zzz / openhands /resolver /resolve_issue.py
ar08's picture
Upload 1040 files
246d201 verified
# flake8: noqa: E501
import asyncio
import dataclasses
import json
import os
import pathlib
import shutil
import subprocess
from typing import Any
from uuid import uuid4
from termcolor import colored
import openhands
from openhands.controller.state.state import State
from openhands.core.config import AgentConfig, AppConfig, LLMConfig, SandboxConfig
from openhands.core.logger import openhands_logger as logger
from openhands.core.main import create_runtime, run_controller
from openhands.events.action import CmdRunAction, MessageAction
from openhands.events.observation import (
CmdOutputObservation,
ErrorObservation,
Observation,
)
from openhands.events.stream import EventStreamSubscriber
from openhands.resolver.github_issue import GithubIssue
from openhands.resolver.issue_definitions import (
IssueHandler,
IssueHandlerInterface,
PRHandler,
)
from openhands.resolver.resolver_output import ResolverOutput
from openhands.resolver.utils import (
codeact_user_response,
reset_logger_for_multiprocessing,
)
from openhands.runtime.base import Runtime
# Don't make this confgurable for now, unless we have other competitive agents
AGENT_CLASS = 'CodeActAgent'
def initialize_runtime(
runtime: Runtime,
):
"""Initialize the runtime for the agent.
This function is called before the runtime is used to run the agent.
Currently it does nothing.
"""
logger.info('-' * 30)
logger.info('BEGIN Runtime Completion Fn')
logger.info('-' * 30)
obs: Observation
action = CmdRunAction(command='cd /workspace')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(f'Failed to change directory to /workspace.\n{obs}')
action = CmdRunAction(command='git config --global core.pager ""')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(f'Failed to set git config.\n{obs}')
async def complete_runtime(
runtime: Runtime,
base_commit: str,
) -> dict[str, Any]:
"""Complete the runtime for the agent.
This function is called before the runtime is used to run the agent.
If you need to do something in the sandbox to get the correctness metric after
the agent has run, modify this function.
"""
logger.info('-' * 30)
logger.info('BEGIN Runtime Completion Fn')
logger.info('-' * 30)
obs: Observation
action = CmdRunAction(command='cd /workspace')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(
f'Failed to change directory to /workspace. Observation: {obs}'
)
action = CmdRunAction(command='git config --global core.pager ""')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(f'Failed to set git config. Observation: {obs}')
action = CmdRunAction(command='git config --global --add safe.directory /workspace')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(f'Failed to set git config. Observation: {obs}')
action = CmdRunAction(command='git add -A')
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
if not isinstance(obs, CmdOutputObservation) or obs.exit_code != 0:
raise RuntimeError(f'Failed to git add. Observation: {obs}')
n_retries = 0
git_patch = None
while n_retries < 5:
action = CmdRunAction(command=f'git diff --no-color --cached {base_commit}')
action.set_hard_timeout(600 + 100 * n_retries)
logger.info(action, extra={'msg_type': 'ACTION'})
obs = runtime.run_action(action)
logger.info(obs, extra={'msg_type': 'OBSERVATION'})
n_retries += 1
if isinstance(obs, CmdOutputObservation):
if obs.exit_code == 0:
git_patch = obs.content.strip()
break
else:
logger.info('Failed to get git diff, retrying...')
await asyncio.sleep(10)
elif isinstance(obs, ErrorObservation):
logger.error(f'Error occurred: {obs.content}. Retrying...')
await asyncio.sleep(10)
else:
raise ValueError(f'Unexpected observation type: {type(obs)}')
logger.info('-' * 30)
logger.info('END Runtime Completion Fn')
logger.info('-' * 30)
return {'git_patch': git_patch}
async def process_issue(
issue: GithubIssue,
base_commit: str,
max_iterations: int,
llm_config: LLMConfig,
output_dir: str,
runtime_container_image: str | None,
prompt_template: str,
issue_handler: IssueHandlerInterface,
repo_instruction: str | None = None,
reset_logger: bool = False,
) -> ResolverOutput:
# Setup the logger properly, so you can run multi-processing to parallelize processing
if reset_logger:
log_dir = os.path.join(output_dir, 'infer_logs')
reset_logger_for_multiprocessing(logger, str(issue.number), log_dir)
else:
logger.info(f'Starting fixing issue {issue.number}.')
