] | dict[str, Any]

Parameter	Type	Default
`v`	`str \| list[str] \| dict[str, Any] \| None`	required

Returns list[str] | dict[str, Any]

BaseBlock.normalize_output_colsclassmethod

(cls, v: str | list[str] | dict[str, Any] | None) -> list[str] | dict[str, Any]

Parameter	Type	Default
`v`	`str \| list[str] \| dict[str, Any] \| None`	required

Returns list[str] | dict[str, Any]

BlockRegistry

from sdg_hub.core.blocks import BlockRegistry

Registry for block classes with metadata and enhanced error handling.

Methods

BlockRegistry.categoriesclassmethod

(cls) -> list[str]

Get all available categories.

Returns list[str]

BlockRegistry.discover_blocksclassmethod

(cls) -> None

Print a Rich-formatted table of all available blocks.

BlockRegistry.list_blocksclassmethod

(cls, category: Optional[str] = None, *, grouped: bool = False, include_deprecated: bool = True) -> list[str] | dict[str, list[str]]

List registered blocks, optionally filtered by category.

Parameter	Type	Default
`category`	`Optional[str]`	`--`
`grouped`	`bool`	`False`
`include_deprecated`	`bool`	`True`

Returns list[str] | dict[str, list[str]]

BlockRegistry.registerclassmethod

(cls, block_name: str, category: str, description: str = '', deprecated: bool = False, replacement: Optional[str] = None)

Parameter	Type	Default
`block_name`	`str`	required
`category`	`str`	required
`description`	`str`	`""`
`deprecated`	`bool`	`False`
`replacement`	`Optional[str]`	`--`

BlockMetadata

from sdg_hub.core.blocks.registry import BlockMetadata

Metadata for registered blocks.

Fields

Name	Type	Default	Description
`name`	`str`	required	--
`block_class`	`type`	required	--
`category`	`str`	required	--
`description`	`str`	`""`	--
`deprecated`	`bool`	`False`	--
`replacement`	`Optional[str]`	`--`	--

Methods

BlockMetadata.__init__

(self, name: str, block_class: type, category: str, description: str = '', deprecated: bool = False, replacement: Optional[str] = None) -> None

Initialize self. See help(type(self)) for accurate signature.

Parameter	Type	Default
`name`	`str`	required
`block_class`	`type`	required
`category`	`str`	required
`description`	`str`	`""`
`deprecated`	`bool`	`False`
`replacement`	`Optional[str]`	`--`

LLMChatBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import LLMChatBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"llm"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`model`	`Optional[str]`	`--`	Model identifier in LiteLLM format
`api_key`	`Optional[pydantic.types.SecretStr]`	`--`	API key for the provider
`api_base`	`Optional[str]`	`--`	Base URL for the API
`async_mode`	`bool`	`False`	Whether to use async processing
`timeout`	`float`	`120`	Request timeout in seconds
`num_retries`	`int`	`6`	Number of retry attempts (uses LiteLLM's built-in retry mechanism)
`drop_params`	`bool`	`True`	Whether to drop unsupported parameters to prevent API errors
`max_completion_tokens`	`Optional[int]`	`--`	Maximum completion tokens (used by newer models like GPT-5 instead of max_tokens). When set, max_tokens is automatically excluded.

Methods

LLMChatBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate responses from the LLM.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

LLMChatBlock.model_post_init

(self, _LLMChatBlock__context) -> None

Initialize after Pydantic validation.

Parameter	Type	Default
`_LLMChatBlock__context`		required

LLMChatBlock.validate_single_input_colclassmethod

(cls, v)

Ensure exactly one input column.

Parameter	Type	Default
`v`		required

LLMChatBlock.validate_single_output_colclassmethod

(cls, v)

Ensure exactly one output column.

Parameter	Type	Default
`v`		required

PromptBuilderBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import PromptBuilderBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"llm_util"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`prompt_config_path`	`str`	required	Path to YAML file containing the Jinja template configuration
`format_as_messages`	`bool`	`True`	Whether to format output as chat messages
`prompt_template_config`	`Optional[prompt_builder_block.PromptTemplateConfig]`	`--`	Loaded prompt template configuration
`prompt_renderer`	`Optional[prompt_builder_block.PromptRenderer]`	`--`	Prompt renderer instance

Methods

PromptBuilderBlock.generate

(self, samples: pandas.DataFrame, **_kwargs: Any) -> pandas.DataFrame

Generate formatted output for all samples.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`_kwargs`	`Any`	required

Returns pandas.DataFrame

PromptBuilderBlock.model_post_init

(self, _PromptBuilderBlock__context: Any) -> None

Initialize the block after Pydantic validation.

Parameter	Type	Default
`_PromptBuilderBlock__context`	`Any`	required

PromptBuilderBlock.validate_single_output_colclassmethod

(cls, v)

Validate that exactly one output column is specified.

Parameter	Type	Default
`v`		required

ChatMessage

Bases: BaseModel

from sdg_hub.core.blocks.llm.prompt_builder_block import ChatMessage

Pydantic model for chat messages with proper validation.

