Skip to content

LLM module

Source code in kili/llm/presentation/client/llm.py
class LlmClientMethods:
    def __init__(self, kili_api_gateway: KiliAPIGateway):
        self.kili_api_gateway = kili_api_gateway

    def export(
        self,
        project_id: str,
        disable_tqdm: Optional[bool] = False,
        asset_ids: Optional[List[str]] = None,
        external_ids: Optional[List[str]] = None,
    ) -> Optional[List[Dict[str, Union[List[str], str]]]]:
        # pylint: disable=line-too-long
        """Returns an export of llm assets with valid labels.

        Args:
            project_id: Identifier of the project.
            asset_ids: Optional list of the assets internal IDs from which to export the labels.
            disable_tqdm: Disable the progress bar if True.
            external_ids: Optional list of the assets external IDs from which to export the labels.

        !!! Example
            ```python
            kili.llm.export("your_project_id")
            ```
        """
        if asset_ids and external_ids:
            raise ValueError("You cannot provide both asset_ids and external_ids")

        if external_ids is not None and asset_ids is None:
            id_map = AssetUseCasesUtils(self.kili_api_gateway).infer_ids_from_external_ids(
                asset_external_ids=cast(List[AssetExternalId], external_ids),
                project_id=ProjectId(project_id),
            )
            resolved_asset_ids = [id_map[AssetExternalId(i)] for i in external_ids]
        else:
            resolved_asset_ids = (
                list(AssetId(asset_id) for asset_id in asset_ids) if asset_ids else None
            )

        asset_filter = AssetFilters(
            project_id=ProjectId(project_id), asset_id_in=resolved_asset_ids
        )

        try:
            return export(
                self.kili_api_gateway,
                project_id=ProjectId(project_id),
                asset_filter=asset_filter,
                disable_tqdm=disable_tqdm,
            )
        except NoCompatibleJobError as excp:
            warnings.warn(str(excp), stacklevel=2)
            return None

    def create_model(self, organization_id: str, model: dict) -> ModelDict:
        # pylint: disable=line-too-long
        """Create a new model in an organization.

        Args:
            organization_id: Identifier of the organization.
            model: A dictionary representing the model to create, containing:
                - `name`: Name of the model.
                - `type`: Type of the model, one of:
                    - `"AZURE_OPEN_AI"`
                    - `"OPEN_AI_SDK"`
                - `credentials`: Credentials required for the model, depending on the type:
                    - For `"AZURE_OPEN_AI"` type:
                        - `api_key`: The API key for Azure OpenAI.
                        - `deployment_id`: The deployment ID within Azure.
                        - `endpoint`: The endpoint URL of the Azure OpenAI service.
                    - For `"OPEN_AI_SDK"` type:
                        - `api_key`: The API key for OpenAI SDK.
                        - `endpoint`: The endpoint URL of the OpenAI SDK service.

        Returns:
            A dictionary containing the created model's details.

        Examples:
            >>> # Example of creating an OpenAI SDK model
            >>> model_data = {
            ...     "name": "My OpenAI SDK Model",
            ...     "type": "OPEN_AI_SDK",
            ...     "credentials": {
            ...         "api_key": "your_open_ai_api_key",
            ...         "endpoint": "your_open_ai_endpoint"
            ...     }
            ... }
            >>> kili.llm.create_model(organization_id="your_organization_id", model=model_data)
        """
        credentials_data = model["credentials"]
        model_type = ModelType(model["type"])

        if model_type == ModelType.AZURE_OPEN_AI:
            credentials = AzureOpenAICredentials(**credentials_data)
        elif model_type == ModelType.OPEN_AI_SDK:
            credentials = OpenAISDKCredentials(**credentials_data)
        else:
            raise ValueError(f"Unsupported model type: {model['type']}")

        model_input = ModelToCreateInput(
            credentials=credentials,
            name=model["name"],
            type=model_type,
            organization_id=organization_id,
        )
        return self.kili_api_gateway.create_model(model=model_input)

    def models(self, organization_id: str, fields: Optional[List[str]] = None) -> List[ModelDict]:
        # pylint: disable=line-too-long
        """List models in an organization.

        Args:
            organization_id: Identifier of the organization.
            fields: All the fields to request among the possible fields for the models.
                Defaults to ["id", "credentials", "name", "type"].

        Returns:
            A list of models.