workspace_base = os.path.join(
output_dir, 'workspace', f'{issue_handler.issue_type}_{issue.number}'
)
# Get the absolute path of the workspace base
workspace_base = os.path.abspath(workspace_base)
# write the repo to the workspace
if os.path.exists(workspace_base):
shutil.rmtree(workspace_base)
shutil.copytree(os.path.join(output_dir, 'repo'), workspace_base)
config = AppConfig(
default_agent='CodeActAgent',
runtime='docker',
max_budget_per_task=4,
max_iterations=max_iterations,
sandbox=SandboxConfig(
runtime_container_image=runtime_container_image,
enable_auto_lint=False,
use_host_network=False,
# large enough timeout, since some testcases take very long to run
timeout=300,
),
# do not mount workspace
workspace_base=workspace_base,
workspace_mount_path=workspace_base,
agents={'CodeActAgent': AgentConfig(disabled_microagents=['github'])},
)
config.set_llm_config(llm_config)
runtime = create_runtime(config)
await runtime.connect()
def on_event(evt):
logger.info(evt)
runtime.event_stream.subscribe(EventStreamSubscriber.MAIN, on_event, str(uuid4()))
initialize_runtime(runtime)
instruction, images_urls = issue_handler.get_instruction(
issue, prompt_template, repo_instruction
)
# Here's how you can run the agent (similar to the `main` function) and get the final task state
action = MessageAction(content=instruction, image_urls=images_urls)
try:
state: State | None = await run_controller(
config=config,
initial_user_action=action,
runtime=runtime,
fake_user_response_fn=codeact_user_response,
)
if state is None:
raise RuntimeError('Failed to run the agent.')
except (ValueError, RuntimeError) as e:
error_msg = f'Agent failed with error: {str(e)}'
logger.error(error_msg)
state = None
last_error: str | None = error_msg
# Get git patch
return_val = await complete_runtime(runtime, base_commit)
git_patch = return_val['git_patch']
logger.info(
f'Got git diff for instance {issue.number}:\n--------\n{git_patch}\n--------'
)
# Serialize histories and set defaults for failed state
if state is None:
histories = []
metrics = None
success = False
comment_success = None
result_explanation = 'Agent failed to run'
last_error = 'Agent failed to run or crashed'
else:
histories = [dataclasses.asdict(event) for event in state.history]
metrics = state.metrics.get() if state.metrics else None
# determine success based on the history, issue description and git patch
success, comment_success, result_explanation = issue_handler.guess_success(
issue, state.history, git_patch
)
if issue_handler.issue_type == 'pr' and comment_success:
success_log = 'I have updated the PR and resolved some of the issues that were cited in the pull request review. Specifically, I identified the following revision requests, and all the ones that I think I successfully resolved are checked off. All the unchecked ones I was not able to resolve, so manual intervention may be required:\n'
try:
explanations = json.loads(result_explanation)
except json.JSONDecodeError:
logger.error(
f'Failed to parse result_explanation as JSON: {result_explanation}'
)
explanations = [str(result_explanation)] # Use raw string as fallback
for success_indicator, explanation in zip(comment_success, explanations):
status = (
colored('[X]', 'red')
if success_indicator
else colored('[ ]', 'red')
)
bullet_point = colored('-', 'yellow')
success_log += f'\n{bullet_point} {status}: {explanation}'
logger.info(success_log)
last_error = state.last_error if state.last_error else None
# Save the output
output = ResolverOutput(
issue=issue,
issue_type=issue_handler.issue_type,
instruction=instruction,
base_commit=base_commit,
git_patch=git_patch,
history=histories,
metrics=metrics,
success=success,
comment_success=comment_success,
result_explanation=result_explanation,
error=last_error,
)
return output
def issue_handler_factory(
issue_type: str, owner: str, repo: str, token: str, llm_config: LLMConfig
) -> IssueHandlerInterface:
if issue_type == 'issue':
return IssueHandler(owner, repo, token, llm_config)
elif issue_type == 'pr':
return PRHandler(owner, repo, token, llm_config)
else:
raise ValueError(f'Invalid issue type: {issue_type}')
async def resolve_issue(
owner: str,
repo: str,
token: str,
username: str,
max_iterations: int,
output_dir: str,
llm_config: LLMConfig,
runtime_container_image: str | None,
prompt_template: str,
issue_type: str,
repo_instruction: str | None,
issue_number: int,
comment_id: int | None,
target_branch: str | None = None,
reset_logger: bool = False,
) -> None:
"""Resolve a single github issue.
