PR #8
#8
by
dzakaathif
- opened
- rewards/__init__.py +31 -0
- rewards/game_2048.py +168 -0
- rewards/linear.py +255 -0
- rewards/linkedin.py +88 -0
- rewards/minesweeper.py +255 -0
- rewards/salesforce.py +423 -0
- rewards/tic_tac_toe.py +82 -0
rewards/__init__.py
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Modular reward functions organized by game/application.
|
| 3 |
+
|
| 4 |
+
This package contains reward validation functions for different dojos,
|
| 5 |
+
organized into separate modules for maintainability.
|
| 6 |
+
"""
|
| 7 |
+
|
| 8 |
+
from dojo_sdk_core.dojos.rewards.game_2048 import REWARD_FUNCTIONS_2048
|
| 9 |
+
from dojo_sdk_core.dojos.rewards.linear import REWARD_FUNCTIONS_LINEAR
|
| 10 |
+
from dojo_sdk_core.dojos.rewards.linkedin import REWARD_FUNCTIONS_LINKEDIN
|
| 11 |
+
from dojo_sdk_core.dojos.rewards.minesweeper import REWARD_FUNCTIONS_MINESWEEPER
|
| 12 |
+
from dojo_sdk_core.dojos.rewards.salesforce import REWARD_FUNCTIONS_SALESFORCE
|
| 13 |
+
from dojo_sdk_core.dojos.rewards.tic_tac_toe import REWARD_FUNCTIONS_TIC_TAC_TOE
|
| 14 |
+
|
| 15 |
+
# Unified registry of all reward functions
|
| 16 |
+
REWARD_FUNCTIONS = {
|
| 17 |
+
**REWARD_FUNCTIONS_2048,
|
| 18 |
+
**REWARD_FUNCTIONS_LINEAR,
|
| 19 |
+
**REWARD_FUNCTIONS_LINKEDIN,
|
| 20 |
+
**REWARD_FUNCTIONS_SALESFORCE,
|
| 21 |
+
**REWARD_FUNCTIONS_TIC_TAC_TOE,
|
| 22 |
+
**REWARD_FUNCTIONS_MINESWEEPER,
|
| 23 |
+
}
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
def get_reward_function(name: str):
|
| 27 |
+
"""Get a reward function by name from the unified registry."""
|
| 28 |
+
return REWARD_FUNCTIONS.get(name)
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
__all__ = ["REWARD_FUNCTIONS", "get_reward_function"]
|
rewards/game_2048.py
ADDED
|
@@ -0,0 +1,168 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Reward functions for 2048 game tasks.
|
| 3 |
+
|
| 4 |
+
All functions validate the final board state, accounting for randomly spawned tiles.
|
| 5 |
+
Board is a 16-element array representing a 4x4 grid:
|
| 6 |
+
[0, 1, 2, 3,
|
| 7 |
+
4, 5, 6, 7,
|
| 8 |
+
8, 9, 10, 11,
|
| 9 |
+
12, 13, 14, 15]
|
| 10 |
+
"""
|
| 11 |
+
|
| 12 |
+
import logging
|
| 13 |
+
from typing import Any, Dict, Tuple
|
| 14 |
+
|
| 15 |
+
logger = logging.getLogger(__name__)
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
def _validate_get_2048(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 19 |
+
"""Validate that a 2048 tile is present on the board."""
|
| 20 |
+
if "board" not in final_state:
|
| 21 |
+
return 0.0, "No board in final state"
|
| 22 |
+
|
| 23 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 24 |
+
if 2048 in final_state["board"]:
|
| 25 |
+
return 1.0, "A 2048 tile is present."
|
| 26 |
+
return 0.0, "No 2048 tile is present."
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def _validate_2048_make_first_move(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 30 |
+
"""Validate that any valid move was made from the initial board state."""
|
| 31 |
+
if "board" not in final_state:
|
| 32 |
+
return 0.0, "No board in final state"
|
| 33 |
+
|
| 34 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 35 |
+
|
| 36 |
+
# Check if board changed (has non-zero tiles)
|
| 37 |
+
if any(tile != 0 for tile in final_state["board"]):
|
| 38 |
+
return 1.0, "Board changed from initial state - move was made successfully."
|
| 39 |
+
|
| 40 |
+
return 0.0, "Board is still empty - no move was made."