        Examples:
            >>> kili.llm.models(organization_id="your_organization_id")
        """
        converted_filters = OrganizationModelFilters(
            organization_id=organization_id,
        )

        return list(self.kili_api_gateway.list_models(filters=converted_filters, fields=fields))

    def model(self, model_id: str, fields: Optional[List[str]] = None) -> ModelDict:
        # pylint: disable=line-too-long
        """Retrieve a specific model.

        Args:
            model_id: Identifier of the model.
            fields: All the fields to request among the possible fields for the models.
                Defaults to ["id", "credentials", "name", "type"].

        Returns:
            A dictionary representing the model.

        Examples:
            >>> kili.llm.model(model_id="your_model_id")
        """
        return self.kili_api_gateway.get_model(
            model_id=model_id,
            fields=fields,
        )

    def update_properties_in_model(self, model_id: str, model: dict) -> ModelDict:
        # pylint: disable=line-too-long
        """Update properties of an existing model.

        Args:
            model_id: Identifier of the model to update.
            model: A dictionary containing the properties to update, which may include:
                - `name`: New name of the model.
                - `credentials`: Updated credentials for the model, depending on the type:
                    - For `"AZURE_OPEN_AI"` type:
                        - `api_key`: The API key for Azure OpenAI.
                        - `deployment_id`: The deployment ID within Azure.
                        - `endpoint`: The endpoint URL of the Azure OpenAI service.
                    - For `"OPEN_AI_SDK"` type:
                        - `api_key`: The API key for OpenAI SDK.
                        - `endpoint`: The endpoint URL of the OpenAI SDK service.

        Returns:
            A dictionary containing the updated model's details.

        Examples:
            >>> # Update the name of a model
            >>> kili.llm.update_properties_in_model(
            ...     model_id="your_model_id",
            ...     model={"name": "Updated Model Name"}
            ... )
        """
        credentials_data = model.get("credentials")
        credentials = None

        if credentials_data:
            existing_model = self.kili_api_gateway.get_model(
                model_id=model_id, fields=["id", "type"]
            )
            if not existing_model:
                raise ValueError(f"Model with id {model_id} not found")
            model_type = ModelType(existing_model["type"])

            if model_type == ModelType.AZURE_OPEN_AI:
                credentials = AzureOpenAICredentials(**credentials_data)
            elif model_type == ModelType.OPEN_AI_SDK:
                credentials = OpenAISDKCredentials(**credentials_data)
            else:
                raise ValueError(f"Unsupported model type: {model_type}")

        model_input = ModelToUpdateInput(
            credentials=credentials,
            name=model.get("name"),
        )
        return self.kili_api_gateway.update_properties_in_model(
            model_id=model_id, model=model_input
        )

    def delete_model(self, model_id: str) -> bool:
        # pylint: disable=line-too-long
        """Delete a model from an organization.

        Args:
            model_id: Identifier of the model to delete.

        Returns:
            A dictionary indicating the result of the deletion.

        Examples:
            >>> kili.llm.delete_model(model_id="your_model_id")
        """
        return self.kili_api_gateway.delete_model(model_id=model_id)

    def create_project_model(
        self, project_id: str, model_id: str, configuration: dict
    ) -> ProjectModelDict:
        # pylint: disable=line-too-long
        """Associate a model with a project.

        Args:
            project_id: Identifier of the project.
            model_id: Identifier of the model to associate.
            configuration: Configuration parameters for the project model.

        Returns:
            A dictionary containing the created project model's details.

        Examples:
            >>> configuration = {
            ...     # Configuration details specific to your use case
            ... }
            >>> kili.llm.create_project_model(
            ...     project_id="your_project_id",
            ...     model_id="your_model_id",
            ...     configuration={"temperature": 0.7}
            ... )
        """
        project_model_input = ProjectModelToCreateInput(
            project_id=project_id, model_id=model_id, configuration=configuration
        )
        return self.kili_api_gateway.create_project_model(project_model=project_model_input)

    def project_models(
        self, project_id: str, filters: Optional[Dict] = None, fields: Optional[List[str]] = None
    ) -> List[ProjectModelDict]:
        """List models associated with a project.

        Args:
            project_id: Identifier of the project.
            filters: Optional filters to apply. Possible keys:
                - `model_id`: Identifier of a specific model to filter by.
            fields: All the fields to request among the possible fields for the project models.
                Defaults to ["configuration", "id", "model.credentials", "model.name", "model.type"].

