Annotations Config¶

classmethod AnnotationsConfig.builder()¶

Returns a builder for creating AnnotationsConfig instances.

This method provides a convenient way to construct an AnnotationsConfig object by incrementally setting its attributes.

Returns:

AnnotationsConfig.Builder: An instance of the builder that can be used to set the attributes of an AnnotationsConfig.

Example:

from berrydb.model_garden.annotations_config import AnnotationsConfig

# Get a builder instance
config_builder = AnnotationsConfig.builder()

# Configure the annotations settings
my_config = (
    config_builder
    .name("my-custom-annotation-config")
    .input_transform_expression("document.full_text")
    .output_transform_expression("annotations.entities")
    .llm_provider("openai")
    .llm_model("gpt-4o-mini")
    .prompt("Extract all named entities from the text: {{input}}")
    .additional_transform_expressions([
        {"jsonPath": "annotations.sentiment", "expression": "document.sentiment_score"}
    ])
    .build()
)

# The my_config object is now an instance of AnnotationsConfig
# and can be used, for example, with its save() method.

AnnotationsConfig.save(berrydb_api_key)¶

Saves or updates the annotations configuration in BerryDB.

Parameters:

berrydb_api_key (str): The API key for authenticating with BerryDB.

Returns:

AnnotationsConfig: The AnnotationsConfig instance.

Example:

from berrydb.model_garden.annotations_config import AnnotationsConfig

# API key for BerryDB
my_api_key = "YOUR_BERRYDB_API_KEY"

# Example 1: Create and save a new annotations configuration
new_config_builder = AnnotationsConfig.builder()
new_config = (
    new_config_builder
    .name("my-ner-config")
    .input_transform_expression("data.text")
    .output_transform_expression("annotations.ner")
    .llm_provider("openai")
    .llm_model("gpt-4o-mini")
    .prompt("Extract named entities from the following text: {{input}}")
    .build()
)
# Save the new configuration to BerryDB
saved_new_config = new_config.save(my_api_key)
print(f"Saved new config with ID: {saved_new_config.id}")

# Example 2: Get an existing configuration, modify, and save (update)
existing_config_builder = AnnotationsConfig.get_config_as_builder(
    api_key=my_api_key,
    config_name="my-ner-config" # Assuming this config exists
)
existing_config = existing_config_builder.prompt(
    "Identify and label all persons, organizations, and locations in this text: {{input}}"
).build()

# Update the configuration in BerryDB
updated_config = existing_config.save(my_api_key)

static AnnotationsConfig.get_config_as_builder(api_key, config_name)¶

Retrieves an existing annotations configuration from BerryDB as a builder instance.

This allows for easy modification of a previously saved configuration.

Parameters:

api_key (str): The API key for authenticating with BerryDB.
config_name (str): The name of the annotations configuration to retrieve.

Returns:

AnnotationsConfig.Builder: A builder instance pre-populated with the settings of the fetched configuration.

Example:

from berrydb import AnnotationsConfig

my_api_key = "YOUR_BERRYDB_API_KEY"
existing_config_name = "my-ner-config" # Name of a config already in BerryDB

# Get the existing config as a builder
builder = AnnotationsConfig.get_config_as_builder(
    api_key=my_api_key,
    config_name=existing_config_name
)

# Now you can modify it, e.g., change the prompt
modified_config = (
    builder
    .prompt("A new prompt for entity extraction: {{input}}")
    .build()
)
# modified_config.save(my_api_key) # And save the changes

`Annotations Config Builder Methods`¶

The Builder class provides a fluent interface to construct an AnnotationsConfig object.

Builder.name(name)¶

Sets the name for the annotations configuration.

Parameters:

name (str): The name of the configuration.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-new-config")
    # ... other configurations ...
    .build()
)

Builder.input_transform_expression(jsonata_expr)¶

Sets the input transformation expression for the annotations configuration.

This expression, written in JSONata, defines how to extract the input data (e.g., text to be annotated) from a source document or data structure. JSONata provides a powerful way to query and transform JSON-like data.

Parameters:

jsonata_expr (str): The JSONata expression string. For example, source_data.article_body to extract the value of article_body from a source_data object, or more complex JSONata expressions for intricate transformations.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb.model_garden.annotations_config import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-config")
    .input_transform_expression("payload.document.text_content") # JSONata expression
    # ... other configurations ...
    .build()
)

Builder.output_transform_expression(jsonata_expr)¶

Sets the output transformation expression for the annotations configuration.