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
def _validate_move_tiles_right(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 44 |
+
"""Validate that all tiles are in the rightmost column (positions 3, 7, 11, 15)."""
|
| 45 |
+
if "board" not in final_state:
|
| 46 |
+
return 0.0, "No board in final state"
|
| 47 |
+
|
| 48 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 49 |
+
|
| 50 |
+
board = final_state["board"]
|
| 51 |
+
rightmost_positions = [3, 7, 11, 15]
|
| 52 |
+
|
| 53 |
+
# Get all non-zero tiles and their positions
|
| 54 |
+
non_zero_positions = [i for i, tile in enumerate(board) if tile != 0]
|
| 55 |
+
|
| 56 |
+
# Check if all non-zero tiles are in rightmost column
|
| 57 |
+
if all(pos in rightmost_positions for pos in non_zero_positions):
|
| 58 |
+
return 1.0, f"All tiles successfully moved to rightmost column at positions {non_zero_positions}."
|
| 59 |
+
|
| 60 |
+
# Find which tiles are not in the rightmost column
|
| 61 |
+
misplaced = [pos for pos in non_zero_positions if pos not in rightmost_positions]
|
| 62 |
+
return 0.0, f"Some tiles are not in rightmost column. Misplaced at positions: {misplaced}"
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
def _validate_get_32_tile(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 66 |
+
"""Validate that a 32 tile is present on the board."""
|
| 67 |
+
if "board" not in final_state:
|
| 68 |
+
return 0.0, "No board in final state"
|
| 69 |
+
|
| 70 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 71 |
+
|
| 72 |
+
if 32 in final_state["board"]:
|
| 73 |
+
return 1.0, "A 32 tile is present."
|
| 74 |
+
return 0.0, "No 32 tile is present."
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
def _validate_get_128_tile(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 78 |
+
"""Validate that a 128 tile is present on the board."""
|
| 79 |
+
if "board" not in final_state:
|
| 80 |
+
return 0.0, "No board in final state"
|
| 81 |
+
|
| 82 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 83 |
+
|
| 84 |
+
if 128 in final_state["board"]:
|
| 85 |
+
return 1.0, "A 128 tile is present."
|
| 86 |
+
return 0.0, "No 128 tile is present."
|
| 87 |
+
|
| 88 |
+
|
| 89 |
+
def _validate_get_256_tile(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 90 |
+
"""Validate that a 256 tile is present on the board."""
|
| 91 |
+
if "board" not in final_state:
|
| 92 |
+
return 0.0, "No board in final state"
|
| 93 |
+
|
| 94 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 95 |
+
|
| 96 |
+
if 256 in final_state["board"]:
|
| 97 |
+
return 1.0, "A 256 tile is present."
|
| 98 |
+
return 0.0, "No 256 tile is present."
|
| 99 |
+
|
| 100 |
+
|
| 101 |
+
def _validate_get_512_tile(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 102 |
+
"""Validate that a 512 tile is present on the board."""
|
| 103 |
+
if "board" not in final_state:
|
| 104 |
+
return 0.0, "No board in final state"
|
| 105 |
+
|
| 106 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 107 |
+
|
| 108 |
+
if 512 in final_state["board"]:
|
| 109 |
+
return 1.0, "A 512 tile is present."
|
| 110 |
+
return 0.0, "No 512 tile is present."
|
| 111 |
+
|
| 112 |
+
|
| 113 |
+
def _validate_create_two_high_tiles(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 114 |
+
"""Validate that at least two tiles with value 128 or higher are present."""
|
| 115 |
+
if "board" not in final_state:
|
| 116 |
+
return 0.0, "No board in final state"
|
| 117 |
+
|
| 118 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 119 |
+
|
| 120 |
+
high_tiles = [tile for tile in final_state["board"] if tile >= 128]
|
| 121 |
+
count = len(high_tiles)
|
| 122 |
+
|
| 123 |
+
if count >= 2:
|
| 124 |
+
return 1.0, f"Board contains {count} tiles with value 128 or higher: {high_tiles}."
|
| 125 |
+
|
| 126 |
+
return 0.0, f"Board contains only {count} tile(s) with value 128 or higher. Need at least 2."
|
| 127 |
+
|
| 128 |
+
|
| 129 |
+
def _validate_reach_540_sum(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 130 |
+
"""Validate that the sum of all tile values is 540 or greater."""