Fields

Name	Type	Default	Description
`role`	`Literal[system, user, assistant, tool]`	required	--
`content`	`str`	required	--

Methods

ChatMessage.validate_content_not_emptyclassmethod

(cls, v: str) -> str

Ensure content is not empty or just whitespace.

Parameter	Type	Default
`v`	`str`	required

Returns str

MessageTemplate

Bases: BaseModel

from sdg_hub.core.blocks.llm.prompt_builder_block import MessageTemplate

Template for a chat message with Jinja2 template and original source.

Fields

Name	Type	Default	Description
`role`	`Literal[system, user, assistant, tool]`	required	--
`content_template`	`jinja2.environment.Template`	required	--
`original_source`	`str`	required	--

PromptTemplateConfig

from sdg_hub.core.blocks.llm.prompt_builder_block import PromptTemplateConfig

Self-contained class for loading and validating YAML prompt configurations.

Methods

PromptTemplateConfig.__init__

(self, config_path: str)

Initialize with path to YAML config file.

Parameter	Type	Default
`config_path`	`str`	required

PromptTemplateConfig.get_message_templates

(self) -> list[prompt_builder_block.MessageTemplate]

Return the compiled message templates.

Returns list[prompt_builder_block.MessageTemplate]

PromptRenderer

from sdg_hub.core.blocks.llm.prompt_builder_block import PromptRenderer

Handles rendering of message templates with variable substitution.

Methods

PromptRenderer.__init__

(self, message_templates: list[prompt_builder_block.MessageTemplate])

Initialize with a list of message templates.

Parameter	Type	Default
`message_templates`	`list[prompt_builder_block.MessageTemplate]`	required

PromptRenderer.get_required_variables

(self) -> set

Extract all required variables from message templates.

Returns set

PromptRenderer.render_messages

(self, template_vars: dict[str, Any]) -> list[prompt_builder_block.ChatMessage]

Render all message templates with the given variables.

Parameter	Type	Default
`template_vars`	`dict[str, Any]`	required

Returns list[prompt_builder_block.ChatMessage]

PromptRenderer.resolve_template_vars

(self, sample: dict[str, Any], input_cols) -> dict[str, Any]

Resolve template variables from dataset columns based on input_cols.

Parameter	Type	Default
`sample`	`dict[str, Any]`	required
`input_cols`		required

Returns dict[str, Any]

LLMResponseExtractorBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import LLMResponseExtractorBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"llm_util"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`extract_content`	`bool`	`True`	Whether to extract 'content' field from responses.
`extract_reasoning_content`	`bool`	`False`	Whether to extract 'reasoning_content' field from responses.
`extract_tool_calls`	`bool`	`False`	Whether to extract 'tool_calls' field from responses.
`expand_lists`	`bool`	`True`	Whether to expand list inputs into individual rows (True) or preserve lists (False).
`field_prefix`	`str`	`""`	Prefix to add to output field names (e.g., 'llm_' results in 'llm_content', 'llm_reasoning_content').

Methods

LLMResponseExtractorBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Subclass method to implement data generation logic.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

LLMResponseExtractorBlock.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

LLMResponseExtractorBlock.validate_extraction_configuration

(self)

Validate that at least one extraction field is enabled and pre-compute field names.

JSONParserBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import JSONParserBlock

Block for parsing JSON from text and expanding fields into columns.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"parsing"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`field_prefix`	`str`	`""`	Optional prefix to add to extracted column names
`fix_trailing_commas`	`bool`	`True`	Whether to fix trailing commas in JSON (common LLM output issue)
`extract_embedded`	`bool`	`True`	Whether to extract JSON embedded in surrounding text
`drop_input`	`bool`	`False`	Whether to drop the input column after extraction

Methods

JSONParserBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with JSON fields expanded into columns.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

JSONParserBlock.validate_input_colsclassmethod

(cls, v: list[str]) -> list[str]

Validate that exactly one input column is specified.

Parameter	Type	Default
`v`	`list[str]`	required

Returns list[str]

RegexParserBlock

Bases: BaseTextParserBlock, BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import RegexParserBlock

Block for parsing text content using regex patterns.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"parser"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`parser_cleanup_tags`	`Optional[list[str]]`	`--`	Tags to remove from extracted content
`parsing_pattern`	`str`	required	--

Methods

RegexParserBlock.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

TagParserBlock

Bases: BaseTextParserBlock, BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import TagParserBlock

Block for parsing text content using start/end tags.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"parser"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`parser_cleanup_tags`	`Optional[list[str]]`	`--`	Tags to remove from extracted content
`start_tags`	`list[str]`	required	Start tags for extraction
`end_tags`	`list[str]`	required	End tags for extraction

Methods

TagParserBlock.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

TagParserBlock.normalize_tagsclassmethod

(cls, v)

Parameter	Type	Default
`v`		required

TagParserBlock.validate_tags

(self)

DuplicateColumnsBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import DuplicateColumnsBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict

Methods

DuplicateColumnsBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with duplicated columns.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

DuplicateColumnsBlock.model_post_init

(self, _DuplicateColumnsBlock__context: Any) -> None

Initialize derived attributes after Pydantic validation.

Parameter	Type	Default
`_DuplicateColumnsBlock__context`	`Any`	required

DuplicateColumnsBlock.validate_input_colsclassmethod

(cls, v)

Validate that input_cols is a non-empty dict.

