mirror of
https://github.com/Pythagora-io/gpt-pilot.git
synced 2026-01-08 12:53:50 -05:00
5b474ccc1f
This is a complete rewrite of the GPT Pilot core, from the ground up, making the agentic architecture front and center, and also fixing some long-standing problems with the database architecture that weren't feasible to solve without breaking compatibility. As the database structure and config file syntax have changed, we have automatic imports for projects and current configs, see the README.md file for details. This also relicenses the project to FSL-1.1-MIT license.
295 lines
12 KiB
Python
295 lines
12 KiB
Python
from enum import Enum
|
|
from typing import Annotated, Literal, Optional, Union
|
|
from uuid import uuid4
|
|
|
|
from pydantic import BaseModel, Field
|
|
|
|
from core.agents.base import BaseAgent
|
|
from core.agents.convo import AgentConvo
|
|
from core.agents.response import AgentResponse, ResponseType
|
|
from core.llm.parser import JSONParser
|
|
from core.log import get_logger
|
|
|
|
log = get_logger(__name__)
|
|
|
|
|
|
class StepType(str, Enum):
|
|
COMMAND = "command"
|
|
SAVE_FILE = "save_file"
|
|
HUMAN_INTERVENTION = "human_intervention"
|
|
|
|
|
|
class CommandOptions(BaseModel):
|
|
command: str = Field(description="Command to run")
|
|
timeout: int = Field(description="Timeout in seconds")
|
|
success_message: str = ""
|
|
|
|
|
|
class SaveFileOptions(BaseModel):
|
|
path: str
|
|
|
|
|
|
class SaveFileStep(BaseModel):
|
|
type: Literal[StepType.SAVE_FILE] = StepType.SAVE_FILE
|
|
save_file: SaveFileOptions
|
|
|
|
|
|
class CommandStep(BaseModel):
|
|
type: Literal[StepType.COMMAND] = StepType.COMMAND
|
|
command: CommandOptions
|
|
|
|
|
|
class HumanInterventionStep(BaseModel):
|
|
type: Literal[StepType.HUMAN_INTERVENTION] = StepType.HUMAN_INTERVENTION
|
|
human_intervention_description: str
|
|
|
|
|
|
Step = Annotated[
|
|
Union[SaveFileStep, CommandStep, HumanInterventionStep],
|
|
Field(discriminator="type"),
|
|
]
|
|
|
|
|
|
class TaskSteps(BaseModel):
|
|
steps: list[Step]
|
|
|
|
|
|
class Developer(BaseAgent):
|
|
agent_type = "developer"
|
|
display_name = "Developer"
|
|
|
|
async def run(self) -> AgentResponse:
|
|
if self.prev_response and self.prev_response.type == ResponseType.TASK_REVIEW_FEEDBACK:
|
|
return await self.breakdown_current_iteration(self.prev_response.data["feedback"])
|
|
|
|
# If any of the files are missing metadata/descriptions, those need to be filled-in
|
|
missing_descriptions = [file.path for file in self.current_state.files if not file.meta.get("description")]
|
|
if missing_descriptions:
|
|
log.debug(f"Some files are missing descriptions: {', '.join(missing_descriptions)}, reqesting analysis")
|
|
return AgentResponse.describe_files(self)
|
|
|
|
log.debug(f"Current state files: {len(self.current_state.files)}, relevant {self.current_state.relevant_files}")
|
|
# Check which files are relevant to the current task
|
|
if self.current_state.files and not self.current_state.relevant_files:
|
|
await self.get_relevant_files()
|
|
return AgentResponse.done(self)
|
|
|
|
if not self.current_state.unfinished_tasks:
|
|
log.warning("No unfinished tasks found, nothing to do (why am I called? is this a bug?)")
|
|
return AgentResponse.done(self)
|
|
|
|
if self.current_state.unfinished_iterations:
|
|
return await self.breakdown_current_iteration()
|
|
|
|
# By default, we want to ask the user if they want to run the task,
|
|
# except in certain cases (such as they've just edited it).
|
|
if not self.current_state.current_task.get("run_always", False):
|
|
if not await self.ask_to_execute_task():
|
|
return AgentResponse.done(self)
|
|
|
|
return await self.breakdown_current_task()
|
|
|
|
async def breakdown_current_iteration(self, review_feedback: Optional[str] = None) -> AgentResponse:
|
|
"""
|
|
Breaks down current iteration or task review into steps.