        Returns:
            A list of project models.

        Examples:
            >>> kili.llm.project_models(project_id="your_project_id")
        """
        converted_filters = ProjectModelFilters(
            project_id=project_id,
            model_id=filters["model_id"] if filters and "model_id" in filters else None,
        )

        return list(
            self.kili_api_gateway.list_project_models(
                filters=converted_filters,
                fields=fields,
            )
        )

    def update_project_model(self, project_model_id: str, configuration: dict) -> ProjectModelDict:
        """Update the configuration of a project model.

        Args:
            project_model_id: Identifier of the project model to update.
            configuration: New configuration parameters.

        Returns:
            A dictionary containing the updated project model's details.

        Examples:
            >>> configuration = {
            ...     # Updated configuration details
            ... }
            >>> kili.llm.update_project_model(
            ...     project_model_id="your_project_model_id",
            ...     configuration=configuration
            ... )
        """
        project_model_input = ProjectModelToUpdateInput(configuration=configuration)
        return self.kili_api_gateway.update_project_model(
            project_model_id=project_model_id, project_model=project_model_input
        )

    def delete_project_model(self, project_model_id: str) -> bool:
        """Delete a project model.

        Args:
            project_model_id: Identifier of the project model to delete.

        Returns:
            A dictionary indicating the result of the deletion.

        Examples:
            >>> kili.llm.delete_project_model(project_model_id="your_project_model_id")
        """
        return self.kili_api_gateway.delete_project_model(project_model_id)

    def create_conversation(self, project_id: str, prompt: str) -> List[ChatItemDict]:
        # pylint: disable=line-too-long
        """Create a new conversation in an LLM project starting with a user's prompt.

        This method initiates a new conversation in the specified project by:
        - Creating an LLM asset and label associated with the current user.
        - Adding the user's prompt as the first chat item.
        - Automatically generating assistant responses using the project's models.

        Args:
            project_id: The identifier of the project where the conversation will be created.
            prompt: The initial prompt or message from the user to start the conversation.

        Returns:
            A list of chat items in the conversation, including the user's prompt and the assistant's responses.

        Examples:
            >>> PROMPT = "Hello, how can I improve my coding skills?"
            >>> chat_items = kili.llm.create_conversation(project_id="your_project_id", prompt=PROMPT)

        Notes:
            - The first chat item corresponds to the user's prompt.
            - The subsequent chat items are assistant responses generated by the project's models.
            - An LLM asset and a label are created in the project with status "TODO" and labelType "PREDICTION".
        """
        user_id = self.kili_api_gateway.get_current_user(["id"])["id"]
        llm_asset = self.kili_api_gateway.create_llm_asset(
            project_id=project_id,
            author_id=user_id,
            status="TODO",
            label_type="PREDICTION",
        )
        asset_id = llm_asset["id"]
        label_id = llm_asset["latestLabel"]["id"]
        return self.kili_api_gateway.create_chat_item(
            asset_id=asset_id, label_id=label_id, prompt=prompt
        )

create_conversation(self, project_id, prompt)

Create a new conversation in an LLM project starting with a user's prompt.

This method initiates a new conversation in the specified project by: - Creating an LLM asset and label associated with the current user. - Adding the user's prompt as the first chat item. - Automatically generating assistant responses using the project's models.

Parameters:

Name Type Description Default
project_id str

The identifier of the project where the conversation will be created.

required
prompt str

The initial prompt or message from the user to start the conversation.

required

Returns:

Type Description
List[kili.domain.llm.ChatItemDict]

A list of chat items in the conversation, including the user's prompt and the assistant's responses.

Examples:

>>> PROMPT = "Hello, how can I improve my coding skills?"
>>> chat_items = kili.llm.create_conversation(project_id="your_project_id", prompt=PROMPT)

Notes

  • The first chat item corresponds to the user's prompt.
  • The subsequent chat items are assistant responses generated by the project's models.
  • An LLM asset and a label are created in the project with status "TODO" and labelType "PREDICTION".
Source code in kili/llm/presentation/client/llm.py
def create_conversation(self, project_id: str, prompt: str) -> List[ChatItemDict]:
    # pylint: disable=line-too-long
    """Create a new conversation in an LLM project starting with a user's prompt.