Parameter	Type	Default
`v`		required

IndexBasedMapperBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import IndexBasedMapperBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`choice_map`	`dict[str, str]`	required	Dictionary mapping choice values to column names
`choice_cols`	`list[str]`	required	List of column names containing choice values

Methods

IndexBasedMapperBlock.generate

(self, samples: pandas.DataFrame, **kwargs) -> pandas.DataFrame

Generate a new dataset with selected values.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`		required

Returns pandas.DataFrame

IndexBasedMapperBlock.model_post_init

(self, _IndexBasedMapperBlock__context: Any) -> None

Initialize derived attributes after Pydantic validation.

Parameter	Type	Default
`_IndexBasedMapperBlock__context`	`Any`	required

IndexBasedMapperBlock.validate_choice_cols_not_emptyclassmethod

(cls, v)

Validate that choice_cols is not empty.

Parameter	Type	Default
`v`		required

IndexBasedMapperBlock.validate_choice_mapclassmethod

(cls, v)

Validate that choice_map is not empty.

Parameter	Type	Default
`v`		required

IndexBasedMapperBlock.validate_input_output_consistency

(self)

Validate that choice_cols and output_cols have same length and consistency.

JSONStructureBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import JSONStructureBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`ensure_json_serializable`	`bool`	`True`	Whether to ensure all values are JSON serializable
`pretty_print`	`bool`	`False`	Whether to format JSON with indentation

Methods

JSONStructureBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with JSON structured output.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

JSONStructureBlock.validate_output_colsclassmethod

(cls, v)

Validate that exactly one output column is specified.

Parameter	Type	Default
`v`		required

MeltColumnsBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import MeltColumnsBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict

Methods

MeltColumnsBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a flattened dataset in long format.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

MeltColumnsBlock.model_post_init

(self, _MeltColumnsBlock__context: Any) -> None

Initialize derived attributes after Pydantic validation.

Parameter	Type	Default
`_MeltColumnsBlock__context`	`Any`	required

MeltColumnsBlock.validate_input_colsclassmethod

(cls, v)

Validate that input_cols is not empty.

Parameter	Type	Default
`v`		required

MeltColumnsBlock.validate_output_colsclassmethod

(cls, v)

Validate that exactly two output columns are specified.

Parameter	Type	Default
`v`		required

RenameColumnsBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import RenameColumnsBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict

Methods

RenameColumnsBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with renamed columns.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

RenameColumnsBlock.model_post_init

(self, _RenameColumnsBlock__context: Any) -> None

Initialize derived attributes after Pydantic validation.

Parameter	Type	Default
`_RenameColumnsBlock__context`	`Any`	required

RenameColumnsBlock.validate_input_colsclassmethod

(cls, v)

Validate that input_cols is a non-empty dict.

Parameter	Type	Default
`v`		required

RowMultiplierBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import RowMultiplierBlock

Block for duplicating dataset rows.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`num_samples`	`int`	required	Number of times to duplicate each row
`shuffle`	`bool`	`False`	Shuffle output rows after duplication
`random_seed`	`Optional[int]`	`--`	Seed for reproducible shuffling

Methods

RowMultiplierBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with duplicated rows.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

SamplerBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import SamplerBlock

Block for randomly sampling values from list columns or across rows.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`num_samples`	`int`	`5`	Number of values to randomly sample
`random_seed`	`Optional[int]`	`--`	Random seed for reproducibility
`return_scalar`	`bool`	`False`	When num_samples=1, return scalar value instead of single-element list
`source`	`Literal[cell, column]`	`"cell"`	Sampling source: 'cell' samples from a list within each row; 'column' samples scalar values from the column across rows
`exclude_self`	`bool`	`True`	When source='column', exclude the current row's value from the pool
`exclude_by_value`	`bool`	`False`	When source='column' and exclude_self=True, exclude all pool entries matching the current row's value (not just the current index). Use after RowMultiplierBlock to avoid sampling duplicated copies of the same row.
`replace`	`bool`	`False`	Sample with replacement (True) or without (False)
`sample_range`	`Optional[list[int]]`	`--`	When source='column', restrict sampling pool to rows [start, end). Default None uses all rows.

Methods

SamplerBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with sampled values.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

SamplerBlock.validate_input_colsclassmethod

(cls, v: list[str]) -> list[str]

Validate that exactly one input column is specified.

Parameter	Type	Default
`v`	`list[str]`	required

Returns list[str]

SamplerBlock.validate_mode_specific_params

(self) -> 'SamplerBlock'

Reject mode-irrelevant parameters set to non-default values.

Returns SamplerBlock

SamplerBlock.validate_num_samplesclassmethod

(cls, v: int) -> int

Validate that num_samples is at least 1.

Parameter	Type	Default
`v`	`int`	required

Returns int

SamplerBlock.validate_output_cols_for_source

(self) -> 'SamplerBlock'

Validate output_cols length based on source mode.

Returns SamplerBlock

SamplerBlock.validate_sample_rangeclassmethod

(cls, v: Optional[list[int]]) -> Optional[list[int]]

Validate sample_range is a valid [start, end) pair.