|
| 131 |
+
if "board" not in final_state:
|
| 132 |
+
return 0.0, "No board in final state"
|
| 133 |
+
|
| 134 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 135 |
+
|
| 136 |
+
total_sum = sum(final_state["board"])
|
| 137 |
+
|
| 138 |
+
if total_sum >= 540:
|
| 139 |
+
return 1.0, f"Board sum is {total_sum}, which meets the requirement of 540 or greater."
|
| 140 |
+
|
| 141 |
+
return 0.0, f"Board sum is {total_sum}, which is below the required 540."
|
| 142 |
+
|
| 143 |
+
|
| 144 |
+
def _validate_strategic_32(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 145 |
+
"""Validate that a 32 tile was created strategically from an empty board."""
|
| 146 |
+
if "board" not in final_state:
|
| 147 |
+
return 0.0, "No board in final state"
|
| 148 |
+
|
| 149 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 150 |
+
|
| 151 |
+
if 32 in final_state["board"]:
|
| 152 |
+
return 1.0, "A 32 tile was successfully created from empty board."
|
| 153 |
+
return 0.0, "No 32 tile is present."
|
| 154 |
+
|
| 155 |
+
|
| 156 |
+
# Registry of all 2048 reward functions
|
| 157 |
+
REWARD_FUNCTIONS_2048 = {
|
| 158 |
+
"_validate_get_2048": _validate_get_2048,
|
| 159 |
+
"_validate_2048_make_first_move": _validate_2048_make_first_move,
|
| 160 |
+
"_validate_move_tiles_right": _validate_move_tiles_right,
|
| 161 |
+
"_validate_get_32_tile": _validate_get_32_tile,
|
| 162 |
+
"_validate_get_128_tile": _validate_get_128_tile,
|
| 163 |
+
"_validate_get_256_tile": _validate_get_256_tile,
|
| 164 |
+
"_validate_get_512_tile": _validate_get_512_tile,
|
| 165 |
+
"_validate_create_two_high_tiles": _validate_create_two_high_tiles,
|
| 166 |
+
"_validate_reach_540_sum": _validate_reach_540_sum,
|
| 167 |
+
"_validate_strategic_32": _validate_strategic_32,
|
| 168 |
+
}
|
rewards/linear.py
ADDED
|
@@ -0,0 +1,255 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Reward functions for Linear app tasks.
|
| 3 |
+
"""
|
| 4 |
+
|
| 5 |
+
import logging
|
| 6 |
+
from typing import Any, Dict, Tuple
|
| 7 |
+
|
| 8 |
+
logger = logging.getLogger(__name__)
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
def _validate_drag_to_different_column(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 12 |
+
"""Validate that an issue was moved to a different column."""
|
| 13 |
+
if "issues" not in final_state:
|
| 14 |
+
return 0.0, "No issues in final state"
|
| 15 |
+
|
| 16 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 17 |
+
|
| 18 |
+
# Find the issue we're tracking (VSS-101)
|
| 19 |
+
target_issue = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-101"), None)
|
| 20 |
+
|
| 21 |
+
if not target_issue:
|
| 22 |
+
return 0.0, "Target issue VSS-101 not found in final state"
|
| 23 |
+
|
| 24 |
+
# Check if it moved to in_progress status
|
| 25 |
+
if target_issue.get("status") == "in_progress" and target_issue.get("assigneeId") == "1":
|
| 26 |
+
return 1.0, "Issue VSS-101 successfully moved to In Progress column for user 1"
|
| 27 |
+
|
| 28 |
+
return (
|
| 29 |
+
0.0,
|
| 30 |
+
f"Issue VSS-101 has status={target_issue.get('status')}, "
|
| 31 |
+
f"assigneeId={target_issue.get('assigneeId')}, expected status=in_progress, assigneeId=1",
|
| 32 |
+
)
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
def _validate_drag_two_issues_same_user(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 36 |
+
"""Validate that two issues were moved within the same user's board."""
