Spaces:
Runtime error
Runtime error
File size: 11,599 Bytes
27d7338 f8cc947 27d7338 c43ae5c f8cc947 27d7338 c43ae5c f8cc947 27d7338 f8cc947 27d7338 c43ae5c f8cc947 c43ae5c f8cc947 27d7338 f8cc947 27d7338 f8cc947 27d7338 351c200 c43ae5c f8cc947 351c200 f8cc947 27d7338 f8cc947 27d7338 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 | """Core environment implementation for Code Review OpenEnv."""
from __future__ import annotations
from typing import Dict, List, Optional, Tuple
from env.models import CodeReviewAction, CodeReviewObservation, ReviewComment
from env.reward_engine import RewardEngine
from env.state_manager import StateManager
from env.tasks.task_easy import get_task as get_easy
from env.tasks.task_hard import get_task as get_hard
from env.tasks.task_medium import get_task as get_medium
class CodeReviewEnv:
"""Gym-like environment for evaluating code-review agents."""
def __init__(self) -> None:
"""Initialize environment with no active episode."""
self._task_id: str | None = None
self._max_steps: int = 0
self._pr_title: str = ""
self._pr_description: str = ""
self._full_file: str = ""
self._code_diff: str = ""
self._ground_truth = []
self._state: StateManager | None = None
self._reward_engine: RewardEngine | None = None
# Multi-file repository support
self._repository_files: Optional[Dict[str, str]] = None
self._available_files: Optional[List[str]] = None
def reset(self, task_id: str) -> CodeReviewObservation:
"""Reset the environment to a fresh episode for the given task.
Args:
task_id: One of "easy", "medium", "hard".
Returns:
Initial observation with empty existing_comments.
"""
if task_id == "easy":
task = get_easy()
elif task_id == "medium":
task = get_medium()
elif task_id == "hard":
task = get_hard()
else:
raise ValueError(f"Unknown task_id: {task_id}")
self._task_id = task.task_id
self._max_steps = task.max_steps
self._pr_title = task.pr_title
self._pr_description = task.pr_description
self._full_file = task.full_file
self._code_diff = task.code_diff
self._ground_truth = task.ground_truth
# Store repository files if available
self._repository_files = getattr(task, 'repository_files', None)
self._available_files = getattr(task, 'available_files', None)
self._state = StateManager(task_id=task.task_id)
self._reward_engine = RewardEngine(task_id=task.task_id, ground_truth=task.ground_truth, max_steps=task.max_steps)
return CodeReviewObservation(
task_id=task.task_id,
language="python",
pr_title=self._pr_title,
pr_description=self._pr_description,
code_diff=self._code_diff,
full_file=self._full_file,
existing_comments=[],
step_number=1,
max_steps=self._max_steps,
review_status="pending",
repository_files=self._repository_files,
available_files=self._available_files,
)
def step(self, action: CodeReviewAction) -> Tuple[CodeReviewObservation, float, bool, dict]:
"""Apply an action and advance the environment by one step.
Args:
action: CodeReviewAction describing the agent's operation.
Returns:
Tuple of (updated_observation, reward, done, info).
"""
if self._state is None or self._reward_engine is None or self._task_id is None:
raise RuntimeError("Environment must be reset() before step().")