Parameter	Type	Default
`v`	`Optional[list[int]]`	required

Returns Optional[list[int]]

TextConcatBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import TextConcatBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`separator`	`str`	`" "`	Separator to use between combined values

Methods

TextConcatBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate a dataset with combined columns.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

TextConcatBlock.validate_input_colsclassmethod

(cls, v)

Validate that input_cols is a non-empty list.

Parameter	Type	Default
`v`		required

TextConcatBlock.validate_output_colsclassmethod

(cls, v)

Validate that exactly one output column is specified.

Parameter	Type	Default
`v`		required

UniformColumnValueSetter

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import UniformColumnValueSetter

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"transform"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`reduction_strategy`	`Literal[mode, min, max, mean, median]`	`"mode"`	--

Methods

UniformColumnValueSetter.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Subclass method to implement data generation logic.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

UniformColumnValueSetter.model_post_init

(self, _UniformColumnValueSetter__context: Any) -> None

Override this method to perform additional initialization after `__init__` and `model_construct`. This is useful if you want to do some validation that requires the entire model to be initialized.

Parameter	Type	Default
`_UniformColumnValueSetter__context`	`Any`	required

UniformColumnValueSetter.validate_input_cols_singleclassmethod

(cls, v)

Parameter	Type	Default
`v`		required

ColumnValueFilterBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import ColumnValueFilterBlock

Base class for all blocks, with standardized patterns and full Pydantic compatibility.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"filtering"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`filter_value`	`Union[Any, list[Any]]`	required	The value(s) to filter by
`operation`	`str`	required	String name of binary operator for comparison (e.g., 'eq', 'contains')
`convert_dtype`	`Optional[str]`	`--`	String name of type to convert filter column to ('float' or 'int')

Methods

ColumnValueFilterBlock.generate

(self, samples: pandas.DataFrame, **_kwargs: Any) -> pandas.DataFrame

Generate filtered dataset based on specified conditions.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`_kwargs`	`Any`	required

Returns pandas.DataFrame

ColumnValueFilterBlock.model_post_init

(self, _ColumnValueFilterBlock__context: Any) -> None

Initialize derived attributes after Pydantic validation.

Parameter	Type	Default
`_ColumnValueFilterBlock__context`	`Any`	required

ColumnValueFilterBlock.validate_convert_dtypeclassmethod

(cls, v)

Validate that convert_dtype is a supported type string.

Parameter	Type	Default
`v`		required

ColumnValueFilterBlock.validate_input_cols_not_emptyclassmethod

(cls, v)

Validate that we have at least one input column.

Parameter	Type	Default
`v`		required

ColumnValueFilterBlock.validate_operationclassmethod

(cls, v)

Validate that operation is a supported operation string.

Parameter	Type	Default
`v`		required

AgentBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import AgentBlock

Block for executing external agent frameworks on DataFrame rows.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"agent"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`agent_framework`	`str`	required	Connector name (e.g., 'langflow')
`agent_url`	`str`	required	Agent API endpoint URL
`agent_api_key`	`Optional[str]`	`--`	API key for authentication
`timeout`	`float`	`120`	Request timeout in seconds
`max_retries`	`int`	`3`	Maximum retry attempts
`session_id_col`	`Optional[str]`	`--`	Column containing session IDs
`async_mode`	`bool`	`False`	Use async execution for better throughput
`max_concurrency`	`int`	`10`	Maximum concurrent requests in async mode
`connector_kwargs`	`dict[str, Any]`	`[object Object]`	Extra keyword arguments passed to the connector constructor. Use for framework-specific settings like assistant_id for LangGraph.

Methods

AgentBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Process DataFrame rows through the agent.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

AgentBlock.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

AgentResponseExtractorBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import AgentResponseExtractorBlock

Block for extracting fields from standardized agent response objects.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"agent_util"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`extract_text`	`bool`	`True`	Whether to extract text content from responses.
`extract_session_id`	`bool`	`False`	Whether to extract session_id from responses.
`extract_tool_trace`	`bool`	`False`	Whether to extract the full tool call trace from agent responses. Collects all assistant messages with tool_calls and tool result messages into a structured trace list.
`expand_lists`	`bool`	`True`	Whether to expand list inputs into individual rows (True) or preserve lists (False).
`field_prefix`	`str`	`""`	Prefix to add to output field names (e.g., 'agent_' results in 'agent_text').

Methods

AgentResponseExtractorBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Subclass method to implement data generation logic.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

AgentResponseExtractorBlock.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

AgentResponseExtractorBlock.validate_extraction_configuration

(self)

Validate that at least one extraction field is enabled and pre-compute field names.

MCPAgentBlock

Bases: BaseBlock, BaseModel, ABC

from sdg_hub.core.blocks import MCPAgentBlock

LLM agent block that connects to remote MCP servers for tool use.