Args:
owner: Github owner of the repo.
repo: Github repository to resolve issues in form of `owner/repo`.
token: Github token to access the repository.
username: Github username to access the repository.
max_iterations: Maximum number of iterations to run.
output_dir: Output directory to write the results.
llm_config: Configuration for the language model.
runtime_container_image: Container image to use.
prompt_template: Prompt template to use.
issue_type: Type of issue to resolve (issue or pr).
repo_instruction: Repository instruction to use.
issue_number: Issue number to resolve.
comment_id: Optional ID of a specific comment to focus on.
target_branch: Optional target branch to create PR against (for PRs).
reset_logger: Whether to reset the logger for multiprocessing.
"""
issue_handler = issue_handler_factory(issue_type, owner, repo, token, llm_config)
# Load dataset
issues: list[GithubIssue] = issue_handler.get_converted_issues(
issue_numbers=[issue_number], comment_id=comment_id
)
if not issues:
raise ValueError(
f'No issues found for issue number {issue_number}. Please verify that:\n'
f'1. The issue/PR #{issue_number} exists in the repository {owner}/{repo}\n'
f'2. You have the correct permissions to access it\n'
f'3. The repository name is spelled correctly'
)
issue = issues[0]
if comment_id is not None:
if (
issue_type == 'pr'
and not issue.review_comments
and not issue.review_threads
and not issue.thread_comments
):
raise ValueError(
f'Comment ID {comment_id} did not have a match for issue {issue.number}'
)
if issue_type == 'issue' and not issue.thread_comments:
raise ValueError(
f'Comment ID {comment_id} did not have a match for issue {issue.number}'
)
# TEST METADATA
model_name = llm_config.model.split('/')[-1]
pathlib.Path(output_dir).mkdir(parents=True, exist_ok=True)
pathlib.Path(os.path.join(output_dir, 'infer_logs')).mkdir(
parents=True, exist_ok=True
)
logger.info(f'Using output directory: {output_dir}')
# checkout the repo
repo_dir = os.path.join(output_dir, 'repo')
if not os.path.exists(repo_dir):
checkout_output = subprocess.check_output(
[
'git',
'clone',
f'https://{username}:{token}@github.com/{owner}/{repo}',
f'{output_dir}/repo',
]
).decode('utf-8')
if 'fatal' in checkout_output:
raise RuntimeError(f'Failed to clone repository: {checkout_output}')
# get the commit id of current repo for reproducibility
base_commit = (
subprocess.check_output(['git', 'rev-parse', 'HEAD'], cwd=repo_dir)
.decode('utf-8')
.strip()
)
logger.info(f'Base commit: {base_commit}')
if repo_instruction is None:
# Check for .openhands_instructions file in the workspace directory
openhands_instructions_path = os.path.join(repo_dir, '.openhands_instructions')
if os.path.exists(openhands_instructions_path):
with open(openhands_instructions_path, 'r') as f:
repo_instruction = f.read()
# OUTPUT FILE
output_file = os.path.join(output_dir, 'output.jsonl')
logger.info(f'Writing output to {output_file}')
# Check if this issue was already processed
if os.path.exists(output_file):
with open(output_file, 'r') as f:
for line in f:
data = ResolverOutput.model_validate_json(line)
if data.issue.number == issue_number:
logger.warning(
f'Issue {issue_number} was already processed. Skipping.'
)
return
output_fp = open(output_file, 'a')
logger.info(
f'Resolving issue {issue_number} with Agent {AGENT_CLASS}, model {model_name}, max iterations {max_iterations}.'
)
try:
# checkout to pr branch if needed
if issue_type == 'pr':
branch_to_use = target_branch if target_branch else issue.head_branch
logger.info(
f'Checking out to PR branch {target_branch} for issue {issue.number}'
)
if not branch_to_use:
raise ValueError('Branch name cannot be None')
# Fetch the branch first to ensure it exists locally
fetch_cmd = ['git', 'fetch', 'origin', branch_to_use]
subprocess.check_output(
fetch_cmd,
cwd=repo_dir,
)
# Checkout the branch
checkout_cmd = ['git', 'checkout', branch_to_use]
subprocess.check_output(
checkout_cmd,
cwd=repo_dir,
)
# Update issue's base_branch if using custom target branch
if target_branch:
issue.base_branch = target_branch
base_commit = (
subprocess.check_output(['git', 'rev-parse', 'HEAD'], cwd=repo_dir)
.decode('utf-8')
.strip()
)
output = await process_issue(
issue,
base_commit,
max_iterations,
llm_config,
output_dir,
runtime_container_image,
prompt_template,
issue_handler,
repo_instruction,
reset_logger,
)
output_fp.write(output.model_dump_json() + '\n')
output_fp.flush()
finally:
output_fp.close()
logger.info('Finished.')
