PrivateAssistant

An AI assistant that privately helps one or more speakers in a Conversation. See ConvoKit AI.

The PrivateAssistant class: an AI assistant that privately supports speakers in a Conversation.

class convokit.model.privateAssistant.PrivateAssistant(id: str, config: Optional[Dict] = None, supports: Optional[Iterable] = None, speakers: Optional[Iterable[str]] = None, conversation_id: Optional[str] = None)

Represents an AI private assistant that sends Supports to one or more speakers in a Conversation.

Private assistants are not speakers in the Conversation: their messages (Supports) are seen only by the speakers they assist.

Parameters

  • id – the unique id of the assistant

  • config – configuration of the assistant, e.g. {"prompt": "...", "model": "gemini-2.5-flash", "temperature": 0.7}. config["prompt"] is the prompt that generates a support message from this assistant (required by generate(); see convokit.model.aiGeneration for the other keys).

  • supports – the Supports (or their dict forms) produced by this assistant

  • speakers – ids of the Speakers that can see this assistant’s Supports

  • conversation_id – id of the Conversation the assistant belongs to

Variables

  • id – the unique id of the assistant

  • config – configuration of the assistant

  • supports – list of the Supports produced by this assistant

  • speakers – ids of the Speakers that can see this assistant’s Supports

  • conversation_id – id of the Conversation the assistant belongs to

classmethod from_dict(data: Dict) → Optional[convokit.model.privateAssistant.PrivateAssistant]

Build a PrivateAssistant from its dict form (as produced by to_dict()).

Parameters

data – the dict form of the assistant

Returns

the PrivateAssistant, or None if data is not a dict

generate(conversation=None, reply_to: Optional[str] = None, draft: str = '', append: bool = False, client=None, config_manager=None, id: Optional[str] = None, timestamp: Optional[int] = None) → convokit.model.support.Support

Generate a Support from this assistant with convokit.genai, using the prompt in config["prompt"] (see convokit.model.aiGeneration for the recognized config keys).

Example:

assistant = conversation.get_private_assistant("assistant_1")
support = assistant.generate(conversation, draft="I think", append=True)
Parameters

  • conversation – if given, the conversation transcript (with Supports) up to reply_to is included in the prompt, along with earlier Supports replying to reply_to

  • reply_to – id of the utterance the assisted speaker is replying to (defaults to the last utterance of conversation)

  • draft – the draft the assisted speaker has written so far

  • append – whether to add the generated support to this assistant and conversation

  • client – a convokit.genai LLMClient to use instead of building one from the config

  • config_manager – GenAIConfigManager used to build the client (defaults to one reading ~/.convokit/config.yml)

  • id – id of the generated support (random if not given)

  • timestamp – timestamp of the generated support (defaults to after the conversation’s last message)

Returns

the generated Support

Raises

ValueError – if config has no "prompt", or if append is True and no conversation is given

static normalize_list(value) → List[convokit.model.privateAssistant.PrivateAssistant]

Convert a list (or dict of id -> value) of PrivateAssistants and/or dicts into a list of PrivateAssistants. Entries that are neither are dropped.

Parameters

value – a list or dict of PrivateAssistants and/or their dict forms (or None)

Returns

a list of PrivateAssistants

to_dict() → Dict

Convert this assistant (including its Supports) into a JSON-serializable dict.

Returns

dict with keys "id", "config", "supports", "speakers", and "conversation_id"