|
| 37 |
+
if "issues" not in final_state:
|
| 38 |
+
return 0.0, "No issues in final state"
|
| 39 |
+
|
| 40 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 41 |
+
|
| 42 |
+
# Find both target issues
|
| 43 |
+
issue_101 = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-101"), None)
|
| 44 |
+
issue_106 = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-106"), None)
|
| 45 |
+
|
| 46 |
+
if not issue_101:
|
| 47 |
+
return 0.0, "Target issue VSS-101 not found in final state"
|
| 48 |
+
if not issue_106:
|
| 49 |
+
return 0.0, "Target issue VSS-106 not found in final state"
|
| 50 |
+
|
| 51 |
+
# Check both issues
|
| 52 |
+
issue_101_correct = issue_101.get("status") == "in_progress" and issue_101.get("assigneeId") == "1"
|
| 53 |
+
issue_106_correct = issue_106.get("status") == "queued" and issue_106.get("assigneeId") == "1"
|
| 54 |
+
|
| 55 |
+
if issue_101_correct and issue_106_correct:
|
| 56 |
+
return 1.0, "Both issues successfully moved to target columns for user 1"
|
| 57 |
+
|
| 58 |
+
errors = []
|
| 59 |
+
if not issue_101_correct:
|
| 60 |
+
errors.append(
|
| 61 |
+
f"VSS-101: status={issue_101.get('status')}, assigneeId={issue_101.get('assigneeId')} (expected in_progress, 1)"
|
| 62 |
+
)
|
| 63 |
+
if not issue_106_correct:
|
| 64 |
+
errors.append(
|
| 65 |
+
f"VSS-106: status={issue_106.get('status')}, assigneeId={issue_106.get('assigneeId')} (expected queued, 1)"
|
| 66 |
+
)
|
| 67 |
+
|
| 68 |
+
return 0.0, "; ".join(errors)
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
def _validate_create_new_issue(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 72 |
+
"""Validate that a new issue was created with correct properties."""
|
| 73 |
+
if "issues" not in final_state:
|
| 74 |
+
return 0.0, "No issues in final state"
|
| 75 |
+
|
| 76 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 77 |
+
|
| 78 |
+
# Find issue by title
|
| 79 |
+
target_issue = next(
|
| 80 |
+
(issue for issue in final_state["issues"] if issue.get("title") == "Implement user authentication system"), None
|
| 81 |
+
)
|
| 82 |
+
|
| 83 |
+
if not target_issue:
|
| 84 |
+
return 0.0, "Issue with title 'Implement user authentication system' not found in final state"
|
| 85 |
+
|
| 86 |
+
# Validate properties
|
| 87 |
+
errors = []
|
| 88 |
+
|
| 89 |
+
if target_issue.get("assigneeId") != "2":
|
| 90 |
+
errors.append(f"assigneeId={target_issue.get('assigneeId')} (expected '2')")
|
| 91 |
+
|
| 92 |
+
if target_issue.get("priority") != "high":
|
| 93 |
+
errors.append(f"priority={target_issue.get('priority')} (expected 'high')")
|
| 94 |
+
|
| 95 |
+
description = target_issue.get("description", "")
|
| 96 |
+
if "JWT-based authentication" not in description:
|
| 97 |
+
errors.append("description does not contain 'JWT-based authentication'")
|
| 98 |
+
|
| 99 |
+
if errors:
|
| 100 |
+
return 0.0, "; ".join(errors)
|
| 101 |
+
|
| 102 |
+
return 1.0, "Issue successfully created with correct properties"
|
| 103 |
+
|
| 104 |
+
|
| 105 |
+
def _validate_drag_and_reassign_issue(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 106 |
+
"""Validate that issue VSS-106 was dragged to different status and assignee."""
|
| 107 |
+
if "issues" not in final_state:
|
| 108 |
+
return 0.0, "No issues in final state"
|
| 109 |
+
|
| 110 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 111 |
+
|
| 112 |
+
# Find the target issue
|
| 113 |
+
target_issue = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-106"), None)
|
| 114 |
+
|
| 115 |
+
if not target_issue:
|
| 116 |
+
return 0.0, "Target issue VSS-106 not found in final state"
|
| 117 |
+
|
| 118 |
+
# Check if it has correct status and assigneeId
|
| 119 |
+
if target_issue.get("status") == "in_progress" and target_issue.get("assigneeId") == "2":
|
| 120 |
+
return 1.0, "Issue VSS-106 successfully moved to In Progress column for Chen (assigneeId=2)"
|
| 121 |
+
|
| 122 |
+
return (
|
| 123 |
+
0.0,
|
| 124 |
+
f"Issue VSS-106 has status={target_issue.get('status')}, "
|
| 125 |
+
f"assigneeId={target_issue.get('assigneeId')}, expected status=in_progress, assigneeId=2",
|
| 126 |
+
)
|
| 127 |
+
|
| 128 |
+
|
| 129 |
+
def _validate_update_issue_properties(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 130 |
+
"""Validate that issue VSS-105 has updated priority and label."""