    This method initiates a new conversation in the specified project by:
    - Creating an LLM asset and label associated with the current user.
    - Adding the user's prompt as the first chat item.
    - Automatically generating assistant responses using the project's models.

    Args:
        project_id: The identifier of the project where the conversation will be created.
        prompt: The initial prompt or message from the user to start the conversation.

    Returns:
        A list of chat items in the conversation, including the user's prompt and the assistant's responses.

    Examples:
        >>> PROMPT = "Hello, how can I improve my coding skills?"
        >>> chat_items = kili.llm.create_conversation(project_id="your_project_id", prompt=PROMPT)

    Notes:
        - The first chat item corresponds to the user's prompt.
        - The subsequent chat items are assistant responses generated by the project's models.
        - An LLM asset and a label are created in the project with status "TODO" and labelType "PREDICTION".
    """
    user_id = self.kili_api_gateway.get_current_user(["id"])["id"]
    llm_asset = self.kili_api_gateway.create_llm_asset(
        project_id=project_id,
        author_id=user_id,
        status="TODO",
        label_type="PREDICTION",
    )
    asset_id = llm_asset["id"]
    label_id = llm_asset["latestLabel"]["id"]
    return self.kili_api_gateway.create_chat_item(
        asset_id=asset_id, label_id=label_id, prompt=prompt
    )

create_model(self, organization_id, model)

Create a new model in an organization.

Parameters:

Name Type Description Default
organization_id str

Identifier of the organization.

required
model dict

A dictionary representing the model to create, containing: - name: Name of the model. - type: Type of the model, one of: - "AZURE_OPEN_AI" - "OPEN_AI_SDK" - credentials: Credentials required for the model, depending on the type: - For "AZURE_OPEN_AI" type: - api_key: The API key for Azure OpenAI. - deployment_id: The deployment ID within Azure. - endpoint: The endpoint URL of the Azure OpenAI service. - For "OPEN_AI_SDK" type: - api_key: The API key for OpenAI SDK. - endpoint: The endpoint URL of the OpenAI SDK service.

required

Returns:

Type Description
ModelDict

A dictionary containing the created model's details.

Examples:

>>> # Example of creating an OpenAI SDK model
>>> model_data = {
...     "name": "My OpenAI SDK Model",
...     "type": "OPEN_AI_SDK",
...     "credentials": {
...         "api_key": "your_open_ai_api_key",
...         "endpoint": "your_open_ai_endpoint"
...     }
... }
>>> kili.llm.create_model(organization_id="your_organization_id", model=model_data)
Source code in kili/llm/presentation/client/llm.py
def create_model(self, organization_id: str, model: dict) -> ModelDict:
    # pylint: disable=line-too-long
    """Create a new model in an organization.

    Args:
        organization_id: Identifier of the organization.
        model: A dictionary representing the model to create, containing:
            - `name`: Name of the model.
            - `type`: Type of the model, one of:
                - `"AZURE_OPEN_AI"`
                - `"OPEN_AI_SDK"`
            - `credentials`: Credentials required for the model, depending on the type:
                - For `"AZURE_OPEN_AI"` type:
                    - `api_key`: The API key for Azure OpenAI.
                    - `deployment_id`: The deployment ID within Azure.
                    - `endpoint`: The endpoint URL of the Azure OpenAI service.
                - For `"OPEN_AI_SDK"` type:
                    - `api_key`: The API key for OpenAI SDK.
                    - `endpoint`: The endpoint URL of the OpenAI SDK service.

    Returns:
        A dictionary containing the created model's details.

    Examples:
        >>> # Example of creating an OpenAI SDK model
        >>> model_data = {
        ...     "name": "My OpenAI SDK Model",
        ...     "type": "OPEN_AI_SDK",
        ...     "credentials": {
        ...         "api_key": "your_open_ai_api_key",
        ...         "endpoint": "your_open_ai_endpoint"
        ...     }
        ... }
        >>> kili.llm.create_model(organization_id="your_organization_id", model=model_data)
    """
    credentials_data = model["credentials"]
    model_type = ModelType(model["type"])

    if model_type == ModelType.AZURE_OPEN_AI:
        credentials = AzureOpenAICredentials(**credentials_data)
    elif model_type == ModelType.OPEN_AI_SDK:
        credentials = OpenAISDKCredentials(**credentials_data)
    else:
        raise ValueError(f"Unsupported model type: {model['type']}")

    model_input = ModelToCreateInput(
        credentials=credentials,
        name=model["name"],
        type=model_type,
        organization_id=organization_id,
    )
    return self.kili_api_gateway.create_model(model=model_input)

create_project_model(self, project_id, model_id, configuration)

Associate a model with a project.