Fields

Name	Type	Default	Description
`block_name`	`str`	required	Unique identifier for this block instance
`block_type`	`str`	`"mcp"`	--
`input_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Input columns: str, list, or dict
`output_cols`	`Union[str, list[str], dict[str, Any], NoneType]`	`--`	Output columns: str, list, or dict
`mcp_server_url`	`str`	required	URL of the remote MCP server
`mcp_headers`	`Union[dict[str, str], NoneType]`	`--`	HTTP headers for MCP server authentication
`model`	`str`	required	Model identifier in LiteLLM format
`api_key`	`Optional[pydantic.types.SecretStr]`	`--`	API key for the LLM provider
`api_base`	`Optional[str]`	`--`	Base URL for the LLM API
`max_iterations`	`int`	`10`	Maximum number of agentic loop iterations
`system_prompt`	`Optional[str]`	`--`	System prompt to prepend to conversations

Methods

MCPAgentBlock.generate

(self, samples: pandas.DataFrame, **kwargs: Any) -> pandas.DataFrame

Generate responses using LLM with MCP tools.

Parameter	Type	Default
`samples`	`pandas.DataFrame`	required
`kwargs`	`Any`	required

Returns pandas.DataFrame

MCPAgentBlock.model_post_init

(self, _MCPAgentBlock__context) -> None

Initialize after Pydantic validation.

Parameter	Type	Default
`_MCPAgentBlock__context`		required

MCPAgentBlock.validate_max_iterationsclassmethod

(cls, v)

Ensure max_iterations is positive.

Parameter	Type	Default
`v`		required

MCPAgentBlock.validate_single_input_colclassmethod

(cls, v)

Ensure exactly one input column.

Parameter	Type	Default
`v`		required

MCPAgentBlock.validate_single_output_colclassmethod

(cls, v)

Ensure exactly one output column.

Parameter	Type	Default
`v`		required

Flow

Bases: BaseModel

from sdg_hub import Flow

Pydantic-based flow for chaining data generation blocks.

Fields

Name	Type	Default	Description
`blocks`	`list[BaseBlock]`		Ordered list of blocks to execute in the flow
`metadata`	`FlowMetadata`	required	Flow metadata including name, version, author, etc.

Methods

Flow.add_block

(self, block: BaseBlock) -> 'Flow'

Add a block to the flow, returning a new Flow instance.

Parameter	Type	Default
`block`	`BaseBlock`	required

Returns Flow

Flow.dry_run

(self, dataset: Union[pandas.DataFrame, Dataset], sample_size: int = 2, runtime_params: Optional[dict[str, dict[str, Any]]] = None, max_concurrency: Optional[int] = None, enable_time_estimation: bool = False) -> dict[str, Any]

Perform a dry run of the flow with a subset of data.

Parameter	Type	Default
`dataset`	`Union[pandas.DataFrame, Dataset]`	required
`sample_size`	`int`	`2`
`runtime_params`	`Union[dict[str, dict[str, Any]], NoneType]`	`--`
`max_concurrency`	`Optional[int]`	`--`
`enable_time_estimation`	`bool`	`False`

Returns dict[str, Any]

Flow.from_yamlclassmethod

(cls, yaml_path: str) -> 'Flow'

Load flow from YAML configuration file.

Parameter	Type	Default
`yaml_path`	`str`	required

Returns Flow

Flow.generate

(self, dataset: Union[pandas.DataFrame, Dataset], runtime_params: Optional[dict[str, dict[str, Any]]] = None, checkpoint_dir: Optional[str] = None, save_freq: Optional[int] = None, log_dir: Optional[str] = None, max_concurrency: Optional[int] = None) -> Union[pandas.DataFrame, Dataset]

Execute the flow blocks in sequence to generate data.

Parameter	Type	Default
`dataset`	`Union[pandas.DataFrame, Dataset]`	required
`runtime_params`	`Union[dict[str, dict[str, Any]], NoneType]`	`--`
`checkpoint_dir`	`Optional[str]`	`--`
`save_freq`	`Optional[int]`	`--`
`log_dir`	`Optional[str]`	`--`
`max_concurrency`	`Optional[int]`	`--`

Returns Union[pandas.DataFrame, Dataset]

Flow.get_dataset_requirements

(self) -> Optional[DatasetRequirements]

Get the dataset requirements for this flow, or None if not defined.

Returns Optional[DatasetRequirements]

Flow.get_dataset_schema

(self) -> pandas.DataFrame

Get an empty DataFrame with the correct schema for this flow.

Returns pandas.DataFrame

Flow.get_default_model

(self) -> Optional[str]

Get the default recommended model for this flow, or None if unspecified.

Returns Optional[str]

Flow.get_info

(self) -> dict[str, Any]

Get information about the flow.

Returns dict[str, Any]

Flow.get_model_recommendations

(self) -> dict[str, Any]

Get model recommendations dict with 'default', 'compatible', 'experimental' keys.

Returns dict[str, Any]

Flow.is_agent_config_required

(self) -> bool

Check if agent configuration is required (True if flow has agent blocks).

Returns bool

Flow.is_agent_config_set

(self) -> bool

Check if agent configuration has been set or is not required.

Returns bool

Flow.is_model_config_required

(self) -> bool

Check if model configuration is required (True if flow has LLM blocks).

Returns bool

Flow.is_model_config_set

(self) -> bool

Check if model configuration has been set or is not required.

Returns bool

Flow.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

Flow.print_info

(self) -> None

Print an interactive summary of the Flow in the console using rich.