|
|
|
|
:param review_feedback: If provided, the task review feedback is broken down instead of the current iteration
|
|
:return: AgentResponse.done(self) when the breakdown is done
|
|
"""
|
|
if self.current_state.unfinished_steps:
|
|
# if this happens, it's most probably a bug as we should have gone through all the
|
|
# steps before getting new new iteration instructions
|
|
log.warning(
|
|
f"Unfinished steps found before the next iteration is broken down: {self.current_state.unfinished_steps}"
|
|
)
|
|
|
|
if review_feedback is not None:
|
|
iteration = None
|
|
description = review_feedback
|
|
user_feedback = ""
|
|
source = "review"
|
|
n_tasks = 1
|
|
log.debug(f"Breaking down the task review feedback {review_feedback}")
|
|
await self.send_message("Breaking down the task review feedback...")
|
|
else:
|
|
iteration = self.current_state.current_iteration
|
|
if iteration is None:
|
|
log.error("Iteration breakdown called but there's no current iteration or task review, possible bug?")
|
|
return AgentResponse.done(self)
|
|
|
|
description = iteration["description"]
|
|
user_feedback = iteration["user_feedback"]
|
|
source = "troubleshooting"
|
|
n_tasks = len(self.next_state.iterations)
|
|
log.debug(f"Breaking down the iteration {description}")
|
|
await self.send_message("Breaking down the current task iteration ...")
|
|
|
|
await self.ui.send_task_progress(
|
|
n_tasks, # iterations and reviews can be created only one at a time, so we are always on last one
|
|
n_tasks,
|
|
self.current_state.current_task["description"],
|
|
source,
|
|
"in-progress",
|
|
)
|
|
llm = self.get_llm()
|
|
# FIXME: In case of iteration, parse_task depends on the context (files, tasks, etc) set there.
|
|
# Ideally this prompt would be self-contained.
|
|
convo = (
|
|
AgentConvo(self)
|
|
.template(
|
|
"iteration",
|
|
current_task=self.current_state.current_task,
|
|
user_feedback=user_feedback,
|
|
user_feedback_qa=None,
|
|
next_solution_to_try=None,
|
|
)
|
|
.assistant(description)
|
|
.template("parse_task")
|
|
.require_schema(TaskSteps)
|
|
)
|
|
response: TaskSteps = await llm(convo, parser=JSONParser(TaskSteps), temperature=0)
|
|
|
|
self.set_next_steps(response, source)
|
|
|
|
if iteration:
|
|
self.next_state.complete_iteration()
|
|
|
|
return AgentResponse.done(self)
|
|
|
|
async def breakdown_current_task(self) -> AgentResponse:
|
|
task = self.current_state.current_task
|
|
source = self.current_state.current_epic.get("source", "app")
|
|
await self.ui.send_task_progress(
|
|
self.current_state.tasks.index(self.current_state.current_task) + 1,
|
|
len(self.current_state.tasks),
|
|
self.current_state.current_task["description"],
|
|
source,
|
|
"in-progress",
|
|
)
|
|
|
|
log.debug(f"Breaking down the current task: {task['description']}")
|
|
await self.send_message("Thinking about how to implement this task ...")
|
|
|
|
current_task_index = self.current_state.tasks.index(task)
|
|
|
|
llm = self.get_llm()
|
|
convo = AgentConvo(self).template(
|
|
"breakdown",
|
|
task=task,
|
|
iteration=None,
|
|
current_task_index=current_task_index,
|
|
)
|
|
response: str = await llm(convo)
|
|
|
|
# FIXME: check if this is correct, as sqlalchemy can't figure out modifications
|
|
# to attributes; however, self.next is not saved yet so maybe this is fine
|
|
self.next_state.tasks[current_task_index] = {
|
|
**task,
|
|
"instructions": response,
|
|
}
|
|
|
|
await self.send_message("Breaking down the task into steps ...")