error: str | None = None
reward: float
new_comment: ReviewComment | None = None
# Handle inspect_file action
if action.operation == "inspect_file":
if self._repository_files and action.filename and action.filename in self._repository_files:
outcome = self._reward_engine.compute(
action,
comments_so_far=self._state.comments,
correctly_identified_bug_lines=self._state.correctly_identified_bug_lines,
step_number=self._state.step_number,
steps_used_after_this=self._state.step_number,
)
reward = outcome.reward
self._state.record_action(action, reward, error=None)
else:
reward = 0.0
error = f"File not found: {action.filename}"
self._state.record_action(action, reward, error=error)
# Handle inspect_lines action
elif action.operation == "inspect_lines":
if action.start_line is not None and action.end_line is not None:
if action.end_line - action.start_line > 40:
reward = 0.0
error = "inspect_lines max range is 40 lines"
self._state.record_action(action, reward, error=error)
elif self._repository_files and action.filename and action.filename in self._repository_files:
outcome = self._reward_engine.compute(
action,
comments_so_far=self._state.comments,
correctly_identified_bug_lines=self._state.correctly_identified_bug_lines,
step_number=self._state.step_number,
steps_used_after_this=self._state.step_number,
)
reward = outcome.reward
self._state.record_action(action, reward, error=None)
else:
reward = 0.0
error = f"File not found: {action.filename}"
self._state.record_action(action, reward, error=error)
else:
reward = 0.0
error = "inspect_lines requires start_line and end_line"
self._state.record_action(action, reward, error=error)
elif action.operation == "add_comment":
if action.line_number is None:
outcome = self._reward_engine.compute(
action,
comments_so_far=self._state.comments,
correctly_identified_bug_lines=self._state.correctly_identified_bug_lines,
step_number=self._state.step_number,
steps_used_after_this=self._state.step_number,
)
reward = outcome.reward
error = "Missing line_number for add_comment"
self._state.record_action(
action,
reward,
new_comment=None,
correctly_identified_bug_line=None,
is_false_positive=True,
is_red_herring_flag=False,
error=error,
confidence_modifier=outcome.confidence_modifier,
)
else:
new_comment = ReviewComment(
line_number=action.line_number,
severity=action.severity or "minor",
category=action.category or "bug",
message=action.message or "Issue detected",
step_added=self._state.step_number,
)
outcome = self._reward_engine.compute(
action,
comments_so_far=self._state.comments + [new_comment],
correctly_identified_bug_lines=self._state.correctly_identified_bug_lines,
step_number=self._state.step_number,
steps_used_after_this=self._state.step_number,
)
reward = outcome.reward
self._state.record_action(
action,
reward,
new_comment=new_comment,
correctly_identified_bug_line=outcome.correctly_identified_bug_line,
is_false_positive=outcome.is_false_positive,
is_red_herring_flag=outcome.is_red_herring_flag,
error=None,
confidence_modifier=outcome.confidence_modifier,
explanation_depth=outcome.explanation_depth,
)
else:
outcome = self._reward_engine.compute(
action,
comments_so_far=self._state.comments,
correctly_identified_bug_lines=self._state.correctly_identified_bug_lines,
step_number=self._state.step_number,
steps_used_after_this=self._state.step_number,
)
reward = outcome.reward
self._state.record_action(action, reward, error=None)
done = False
if action.operation in {"done", "approve", "request_changes"}:
done = True
if self._state.step_number > self._max_steps:
done = True
if action.operation != "done":
self._state.cumulative_reward += -0.20
# On terminal actions, replace cumulative_reward with the final F1
# grader score. The per-step shaped rewards guided the agent during
# the episode, but the final reported score must be the deterministic
# F1 result — NOT the sum of shaped rewards + F1.
if done and action.operation in {"done", "approve", "request_changes"}:
final_f1 = outcome.final_score if hasattr(outcome, "final_score") and outcome.final_score is not None else None
if final_f1 is not None:
# Replace, don't add — this is the fix for the 0.999 ceiling bug.
self._state.cumulative_reward = final_f1
else:
# approve/request_changes or done without final_score — run grader.
final_f1 = self._reward_engine._grade(self._state.comments)
self._state.cumulative_reward = final_f1
# Compute injection resistance at episode end for hard task
if done and self._task_id == "hard":
# The injected lines are the real bug lines that have adversarial comments above them
# ECB bug (line 35) and race condition bug (line 47)
injected_lines = [35, 47]
self._state.compute_injection_resistance(self._ground_truth, injected_lines)
# Clamp cumulative score to (0.0, 1.0) per OpenEnv strictly between bounds spec.
clamped_score = max(0.001, min(0.999, self._state.cumulative_reward))
info = {
"bugs_found": len(self._state.correctly_identified_bug_lines),
"false_positives": self._state.get_false_positive_count(),
"current_score": clamped_score,
"error": error,
}
obs = CodeReviewObservation(
task_id=self._task_id,
language="python",
pr_title=self._pr_title,
pr_description=self._pr_description,
code_diff=self._code_diff,
full_file=self._full_file,
existing_comments=list(self._state.comments),
step_number=max(1, self._state.step_number),
max_steps=self._max_steps,
review_status="submitted" if done else "in_review",
repository_files=self._repository_files,
available_files=self._available_files,
)
return obs, float(round(min(max(reward, 0.01), 0.99), 3)), bool(done), info
def state(self) -> dict:
"""Return full current state as a plain dict."""
if self._state is None:
return {"task_id": None, "step_number": 0, "comments": [], "running_score": 0.01, "bugs_found": 0, "false_positives": 0}
return self._state.to_dict()
|