Parameters:

Name Type Description Default
project_id str

Identifier of the project.

required
model_id str

Identifier of the model to associate.

required
configuration dict

Configuration parameters for the project model.

required

Returns:

Type Description
ProjectModelDict

A dictionary containing the created project model's details.

Examples:

>>> configuration = {
...     # Configuration details specific to your use case
... }
>>> kili.llm.create_project_model(
...     project_id="your_project_id",
...     model_id="your_model_id",
...     configuration={"temperature": 0.7}
... )
Source code in kili/llm/presentation/client/llm.py
def create_project_model(
    self, project_id: str, model_id: str, configuration: dict
) -> ProjectModelDict:
    # pylint: disable=line-too-long
    """Associate a model with a project.

    Args:
        project_id: Identifier of the project.
        model_id: Identifier of the model to associate.
        configuration: Configuration parameters for the project model.

    Returns:
        A dictionary containing the created project model's details.

    Examples:
        >>> configuration = {
        ...     # Configuration details specific to your use case
        ... }
        >>> kili.llm.create_project_model(
        ...     project_id="your_project_id",
        ...     model_id="your_model_id",
        ...     configuration={"temperature": 0.7}
        ... )
    """
    project_model_input = ProjectModelToCreateInput(
        project_id=project_id, model_id=model_id, configuration=configuration
    )
    return self.kili_api_gateway.create_project_model(project_model=project_model_input)

delete_model(self, model_id)

Delete a model from an organization.

Parameters:

Name Type Description Default
model_id str

Identifier of the model to delete.

required

Returns:

Type Description
bool

A dictionary indicating the result of the deletion.

Examples:

>>> kili.llm.delete_model(model_id="your_model_id")
Source code in kili/llm/presentation/client/llm.py
def delete_model(self, model_id: str) -> bool:
    # pylint: disable=line-too-long
    """Delete a model from an organization.

    Args:
        model_id: Identifier of the model to delete.

    Returns:
        A dictionary indicating the result of the deletion.

    Examples:
        >>> kili.llm.delete_model(model_id="your_model_id")
    """
    return self.kili_api_gateway.delete_model(model_id=model_id)

delete_project_model(self, project_model_id)

Delete a project model.

Parameters:

Name Type Description Default
project_model_id str

Identifier of the project model to delete.

required

Returns:

Type Description
bool

A dictionary indicating the result of the deletion.

Examples:

>>> kili.llm.delete_project_model(project_model_id="your_project_model_id")
Source code in kili/llm/presentation/client/llm.py
def delete_project_model(self, project_model_id: str) -> bool:
    """Delete a project model.

    Args:
        project_model_id: Identifier of the project model to delete.

    Returns:
        A dictionary indicating the result of the deletion.

    Examples:
        >>> kili.llm.delete_project_model(project_model_id="your_project_model_id")
    """
    return self.kili_api_gateway.delete_project_model(project_model_id)

export(self, project_id, disable_tqdm=False, asset_ids=None, external_ids=None)

Returns an export of llm assets with valid labels.

Parameters:

Name Type Description Default
project_id str

Identifier of the project.

required
asset_ids Optional[List[str]]

Optional list of the assets internal IDs from which to export the labels.

None
disable_tqdm Optional[bool]

Disable the progress bar if True.

False
external_ids Optional[List[str]]

Optional list of the assets external IDs from which to export the labels.

None

Example

kili.llm.export("your_project_id")
Source code in kili/llm/presentation/client/llm.py
def export(
    self,
    project_id: str,
    disable_tqdm: Optional[bool] = False,
    asset_ids: Optional[List[str]] = None,
    external_ids: Optional[List[str]] = None,
) -> Optional[List[Dict[str, Union[List[str], str]]]]:
    # pylint: disable=line-too-long
    """Returns an export of llm assets with valid labels.

    Args:
        project_id: Identifier of the project.
        asset_ids: Optional list of the assets internal IDs from which to export the labels.
        disable_tqdm: Disable the progress bar if True.
        external_ids: Optional list of the assets external IDs from which to export the labels.