This expression, written in JSONata, defines how to process and structure the output received from the model or LLM or other annotation source before it is stored. It maps the raw output into the desired annotation format.

Parameters:

jsonata_expr (str): The JSONata expression string. For example, llm_output.entities to extract an ‘entities’ field from the LLM’s output, or a more complex expression to reshape the data.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-annotation-processor")
    .output_transform_expression("llm_response.parsed_output.labels") # JSONata expression
    .input_transform_expression("document.text")
    # ... other configurations ...
    .build()
)

Builder.additional_transform_expressions(jsonata_expressions)¶

Sets a list of additional transformation expressions for the annotations configuration.

These expressions, written in JSONata, allow you to apply further transformations to the data. This is applied when the annotation is being submitted to BerryDB. Each expression in the list should specify a jsonPath where the result of the expression should be placed in the final output structure.

Parameters:

jsonata_expressions (List[Dict[str, str]]): A list of dictionaries. Each dictionary must contain two keys:
- ‘jsonPath’ (str): The JSONPath where the result of the expression should be placed.
- ‘expression’ (str): The JSONata expression to evaluate.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-complex-annotation-config")
    .input_transform_expression("document.text")
    .output_transform_expression("llm_response.main_result")
    .additional_transform_expressions([
        {"jsonPath": "metadata.sentiment_score", "expression": "document.sentiment"},
        {"jsonPath": "summary.length", "expression": "llm_response.summary ~> $length()"}
    ])
    .build()
)

Builder.llm_provider(provider='openai')¶

Sets the Large Language Model (LLM) provider for the annotations configuration.

This specifies which LLM provider (e.g., “openai”, “anthropic”) will be used if this configuration is leveraged.

Parameters:

provider (str, optional): The name of the LLM provider.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config_with_openai = (
    builder
    .name("openai-config")
    .llm_provider("openai") # Explicitly set to OpenAI
    .llm_model("gpt-4o-mini") # A model from OpenAI
    .prompt("Extract entities: {{input}}")
    .input_transform_expression("data.text")
    .output_transform_expression("annotations.ner")
    .build()
)

builder_default_provider = AnnotationsConfig.builder()
config_with_default = (
    builder_default_provider
    .name("default-provider-config")
    # .llm_provider() is not called, so it uses the default
    # .llm_model("some-default-model") # Uses default LLM model
    .prompt("Summarize: {{input}}")
    .input_transform_expression("data.text")
    .output_transform_expression("annotations.summary")
    .build()
)

Builder.llm_model(model)¶

Sets the specific Large Language Model (LLM) to be used for the annotations configuration.

This specifies the particular model (e.g., “gpt-4o-mini”, “claude-3-opus-20240229”) from the selected llm_provider that will be used if this configuration is leveraged for automated annotation generation.

Ensure the chosen model is compatible with and available from the previously set llm_provider.

Parameters:

model (str): The name or identifier of the LLM model.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-ner-config-with-specific-model")
    .llm_provider("openai")  # First, set the provider
    .llm_model("gpt-4o-mini") # Then, set the specific model from that provider
    .prompt("Extract all named entities from the following text: {{input}}")
    .input_transform_expression("data.text_content")
    .output_transform_expression("annotations.entities")
    .build()
)

Builder.prompt(prompt)¶

Sets the prompt template for the LLM in the annotations configuration.

This template defines the instructions and context provided to the Large Language Model (LLM).

Parameters:

prompt (str): The prompt template string.

Returns:

AnnotationsConfig.Builder: The builder instance, allowing for method chaining.

Example:

from berrydb import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-entity-extraction-config")
    .prompt("Extract all person names and locations from the following text: {{input}}")
    .llm_provider("openai")
    .llm_model("gpt-4o-mini")
    .input_transform_expression("document.text_content")
    .output_transform_expression("annotations.extracted_entities")
    .build()
)

Builder.build(config_name=None)¶

Constructs an AnnotationsConfig instance using the configured settings.

Parameters:

config_name (str, optional): An optional name for the configuration. If not provided, the name set using the name() builder method will be used.

Returns:

AnnotationsConfig: A new AnnotationsConfig instance.

Example:

from berrydb.model_garden.annotations_config import AnnotationsConfig

builder = AnnotationsConfig.builder()
config = (
    builder
    .name("my-final-config")
    # ... other configurations ...
    .build()
)

Annotations Config¶

Annotations Config Builder Methods¶

`Annotations Config Builder Methods`¶