vibe.assistant.services.provider_factory

ProviderFactory - creates and configures LLM providers.

Responsibilities: - Endpoint resolution (including Jinja template evaluation) - Development overrides (endpoint switching, tool toggles) - Playback from log files - SystemProxyProvider wrapping for assistant mode - Tool configuration

ProviderWithConfig

Result of creating a provider, includes both provider and config.

ProviderFactory

Factory for creating and configuring LLM providers.

This class consolidates: - get_assistant_provider - _get_assisted_provider - _get_endpoint_name_for_assistant - _get_endpoint_config_for_provider - _apply_dev_overrides

__init__ 

__init__(assistant_name: str, endpoints: dict[str, Any], session_getter: Callable[[], dict[str, Any]], debug_mode: bool = False, interview_mode: str = 'standard', question_definitions: dict[str, Any] | None = None) -> None

Initialize the provider factory.

Parameters:
  • assistant_name (str) –

    Name of the assistant

  • endpoints (dict[str, Any]) –

    Dict of endpoint configurations (from app.LLM_ENDPOINTS)

  • session_getter (Callable[[], dict[str, Any]]) –

    Callable that returns the Flask session dict

  • debug_mode (bool, default: False ) –

    Whether the app is in debug mode

  • interview_mode (str, default: 'standard' ) –

    Interview mode ("standard" or "assistant")

  • question_definitions (dict[str, Any] | None, default: None ) –

    Optional dict of question definitions

resolve_endpoint_name 

resolve_endpoint_name(endpoint_template: str) -> str

Resolve endpoint name from template string.

Parameters:
  • endpoint_template (str) –

    Endpoint name or Jinja template string
Returns:
  • str –

    Resolved endpoint name

get_endpoint_config 

get_endpoint_config(endpoint_name: str) -> dict[str, Any]

Get the full endpoint configuration.

Parameters:
  • endpoint_name (str) –

    Resolved endpoint name
Returns:
  • dict[str, Any] –

    Endpoint configuration dict
Raises:
  • ValueError –

    If endpoint is not defined

create 

create(endpoint_template: str, get_pending_questions: Callable[[], list[str]] | None = None) -> ProviderWithConfig

Create a configured provider instance.

Parameters:
  • endpoint_template (str) –

    Endpoint name or Jinja template

  • get_pending_questions (Callable[[], list[str]] | None, default: None ) –

    Optional callback for SystemProxyProvider (required if interview_mode == "assistant")
Returns:

configure_tools 

configure_tools(provider: LLMProvider, unanswered_questions: list[dict[str, Any]]) -> None

Configure tools on a provider.

Parameters:
  • provider (LLMProvider) –

    Provider instance to configure

  • unanswered_questions (list[dict[str, Any]]) –

    List of unanswered predefined questions

create_assisted_provider 

create_assisted_provider(endpoint_name: str) -> LLMProvider

Create a provider for assisted text generation (tools disabled).

Parameters:
  • endpoint_name (str) –

    Endpoint name to use
Returns:
  • LLMProvider –

    Provider instance with tools disabled
Raises:
  • ValueError –

    If endpoint not found and no default