    !!! Example
        ```python
        kili.llm.export("your_project_id")
        ```
    """
    if asset_ids and external_ids:
        raise ValueError("You cannot provide both asset_ids and external_ids")

    if external_ids is not None and asset_ids is None:
        id_map = AssetUseCasesUtils(self.kili_api_gateway).infer_ids_from_external_ids(
            asset_external_ids=cast(List[AssetExternalId], external_ids),
            project_id=ProjectId(project_id),
        )
        resolved_asset_ids = [id_map[AssetExternalId(i)] for i in external_ids]
    else:
        resolved_asset_ids = (
            list(AssetId(asset_id) for asset_id in asset_ids) if asset_ids else None
        )

    asset_filter = AssetFilters(
        project_id=ProjectId(project_id), asset_id_in=resolved_asset_ids
    )

    try:
        return export(
            self.kili_api_gateway,
            project_id=ProjectId(project_id),
            asset_filter=asset_filter,
            disable_tqdm=disable_tqdm,
        )
    except NoCompatibleJobError as excp:
        warnings.warn(str(excp), stacklevel=2)
        return None

model(self, model_id, fields=None)

Retrieve a specific model.

Parameters:

Name Type Description Default
model_id str

Identifier of the model.

required
fields Optional[List[str]]

All the fields to request among the possible fields for the models. Defaults to ["id", "credentials", "name", "type"].

None

Returns:

Type Description
ModelDict

A dictionary representing the model.

Examples:

>>> kili.llm.model(model_id="your_model_id")
Source code in kili/llm/presentation/client/llm.py
def model(self, model_id: str, fields: Optional[List[str]] = None) -> ModelDict:
    # pylint: disable=line-too-long
    """Retrieve a specific model.

    Args:
        model_id: Identifier of the model.
        fields: All the fields to request among the possible fields for the models.
            Defaults to ["id", "credentials", "name", "type"].

    Returns:
        A dictionary representing the model.

    Examples:
        >>> kili.llm.model(model_id="your_model_id")
    """
    return self.kili_api_gateway.get_model(
        model_id=model_id,
        fields=fields,
    )

models(self, organization_id, fields=None)

List models in an organization.

Parameters:

Name Type Description Default
organization_id str

Identifier of the organization.

required
fields Optional[List[str]]

All the fields to request among the possible fields for the models. Defaults to ["id", "credentials", "name", "type"].

None

Returns:

Type Description
List[kili.domain.llm.ModelDict]

A list of models.

Examples:

>>> kili.llm.models(organization_id="your_organization_id")
Source code in kili/llm/presentation/client/llm.py
def models(self, organization_id: str, fields: Optional[List[str]] = None) -> List[ModelDict]:
    # pylint: disable=line-too-long
    """List models in an organization.

    Args:
        organization_id: Identifier of the organization.
        fields: All the fields to request among the possible fields for the models.
            Defaults to ["id", "credentials", "name", "type"].

    Returns:
        A list of models.

    Examples:
        >>> kili.llm.models(organization_id="your_organization_id")
    """
    converted_filters = OrganizationModelFilters(
        organization_id=organization_id,
    )

    return list(self.kili_api_gateway.list_models(filters=converted_filters, fields=fields))

project_models(self, project_id, filters=None, fields=None)

List models associated with a project.

Parameters:

Name Type Description Default
project_id str

Identifier of the project.

required
filters Optional[Dict]

Optional filters to apply. Possible keys: - model_id: Identifier of a specific model to filter by.

None
fields Optional[List[str]]

All the fields to request among the possible fields for the project models. Defaults to ["configuration", "id", "model.credentials", "model.name", "model.type"].

None

Returns:

Type Description
List[kili.domain.llm.ProjectModelDict]

A list of project models.

Examples:

>>> kili.llm.project_models(project_id="your_project_id")
Source code in kili/llm/presentation/client/llm.py
def project_models(
    self, project_id: str, filters: Optional[Dict] = None, fields: Optional[List[str]] = None
) -> List[ProjectModelDict]:
    """List models associated with a project.

    Args:
        project_id: Identifier of the project.
        filters: Optional filters to apply. Possible keys:
            - `model_id`: Identifier of a specific model to filter by.
        fields: All the fields to request among the possible fields for the project models.
            Defaults to ["configuration", "id", "model.credentials", "model.name", "model.type"].

    Returns:
        A list of project models.

