diff options
Diffstat (limited to 'autogpts/autogpt/autogpt/core/planning/prompt_strategies/initial_plan.py')
| -rw-r--r-- | autogpts/autogpt/autogpt/core/planning/prompt_strategies/initial_plan.py | 206 |
1 files changed, 206 insertions, 0 deletions
diff --git a/autogpts/autogpt/autogpt/core/planning/prompt_strategies/initial_plan.py b/autogpts/autogpt/autogpt/core/planning/prompt_strategies/initial_plan.py new file mode 100644 index 000000000..9912fb6e2 --- /dev/null +++ b/autogpts/autogpt/autogpt/core/planning/prompt_strategies/initial_plan.py @@ -0,0 +1,206 @@ +import logging + +from autogpt.core.configuration import SystemConfiguration, UserConfigurable +from autogpt.core.planning.schema import Task, TaskType +from autogpt.core.prompting import PromptStrategy +from autogpt.core.prompting.schema import ChatPrompt, LanguageModelClassification +from autogpt.core.prompting.utils import json_loads, to_numbered_list +from autogpt.core.resource.model_providers import ( + AssistantChatMessage, + ChatMessage, + CompletionModelFunction, +) +from autogpt.core.utils.json_schema import JSONSchema + +logger = logging.getLogger(__name__) + + +class InitialPlanConfiguration(SystemConfiguration): + model_classification: LanguageModelClassification = UserConfigurable() + system_prompt_template: str = UserConfigurable() + system_info: list[str] = UserConfigurable() + user_prompt_template: str = UserConfigurable() + create_plan_function: dict = UserConfigurable() + + +class InitialPlan(PromptStrategy): + DEFAULT_SYSTEM_PROMPT_TEMPLATE = ( + "You are an expert project planner. " + "Your responsibility is to create work plans for autonomous agents. " + "You will be given a name, a role, set of goals for the agent to accomplish. " + "Your job is to break down those goals into a set of tasks that the agent can" + " accomplish to achieve those goals. " + "Agents are resourceful, but require clear instructions." + " Each task you create should have clearly defined `ready_criteria` that the" + " agent can check to see if the task is ready to be started." + " Each task should also have clearly defined `acceptance_criteria` that the" + " agent can check to evaluate if the task is complete. " + "You should create as many tasks as you think is necessary to accomplish" + " the goals.\n\n" + "System Info:\n{system_info}" + ) + + DEFAULT_SYSTEM_INFO = [ + "The OS you are running on is: {os_info}", + "It takes money to let you run. Your API budget is ${api_budget:.3f}", + "The current time and date is {current_time}", + ] + + DEFAULT_USER_PROMPT_TEMPLATE = ( + "You are {agent_name}, {agent_role}\n" "Your goals are:\n" "{agent_goals}" + ) + + DEFAULT_CREATE_PLAN_FUNCTION = CompletionModelFunction( + name="create_initial_agent_plan", + description=( + "Creates a set of tasks that forms the initial plan of an autonomous agent." + ), + parameters={ + "task_list": JSONSchema( + type=JSONSchema.Type.ARRAY, + items=JSONSchema( + type=JSONSchema.Type.OBJECT, + properties={ + "objective": JSONSchema( + type=JSONSchema.Type.STRING, + description=( + "An imperative verb phrase that succinctly describes " + "the task." + ), + ), + "type": JSONSchema( + type=JSONSchema.Type.STRING, + description="A categorization for the task.", + enum=[t.value for t in TaskType], + ), + "acceptance_criteria": JSONSchema( + type=JSONSchema.Type.ARRAY, + items=JSONSchema( + type=JSONSchema.Type.STRING, + description=( + "A list of measurable and testable criteria that " + "must be met for the task to be considered " + "complete." + ), + ), + ), + "priority": JSONSchema( + type=JSONSchema.Type.INTEGER, + description=( + "A number between 1 and 10 indicating the priority of " + "the task relative to other generated tasks." + ), + minimum=1, + maximum=10, + ), + "ready_criteria": JSONSchema( + type=JSONSchema.Type.ARRAY, + items=JSONSchema( + type=JSONSchema.Type.STRING, + description=( + "A list of measurable and testable criteria that " + "must be met before the task can be started." + ), + ), + ), + }, + ), + ), + }, + ) + + default_configuration: InitialPlanConfiguration = InitialPlanConfiguration( + model_classification=LanguageModelClassification.SMART_MODEL, + system_prompt_template=DEFAULT_SYSTEM_PROMPT_TEMPLATE, + system_info=DEFAULT_SYSTEM_INFO, + user_prompt_template=DEFAULT_USER_PROMPT_TEMPLATE, + create_plan_function=DEFAULT_CREATE_PLAN_FUNCTION.schema, + ) + + def __init__( + self, + model_classification: LanguageModelClassification, + system_prompt_template: str, + system_info: list[str], + user_prompt_template: str, + create_plan_function: dict, + ): + self._model_classification = model_classification + self._system_prompt_template = system_prompt_template + self._system_info = system_info + self._user_prompt_template = user_prompt_template + self._create_plan_function = CompletionModelFunction.parse(create_plan_function) + + @property + def model_classification(self) -> LanguageModelClassification: + return self._model_classification + + def build_prompt( + self, + agent_name: str, + agent_role: str, + agent_goals: list[str], + abilities: list[str], + os_info: str, + api_budget: float, + current_time: str, + **kwargs, + ) -> ChatPrompt: + template_kwargs = { + "agent_name": agent_name, + "agent_role": agent_role, + "os_info": os_info, + "api_budget": api_budget, + "current_time": current_time, + **kwargs, + } + template_kwargs["agent_goals"] = to_numbered_list( + agent_goals, **template_kwargs + ) + template_kwargs["abilities"] = to_numbered_list(abilities, **template_kwargs) + template_kwargs["system_info"] = to_numbered_list( + self._system_info, **template_kwargs + ) + + system_prompt = ChatMessage.system( + self._system_prompt_template.format(**template_kwargs), + ) + user_prompt = ChatMessage.user( + self._user_prompt_template.format(**template_kwargs), + ) + + return ChatPrompt( + messages=[system_prompt, user_prompt], + functions=[self._create_plan_function], + # TODO: + tokens_used=0, + ) + + def parse_response_content( + self, + response_content: AssistantChatMessage, + ) -> dict: + """Parse the actual text response from the objective model. + + Args: + response_content: The raw response content from the objective model. + + Returns: + The parsed response. + """ + try: + if not response_content.tool_calls: + raise ValueError( + f"LLM did not call {self._create_plan_function.name} function; " + "plan creation failed" + ) + parsed_response: object = json_loads( + response_content.tool_calls[0].function.arguments + ) + parsed_response["task_list"] = [ + Task.parse_obj(task) for task in parsed_response["task_list"] + ] + except KeyError: + logger.debug(f"Failed to parse this response content: {response_content}") + raise + return parsed_response |