|
| 131 |
+
if "issues" not in final_state:
|
| 132 |
+
return 0.0, "No issues in final state"
|
| 133 |
+
|
| 134 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 135 |
+
|
| 136 |
+
# Find the target issue
|
| 137 |
+
target_issue = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-105"), None)
|
| 138 |
+
|
| 139 |
+
if not target_issue:
|
| 140 |
+
return 0.0, "Target issue VSS-105 not found in final state"
|
| 141 |
+
|
| 142 |
+
# Validate properties
|
| 143 |
+
errors = []
|
| 144 |
+
|
| 145 |
+
if target_issue.get("priority") != "urgent":
|
| 146 |
+
errors.append(f"priority={target_issue.get('priority')} (expected 'urgent')")
|
| 147 |
+
|
| 148 |
+
labels = target_issue.get("labels", [])
|
| 149 |
+
has_operations_label = any(label.get("id") == "label2" for label in labels)
|
| 150 |
+
if not has_operations_label:
|
| 151 |
+
errors.append("Operations label (id='label2') not found in labels array")
|
| 152 |
+
|
| 153 |
+
if errors:
|
| 154 |
+
return 0.0, "; ".join(errors)
|
| 155 |
+
|
| 156 |
+
return 1.0, "Issue VSS-105 successfully updated with urgent priority and Operations label"
|
| 157 |
+
|
| 158 |
+
|
| 159 |
+
def _validate_multi_issue_reorganization(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 160 |
+
"""Validate that three issues were reorganized to different columns and assignees."""
|
| 161 |
+
if "issues" not in final_state:
|
| 162 |
+
return 0.0, "No issues in final state"
|
| 163 |
+
|
| 164 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 165 |
+
|
| 166 |
+
# Find all three target issues
|
| 167 |
+
issue_101 = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-101"), None)
|
| 168 |
+
issue_102 = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-102"), None)
|
| 169 |
+
issue_103 = next((issue for issue in final_state["issues"] if issue.get("identifier") == "VSS-103"), None)
|
| 170 |
+
|
| 171 |
+
if not issue_101:
|
| 172 |
+
return 0.0, "Target issue VSS-101 not found in final state"
|
| 173 |
+
if not issue_102:
|
| 174 |
+
return 0.0, "Target issue VSS-102 not found in final state"
|
| 175 |
+
if not issue_103:
|
| 176 |
+
return 0.0, "Target issue VSS-103 not found in final state"
|
| 177 |
+
|
| 178 |
+
# Check all issues
|
| 179 |
+
issue_101_correct = issue_101.get("status") == "in_progress" and issue_101.get("assigneeId") == "2"
|
| 180 |
+
issue_102_correct = issue_102.get("status") == "blocked" and issue_102.get("assigneeId") == "3"
|
| 181 |
+
issue_103_correct = issue_103.get("status") == "in_review" and issue_103.get("assigneeId") == "1"
|
| 182 |
+
|
| 183 |
+
if issue_101_correct and issue_102_correct and issue_103_correct:
|
| 184 |
+
return 1.0, "All three issues successfully reorganized to target columns and assignees"
|
| 185 |
+
|
| 186 |
+
errors = []
|
| 187 |
+
if not issue_101_correct:
|
| 188 |
+
errors.append(
|
| 189 |
+
f"VSS-101: status={issue_101.get('status')}, assigneeId={issue_101.get('assigneeId')} (expected in_progress, 2)"
|
| 190 |
+
)
|
| 191 |
+
if not issue_102_correct:
|
| 192 |
+
errors.append(
|
| 193 |
+
f"VSS-102: status={issue_102.get('status')}, assigneeId={issue_102.get('assigneeId')} (expected blocked, 3)"
|
| 194 |
+
)
|
| 195 |
+
if not issue_103_correct:
|
| 196 |
+
errors.append(
|
| 197 |
+
f"VSS-103: status={issue_103.get('status')}, assigneeId={issue_103.get('assigneeId')} (expected in_review, 1)"
|
| 198 |
+
)
|
| 199 |
+
|
| 200 |
+
return 0.0, "; ".join(errors)
|
| 201 |
+
|
| 202 |
+
|
| 203 |
+
def _validate_complete_issue_workflow(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 204 |
+
"""Validate that a complete issue workflow was executed."""