|
|
convo.template("parse_task").require_schema(TaskSteps)
|
|
response: TaskSteps = await llm(convo, parser=JSONParser(TaskSteps), temperature=0)
|
|
|
|
# There might be state leftovers from previous tasks that we need to clean here
|
|
self.next_state.modified_files = {}
|
|
self.set_next_steps(response, source)
|
|
return AgentResponse.done(self)
|
|
|
|
async def get_relevant_files(self) -> AgentResponse:
|
|
log.debug("Getting relevant files for the current task")
|
|
await self.send_message("Figuring out which project files are relevant for the next task ...")
|
|
|
|
llm = self.get_llm()
|
|
convo = AgentConvo(self).template("filter_files", current_task=self.current_state.current_task)
|
|
|
|
# FIXME: this doesn't validate correct structure format, we should use pydantic for that as well
|
|
llm_response: list[str] = await llm(convo, parser=JSONParser(), temperature=0)
|
|
|
|
existing_files = {file.path for file in self.current_state.files}
|
|
self.next_state.relevant_files = [path for path in llm_response if path in existing_files]
|
|
|
|
return AgentResponse.done(self)
|
|
|
|
def set_next_steps(self, response: TaskSteps, source: str):
|
|
# For logging/debugging purposes, we don't want to remove the finished steps
|
|
# until we're done with the task.
|
|
finished_steps = [step for step in self.current_state.steps if step["completed"]]
|
|
self.next_state.steps = finished_steps + [
|
|
{
|
|
"id": uuid4().hex,
|
|
"completed": False,
|
|
"source": source,
|
|
**step.model_dump(),
|
|
}
|
|
for step in response.steps
|
|
]
|
|
if len(self.next_state.unfinished_steps) > 0:
|
|
self.next_state.steps += [
|
|
# TODO: add refactor step here once we have the refactor agent
|
|
{
|
|
"id": uuid4().hex,
|
|
"completed": False,
|
|
"type": "review_task",
|
|
"source": source,
|
|
},
|
|
{
|
|
"id": uuid4().hex,
|
|
"completed": False,
|
|
"type": "create_readme",
|
|
"source": source,
|
|
},
|
|
]
|
|
log.debug(f"Next steps: {self.next_state.unfinished_steps}")
|
|
|
|
async def ask_to_execute_task(self) -> bool:
|
|
"""
|
|
Asks the user to approve, skip or edit the current task.
|
|
|
|
If task is edited, the method returns False so that the changes are saved. The
|
|
Orchestrator will rerun the agent on the next iteration.
|
|
|
|
:return: True if the task should be executed as is, False if the task is skipped or edited
|
|
"""
|
|
description = self.current_state.current_task["description"]
|
|
user_response = await self.ask_question(
|
|
"Do you want to execute the this task:\n\n" + description,
|
|
buttons={"yes": "Yes", "edit": "Edit Task", "skip": "Skip Task"},
|
|
default="yes",
|
|
buttons_only=True,
|
|
)
|
|
if user_response.button == "yes":
|
|
# Execute the task as is
|
|
return True
|
|
|
|
if user_response.cancelled or user_response.button == "skip":
|
|
log.info(f"Skipping task: {description}")
|
|
self.next_state.current_task["instructions"] = "(skipped on user request)"
|
|
self.next_state.complete_task()
|
|
await self.send_message(f"Skipping task {description}")
|
|
# We're done here, and will pick up the next task (if any) on the next run
|
|
return False
|
|
|
|
user_response = await self.ask_question(
|
|
"Edit the task description:",
|
|
buttons={
|
|
# FIXME: Continue doesn't actually work, VSCode doesn't send the user
|
|
# message if it's clicked. Long term we need to fix the extension.
|
|
# "continue": "Continue",
|
|
"cancel": "Cancel",
|
|
},
|
|
default="continue",
|
|
initial_text=description,
|
|
)
|
|
if user_response.button == "cancel" or user_response.cancelled:
|
|
# User hasn't edited the task so we can execute it immediately as is
|
|
return True
|
|
|
|
self.next_state.current_task["description"] = user_response.text
|
|
self.next_state.current_task["run_always"] = True
|
|
self.next_state.relevant_files = []
|
|
log.info(f"Task description updated to: {user_response.text}")
|
|
# Orchestrator will rerun us with the new task description
|
|
return False
|