Flow.reset_agent_config

(self) -> None

Reset agent configuration flag (useful for testing or reconfiguration).

Flow.reset_model_config

(self) -> None

Reset model configuration flag (useful for testing or reconfiguration).

Flow.set_agent_config

(self, agent_framework: Optional[str] = None, agent_url: Optional[str] = None, agent_api_key: Optional[str] = None, blocks: Optional[list[str]] = None, **kwargs: Any) -> None

Configure agent settings for agent blocks in this flow (in-place).

Parameter	Type	Default
`agent_framework`	`Optional[str]`	`--`
`agent_url`	`Optional[str]`	`--`
`agent_api_key`	`Optional[str]`	`--`
`blocks`	`Optional[list[str]]`	`--`
`kwargs`	`Any`	required

Flow.set_model_config

(self, model: Optional[str] = None, api_base: Optional[str] = None, api_key: Optional[str] = None, blocks: Optional[list[str]] = None, **kwargs: Any) -> None

Configure model settings for LLM blocks in this flow (in-place).

Parameter	Type	Default
`model`	`Optional[str]`	`--`
`api_base`	`Optional[str]`	`--`
`api_key`	`Optional[str]`	`--`
`blocks`	`Optional[list[str]]`	`--`
`kwargs`	`Any`	required

Flow.to_yaml

(self, output_path: str) -> None

Save flow configuration to YAML file.

Parameter	Type	Default
`output_path`	`str`	required

Flow.validate_block_names_unique

(self) -> 'Flow'

Ensure all block names are unique within the flow.

Returns Flow

Flow.validate_blocksclassmethod

(cls, v: list[BaseBlock]) -> list[BaseBlock]

Validate that all blocks are BaseBlock instances.

Parameter	Type	Default
`v`	`list[BaseBlock]`	required

Returns list[BaseBlock]

Flow.validate_dataset

(self, dataset: Union[pandas.DataFrame, Dataset]) -> list[str]

Validate dataset against flow requirements. Returns list of error messages.

Parameter	Type	Default
`dataset`	`Union[pandas.DataFrame, Dataset]`	required

Returns list[str]

FlowRegistry

from sdg_hub import FlowRegistry

Registry for managing contributed flows.

Methods

FlowRegistry.discover_flowsclassmethod

(cls) -> None

Discover and display all flows in a formatted table.

FlowRegistry.get_flow_metadataclassmethod

(cls, flow_name: str) -> Optional[FlowMetadata]

Get metadata for a registered flow.

Parameter	Type	Default
`flow_name`	`str`	required

Returns Optional[FlowMetadata]

FlowRegistry.get_flow_pathclassmethod

(cls, flow_name_or_id: str) -> Optional[str]

Get the path to a registered flow.

Parameter	Type	Default
`flow_name_or_id`	`str`	required

Returns Optional[str]

FlowRegistry.get_flow_path_safeclassmethod

(cls, flow_name_or_id: str) -> str

Get the path to a registered flow with better error handling.

Parameter	Type	Default
`flow_name_or_id`	`str`	required

Returns str

FlowRegistry.get_flows_by_categoryclassmethod

(cls) -> Dict[str, List[Dict[str, str]]]

Get flows organized by their primary tag.

Returns dict[str, list[dict[str, str]]]

FlowRegistry.list_flowsclassmethod

(cls) -> List[Dict[str, str]]

List all registered flows with their IDs.

Returns list[dict[str, str]]

FlowRegistry.register_search_pathclassmethod

(cls, path: str) -> None

Add a directory to search for flows.

Parameter	Type	Default
`path`	`str`	required

FlowRegistry.search_flowsclassmethod

(cls, tag: Optional[str] = None, author: Optional[str] = None) -> List[Dict[str, str]]

Search flows by criteria.

Parameter	Type	Default
`tag`	`Optional[str]`	`--`
`author`	`Optional[str]`	`--`

Returns list[dict[str, str]]

FlowRegistryEntry

from sdg_hub.core.flow.registry import FlowRegistryEntry

Entry in the flow registry.

Fields

Name	Type	Default	Description
`path`	`str`	required	--
`metadata`	`FlowMetadata`	required	--

Methods

FlowRegistryEntry.__init__

(self, path: str, metadata: FlowMetadata) -> None

Initialize self. See help(type(self)) for accurate signature.

Parameter	Type	Default
`path`	`str`	required
`metadata`	`FlowMetadata`	required

FlowMetadata

Bases: BaseModel

from sdg_hub import FlowMetadata

Metadata for flow configuration and open source contributions.

Fields

Name	Type	Default	Description
`name`	`str`	required	Human-readable name
`id`	`str`	`""`	Unique identifier for the flow, generated from name
`description`	`str`	`""`	Detailed description
`version`	`str`	`"1.0.0"`	Semantic version
`author`	`str`	`""`	Author or contributor name
`recommended_models`	`Optional[RecommendedModels]`	`--`	Simplified recommended models structure
`tags`	`list[str]`		Tags for categorization and search
`license`	`str`	`"Apache-2.0"`	License identifier
`dataset_requirements`	`Optional[DatasetRequirements]`	`--`	Requirements for input datasets
`output_columns`	`Optional[list[str]]`	`--`	Columns to keep in the final output. Original input columns are always preserved.