    Examples:
        >>> kili.llm.project_models(project_id="your_project_id")
    """
    converted_filters = ProjectModelFilters(
        project_id=project_id,
        model_id=filters["model_id"] if filters and "model_id" in filters else None,
    )

    return list(
        self.kili_api_gateway.list_project_models(
            filters=converted_filters,
            fields=fields,
        )
    )

update_project_model(self, project_model_id, configuration)

Update the configuration of a project model.

Parameters:

Name Type Description Default
project_model_id str

Identifier of the project model to update.

required
configuration dict

New configuration parameters.

required

Returns:

Type Description
ProjectModelDict

A dictionary containing the updated project model's details.

Examples:

>>> configuration = {
...     # Updated configuration details
... }
>>> kili.llm.update_project_model(
...     project_model_id="your_project_model_id",
...     configuration=configuration
... )
Source code in kili/llm/presentation/client/llm.py
def update_project_model(self, project_model_id: str, configuration: dict) -> ProjectModelDict:
    """Update the configuration of a project model.

    Args:
        project_model_id: Identifier of the project model to update.
        configuration: New configuration parameters.

    Returns:
        A dictionary containing the updated project model's details.

    Examples:
        >>> configuration = {
        ...     # Updated configuration details
        ... }
        >>> kili.llm.update_project_model(
        ...     project_model_id="your_project_model_id",
        ...     configuration=configuration
        ... )
    """
    project_model_input = ProjectModelToUpdateInput(configuration=configuration)
    return self.kili_api_gateway.update_project_model(
        project_model_id=project_model_id, project_model=project_model_input
    )

update_properties_in_model(self, model_id, model)

Update properties of an existing model.

Parameters:

Name Type Description Default
model_id str

Identifier of the model to update.

required
model dict

A dictionary containing the properties to update, which may include: - name: New name of the model. - credentials: Updated credentials for the model, depending on the type: - For "AZURE_OPEN_AI" type: - api_key: The API key for Azure OpenAI. - deployment_id: The deployment ID within Azure. - endpoint: The endpoint URL of the Azure OpenAI service. - For "OPEN_AI_SDK" type: - api_key: The API key for OpenAI SDK. - endpoint: The endpoint URL of the OpenAI SDK service.

required

Returns:

Type Description
ModelDict

A dictionary containing the updated model's details.

Examples:

>>> # Update the name of a model
>>> kili.llm.update_properties_in_model(
...     model_id="your_model_id",
...     model={"name": "Updated Model Name"}
... )
Source code in kili/llm/presentation/client/llm.py
def update_properties_in_model(self, model_id: str, model: dict) -> ModelDict:
    # pylint: disable=line-too-long
    """Update properties of an existing model.

    Args:
        model_id: Identifier of the model to update.
        model: A dictionary containing the properties to update, which may include:
            - `name`: New name of the model.
            - `credentials`: Updated credentials for the model, depending on the type:
                - For `"AZURE_OPEN_AI"` type:
                    - `api_key`: The API key for Azure OpenAI.
                    - `deployment_id`: The deployment ID within Azure.
                    - `endpoint`: The endpoint URL of the Azure OpenAI service.
                - For `"OPEN_AI_SDK"` type:
                    - `api_key`: The API key for OpenAI SDK.
                    - `endpoint`: The endpoint URL of the OpenAI SDK service.

    Returns:
        A dictionary containing the updated model's details.

    Examples:
        >>> # Update the name of a model
        >>> kili.llm.update_properties_in_model(
        ...     model_id="your_model_id",
        ...     model={"name": "Updated Model Name"}
        ... )
    """
    credentials_data = model.get("credentials")
    credentials = None

    if credentials_data:
        existing_model = self.kili_api_gateway.get_model(
            model_id=model_id, fields=["id", "type"]
        )
        if not existing_model:
            raise ValueError(f"Model with id {model_id} not found")
        model_type = ModelType(existing_model["type"])

        if model_type == ModelType.AZURE_OPEN_AI:
            credentials = AzureOpenAICredentials(**credentials_data)
        elif model_type == ModelType.OPEN_AI_SDK:
            credentials = OpenAISDKCredentials(**credentials_data)
        else:
            raise ValueError(f"Unsupported model type: {model_type}")

    model_input = ModelToUpdateInput(
        credentials=credentials,
        name=model.get("name"),
    )
    return self.kili_api_gateway.update_properties_in_model(
        model_id=model_id, model=model_input
    )