def main():
import argparse
def int_or_none(value):
if value.lower() == 'none':
return None
else:
return int(value)
parser = argparse.ArgumentParser(description='Resolve a single issue from Github.')
parser.add_argument(
'--repo',
type=str,
required=True,
help='Github repository to resolve issues in form of `owner/repo`.',
)
parser.add_argument(
'--token',
type=str,
default=None,
help='Github token to access the repository.',
)
parser.add_argument(
'--username',
type=str,
default=None,
help='Github username to access the repository.',
)
parser.add_argument(
'--runtime-container-image',
type=str,
default=None,
help='Container image to use.',
)
parser.add_argument(
'--max-iterations',
type=int,
default=50,
help='Maximum number of iterations to run.',
)
parser.add_argument(
'--issue-number',
type=int,
required=True,
help='Issue number to resolve.',
)
parser.add_argument(
'--comment-id',
type=int_or_none,
required=False,
default=None,
help='Resolve a specific comment',
)
parser.add_argument(
'--output-dir',
type=str,
default='output',
help='Output directory to write the results.',
)
parser.add_argument(
'--llm-model',
type=str,
default=None,
help='LLM model to use.',
)
parser.add_argument(
'--llm-api-key',
type=str,
default=None,
help='LLM API key to use.',
)
parser.add_argument(
'--llm-base-url',
type=str,
default=None,
help='LLM base URL to use.',
)
parser.add_argument(
'--prompt-file',
type=str,
default=None,
help='Path to the prompt template file in Jinja format.',
)
parser.add_argument(
'--repo-instruction-file',
type=str,
default=None,
help='Path to the repository instruction file in text format.',
)
parser.add_argument(
'--issue-type',
type=str,
default='issue',
choices=['issue', 'pr'],
help='Type of issue to resolve, either open issue or pr comments.',
)
parser.add_argument(
'--target-branch',
type=str,
default=None,
help="Target branch to pull and create PR against (for PRs). If not specified, uses the PR's base branch.",
)
parser.add_argument(
'--is-experimental',
type=lambda x: x.lower() == 'true',
help='Whether to run in experimental mode.',
)
my_args = parser.parse_args()
runtime_container_image = my_args.runtime_container_image
if runtime_container_image is None and not my_args.is_experimental:
runtime_container_image = (
f'ghcr.io/all-hands-ai/runtime:{openhands.__version__}-nikolaik'
)
owner, repo = my_args.repo.split('/')
token = my_args.token if my_args.token else os.getenv('GITHUB_TOKEN')
username = my_args.username if my_args.username else os.getenv('GITHUB_USERNAME')
if not username:
raise ValueError('Github username is required.')
if not token:
raise ValueError('Github token is required.')
llm_config = LLMConfig(
model=my_args.llm_model or os.environ['LLM_MODEL'],
api_key=my_args.llm_api_key or os.environ['LLM_API_KEY'],
base_url=my_args.llm_base_url or os.environ.get('LLM_BASE_URL', None),
)
repo_instruction = None
if my_args.repo_instruction_file:
with open(my_args.repo_instruction_file, 'r') as f:
repo_instruction = f.read()
issue_type = my_args.issue_type
# Read the prompt template
prompt_file = my_args.prompt_file
if prompt_file is None:
if issue_type == 'issue':
prompt_file = os.path.join(
os.path.dirname(__file__), 'prompts/resolve/basic-with-tests.jinja'
)
else:
prompt_file = os.path.join(
os.path.dirname(__file__), 'prompts/resolve/basic-followup.jinja'
)
with open(prompt_file, 'r') as f:
prompt_template = f.read()
asyncio.run(
resolve_issue(
owner=owner,
repo=repo,
token=token,
username=username,
runtime_container_image=runtime_container_image,
max_iterations=my_args.max_iterations,
output_dir=my_args.output_dir,
llm_config=llm_config,
prompt_template=prompt_template,
issue_type=issue_type,
repo_instruction=repo_instruction,
issue_number=my_args.issue_number,
comment_id=my_args.comment_id,
target_branch=my_args.target_branch,
)
)
if __name__ == '__main__':
main()