Methods

FlowMetadata.ensure_id

(self) -> 'FlowMetadata'

Ensure id is set.

Returns FlowMetadata

FlowMetadata.get_best_model

(self, available_models: Optional[list[str]] = None) -> Optional[str]

Get the best recommended model based on availability.

Parameter	Type	Default
`available_models`	`Optional[list[str]]`	`--`

Returns Optional[str]

FlowMetadata.validate_idclassmethod

(cls, v: str) -> str

Validate flow id.

Parameter	Type	Default
`v`	`str`	required

Returns str

FlowMetadata.validate_output_columnsclassmethod

(cls, v: Optional[list[str]]) -> Optional[list[str]]

Validate and clean output columns.

Parameter	Type	Default
`v`	`Optional[list[str]]`	required

Returns Optional[list[str]]

FlowMetadata.validate_recommended_modelsclassmethod

(cls, v: Optional[RecommendedModels]) -> Optional[RecommendedModels]

Validate recommended models structure.

Parameter	Type	Default
`v`	`Optional[RecommendedModels]`	required

Returns Optional[RecommendedModels]

FlowMetadata.validate_tagsclassmethod

(cls, v: list[str]) -> list[str]

Validate and clean tags.

Parameter	Type	Default
`v`	`list[str]`	required

Returns list[str]

RecommendedModels

Bases: BaseModel

from sdg_hub.core.flow.metadata import RecommendedModels

Simplified recommended models structure.

Fields

Name	Type	Default	Description
`default`	`str`	required	Default model to use
`compatible`	`list[str]`		Compatible models
`experimental`	`list[str]`		Experimental models

Methods

RecommendedModels.get_all_models

(self) -> list[str]

Get all models (default + compatible + experimental).

Returns list[str]

RecommendedModels.get_best_model

(self, available_models: Optional[list[str]] = None) -> Optional[str]

Get the best model based on availability.

Parameter	Type	Default
`available_models`	`Optional[list[str]]`	`--`

Returns Optional[str]

RecommendedModels.validate_defaultclassmethod

(cls, v: str) -> str

Validate default model name is not empty.

Parameter	Type	Default
`v`	`str`	required

Returns str

RecommendedModels.validate_model_listsclassmethod

(cls, v: list[str]) -> list[str]

Validate model lists contain non-empty names.

Parameter	Type	Default
`v`	`list[str]`	required

Returns list[str]

DatasetRequirements

Bases: BaseModel

from sdg_hub.core.flow.metadata import DatasetRequirements

Dataset requirements for flow execution.

Fields

Name	Type	Default	Description
`required_columns`	`list[str]`		Column names that must be present
`optional_columns`	`list[str]`		Optional columns that can enhance performance
`min_samples`	`int`	`1`	Minimum number of samples required
`max_samples`	`Optional[int]`	`--`	Maximum number of samples to process
`column_types`	`dict[str, str]`	`[object Object]`	Expected types for specific columns
`description`	`str`	`""`	Human-readable description

Methods

DatasetRequirements.validate_column_namesclassmethod

(cls, v: list[str]) -> list[str]

Validate column names are not empty.

Parameter	Type	Default
`v`	`list[str]`	required

Returns list[str]

DatasetRequirements.validate_dataset

(self, dataset_columns: list[str], dataset_size: int) -> list[str]

Validate a dataset against these requirements.

Parameter	Type	Default
`dataset_columns`	`list[str]`	required
`dataset_size`	`int`	required

Returns list[str]

DatasetRequirements.validate_sample_limits

(self) -> 'DatasetRequirements'

Validate sample limits are consistent.

Returns DatasetRequirements

ModelOption

Bases: BaseModel

from sdg_hub.core.flow.metadata import ModelOption

Represents a model option with compatibility level.

Fields

Name	Type	Default	Description
`name`	`str`	required	Model identifier
`compatibility`	`ModelCompatibility`	`"compatible"`	Compatibility level with the flow

Methods

ModelOption.validate_nameclassmethod

(cls, v: str) -> str

Validate model name is not empty.

Parameter	Type	Default
`v`	`str`	required

Returns str

ModelCompatibility

Bases: Enum

from sdg_hub.core.flow.metadata import ModelCompatibility

Model compatibility levels.

BaseConnector

Bases: BaseModel, ABC

from sdg_hub.core.connectors.base import BaseConnector

Abstract base class for all connectors.

Fields

Name	Type	Default	Description
`config`	`ConnectorConfig`	required	Connector configuration

Methods

BaseConnector.aexecute

(self, request: Any) -> Any

Execute an asynchronous request.

Parameter	Type	Default
`request`	`Any`	required

Returns Any

BaseConnector.executeabstract

(self, request: Any) -> Any

Execute a synchronous request.

Parameter	Type	Default
`request`	`Any`	required

Returns Any

ConnectorConfig

Bases: BaseModel

from sdg_hub.core.connectors.base import ConnectorConfig

Base configuration for all connectors.