|
| 205 |
+
if "issues" not in final_state:
|
| 206 |
+
return 0.0, "No issues in final state"
|
| 207 |
+
|
| 208 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 209 |
+
|
| 210 |
+
# Find issue by title
|
| 211 |
+
target_issue = next(
|
| 212 |
+
(issue for issue in final_state["issues"] if issue.get("title") == "Fix critical database connection bug"), None
|
| 213 |
+
)
|
| 214 |
+
|
| 215 |
+
if not target_issue:
|
| 216 |
+
return 0.0, "Issue with title 'Fix critical database connection bug' not found in final state"
|
| 217 |
+
|
| 218 |
+
# Validate all properties
|
| 219 |
+
errors = []
|
| 220 |
+
|
| 221 |
+
if target_issue.get("priority") != "high":
|
| 222 |
+
errors.append(f"priority={target_issue.get('priority')} (expected 'high')")
|
| 223 |
+
|
| 224 |
+
if target_issue.get("assigneeId") != "2":
|
| 225 |
+
errors.append(f"assigneeId={target_issue.get('assigneeId')} (expected '2')")
|
| 226 |
+
|
| 227 |
+
if target_issue.get("status") != "in_progress":
|
| 228 |
+
errors.append(f"status={target_issue.get('status')} (expected 'in_progress')")
|
| 229 |
+
|
| 230 |
+
labels = target_issue.get("labels", [])
|
| 231 |
+
has_bug_label = any(label.get("id") == "label1" for label in labels)
|
| 232 |
+
if not has_bug_label:
|
| 233 |
+
errors.append("Bug label (id='label1') not found in labels array")
|
| 234 |
+
|
| 235 |
+
comments = target_issue.get("comments", [])
|
| 236 |
+
has_comment = any("Starting investigation" in comment.get("text", "") for comment in comments)
|
| 237 |
+
if not has_comment:
|
| 238 |
+
errors.append("Comment containing 'Starting investigation' not found")
|
| 239 |
+
|
| 240 |
+
if errors:
|
| 241 |
+
return 0.0, "; ".join(errors)
|
| 242 |
+
|
| 243 |
+
return 1.0, "Complete issue workflow successfully executed"
|
| 244 |
+
|
| 245 |
+
|
| 246 |
+
# Registry of all Linear reward functions
|
| 247 |
+
REWARD_FUNCTIONS_LINEAR = {
|
| 248 |
+
"_validate_drag_to_different_column": _validate_drag_to_different_column,
|
| 249 |
+
"_validate_drag_two_issues_same_user": _validate_drag_two_issues_same_user,
|
| 250 |
+
"_validate_create_new_issue": _validate_create_new_issue,
|
| 251 |
+
"_validate_drag_and_reassign_issue": _validate_drag_and_reassign_issue,
|
| 252 |
+
"_validate_update_issue_properties": _validate_update_issue_properties,
|
| 253 |
+
"_validate_multi_issue_reorganization": _validate_multi_issue_reorganization,
|
| 254 |
+
"_validate_complete_issue_workflow": _validate_complete_issue_workflow,
|
| 255 |
+
}
|
rewards/linkedin.py
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Reward functions for LinkedIn app tasks.
|
| 3 |
+
"""
|
| 4 |
+
|
| 5 |
+
import logging
|
| 6 |
+
from typing import Any, Dict, Tuple
|
| 7 |
+
|
| 8 |
+
logger = logging.getLogger(__name__)
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
def _validate_search_for_dzaka(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 12 |
+
"""Validate that the user successfully searched for Dzaka Athif."""