Fields

Name	Type	Default	Description
`url`	`Optional[str]`	`--`	Base URL for the service
`api_key`	`Optional[str]`	`--`	API key for authentication
`timeout`	`float`	`120`	Request timeout in seconds
`max_retries`	`int`	`3`	Maximum retry attempts

ConnectorRegistry

from sdg_hub.core.connectors.registry import ConnectorRegistry

Global registry for connector classes.

Methods

ConnectorRegistry.clearclassmethod

(cls) -> 'None'

Clear all registered connectors. Primarily for testing.

Returns None

ConnectorRegistry.getclassmethod

(cls, name: 'str') -> 'type[BaseConnector]'

Get a connector class by name.

Parameter	Type	Default
`name`	`str`	required

Returns type[BaseConnector]

ConnectorRegistry.list_allclassmethod

(cls) -> 'list[str]'

Get all registered connector names.

Returns list[str]

ConnectorRegistry.registerclassmethod

(cls, name: 'str')

Parameter	Type	Default
`name`	`str`	required

BaseAgentConnector

Bases: BaseConnector, BaseModel, ABC

from sdg_hub.core.connectors.agent.base import BaseAgentConnector

Base class for agent framework connectors.

Fields

Name	Type	Default	Description
`config`	`ConnectorConfig`	required	Connector configuration

Methods

BaseAgentConnector.asend

(self, request: mlflow.types.agent.ChatAgentRequest) -> mlflow.types.agent.ChatAgentResponse

Async send - convenience wrapper.

Parameter	Type	Default
`request`	`mlflow.types.agent.ChatAgentRequest`	required

Returns mlflow.types.agent.ChatAgentResponse

BaseAgentConnector.build_requestabstract

(self, request: mlflow.types.agent.ChatAgentRequest) -> dict[str, Any]

Build framework-specific request payload.

Parameter	Type	Default
`request`	`mlflow.types.agent.ChatAgentRequest`	required

Returns dict[str, Any]

BaseAgentConnector.execute

(self, request: dict[str, Any]) -> dict[str, Any]

Execute a request (BaseConnector interface).

Parameter	Type	Default
`request`	`dict[str, Any]`	required

Returns dict[str, Any]

BaseAgentConnector.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

BaseAgentConnector.parse_responseabstract

(self, response: dict[str, Any]) -> mlflow.types.agent.ChatAgentResponse

Parse and validate framework response.

Parameter	Type	Default
`response`	`dict[str, Any]`	required

Returns mlflow.types.agent.ChatAgentResponse

BaseAgentConnector.send

(self, request: mlflow.types.agent.ChatAgentRequest, async_mode: bool = False)

Send a request to the agent.

Parameter	Type	Default
`request`	`mlflow.types.agent.ChatAgentRequest`	required
`async_mode`	`bool`	`False`

LangflowConnector

Bases: BaseAgentConnector, BaseConnector, BaseModel, ABC

from sdg_hub.core.connectors.agent.langflow import LangflowConnector

Connector for Langflow agent framework.

Fields

Name	Type	Default	Description
`config`	`ConnectorConfig`	required	Connector configuration

Methods

LangflowConnector.build_request

(self, request: mlflow.types.agent.ChatAgentRequest) -> dict[str, Any]

Build Langflow-specific request payload.

Parameter	Type	Default
`request`	`mlflow.types.agent.ChatAgentRequest`	required

Returns dict[str, Any]

LangflowConnector.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

LangflowConnector.parse_response

(self, response: dict[str, Any]) -> mlflow.types.agent.ChatAgentResponse

Parse Langflow response into ChatAgentResponse.

Parameter	Type	Default
`response`	`dict[str, Any]`	required

Returns mlflow.types.agent.ChatAgentResponse

LangGraphConnector

Bases: BaseAgentConnector, BaseConnector, BaseModel, ABC

from sdg_hub.core.connectors.agent.langgraph import LangGraphConnector

Connector for LangGraph agent framework.

Fields

Name	Type	Default	Description
`config`	`ConnectorConfig`	required	Connector configuration
`assistant_id`	`str`	`"agent"`	The assistant ID or graph name to run.
`run_config`	`dict[str, Any]`	`[object Object]`	Optional configuration dict passed in the run payload. Merged as the 'config' key in the LangGraph /runs/wait request. Use this to pass runtime parameters to the graph via 'configurable', e.g. ``{'configurable': {'model': 'gpt-4o'}}``.

Methods

LangGraphConnector.build_request

(self, request: mlflow.types.agent.ChatAgentRequest) -> dict[str, Any]

Build LangGraph run request payload.

Parameter	Type	Default
`request`	`mlflow.types.agent.ChatAgentRequest`	required

Returns dict[str, Any]

LangGraphConnector.model_post_init

(self: 'BaseModel', context: 'Any', /) -> 'None'

This function is meant to behave like a BaseModel method to initialize private attributes.

Parameter	Type	Default
`context`	`Any`	required

Returns None

LangGraphConnector.parse_response

(self, response: dict[str, Any]) -> mlflow.types.agent.ChatAgentResponse

Parse LangGraph response into ChatAgentResponse.

Parameter	Type	Default
`response`	`dict[str, Any]`	required

Returns mlflow.types.agent.ChatAgentResponse