|
| 13 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 14 |
+
|
| 15 |
+
# Check 1: currentView is "search"
|
| 16 |
+
if final_state.get("currentView") != "search":
|
| 17 |
+
return 0.0, f"Not on search page, current view: {final_state.get('currentView')}"
|
| 18 |
+
|
| 19 |
+
# Check 2: searchQuery contains "dzaka" (case insensitive)
|
| 20 |
+
query = final_state.get("searchQuery", "").lower()
|
| 21 |
+
if "dzaka" not in query:
|
| 22 |
+
return 0.0, f"Search query doesn't contain 'dzaka': {final_state.get('searchQuery')}"
|
| 23 |
+
|
| 24 |
+
# Check 3: Dzaka Athif in search results
|
| 25 |
+
search_results = final_state.get("searchResults", {})
|
| 26 |
+
people = search_results.get("allPeople", [])
|
| 27 |
+
|
| 28 |
+
dzaka_found = any(user.get("name") == "Dzaka Athif" for user in people)
|
| 29 |
+
|
| 30 |
+
if dzaka_found:
|
| 31 |
+
return 1.0, "Successfully searched for Dzaka Athif"
|
| 32 |
+
|
| 33 |
+
return 0.0, f"Dzaka Athif not found in search results. Found {len(people)} people."
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
def _validate_fractional_find_user_profile(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 37 |
+
"""Validate LinkedIn profile navigation with fractional scoring.
|
| 38 |
+
|
| 39 |
+
Awards 0.33 points for each criterion met:
|
| 40 |
+
- currentView is 'profile'
|
| 41 |
+
- viewedUserId is '2' (John Smith)
|
| 42 |
+
- searchQuery contains 'john smith'
|
| 43 |
+
"""
|
| 44 |
+
logger.debug(f"Running reward function on state: {final_state}")
|
| 45 |
+
|
| 46 |
+
score = 0.0
|
| 47 |
+
passed = []
|
| 48 |
+
failed = []
|
| 49 |
+
|
| 50 |
+
# Check 1: currentView is 'profile' (0.33 points)
|
| 51 |
+
if final_state.get("currentView") == "profile":
|
| 52 |
+
score += 0.33
|
| 53 |
+
passed.append("currentView='profile'")
|
| 54 |
+
else:
|
| 55 |
+
failed.append(f"currentView='{final_state.get('currentView')}' (expected 'profile')")
|
| 56 |
+
|
| 57 |
+
# Check 2: viewedUserId is '2' (0.33 points)
|
| 58 |
+
if final_state.get("viewedUserId") == "2":
|
| 59 |
+
score += 0.33
|
| 60 |
+
passed.append("viewedUserId='2'")
|
| 61 |
+
else:
|
| 62 |
+
failed.append(f"viewedUserId='{final_state.get('viewedUserId')}' (expected '2')")
|
| 63 |
+
|
| 64 |
+
# Check 3: searchQuery contains 'john smith' (0.33 points)
|
| 65 |
+
query = final_state.get("searchQuery", "").lower()
|
| 66 |
+
if "john smith" in query:
|
| 67 |
+
score += 0.34 # Make it add up to 1.0
|
| 68 |
+
passed.append("searchQuery contains 'john smith'")
|
| 69 |
+
else:
|
| 70 |
+
failed.append(f"searchQuery='{final_state.get('searchQuery')}' (expected to contain 'john smith')")
|
| 71 |
+
|
| 72 |
+
# Build reason string
|
| 73 |
+
reason_parts = []
|
| 74 |
+
if passed:
|
| 75 |
+
reason_parts.append(f"Passed: {', '.join(passed)}")
|
| 76 |
+
if failed:
|
| 77 |
+
reason_parts.append(f"Failed: {', '.join(failed)}")
|
| 78 |
+
|
| 79 |
+
reason = "; ".join(reason_parts) if reason_parts else "No criteria evaluated"
|
| 80 |
+
|
| 81 |
+
return round(score, 2), reason
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
# Registry of all LinkedIn reward functions
|
| 85 |
+
REWARD_FUNCTIONS_LINKEDIN = {
|
| 86 |
+
"_validate_search_for_dzaka": _validate_search_for_dzaka,
|
| 87 |
+
"_validate_fractional_find_user_profile": _validate_fractional_find_user_profile,
|
| 88 |
+
}
|
rewards/minesweeper.py
ADDED
|
@@ -0,0 +1,255 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Reward functions for Minesweeper game tasks.
|
| 3 |
+
|
| 4 |
+
Board is a 9x9 grid (2D array) of Cell objects:
|
| 5 |
+
Cell: {
|
| 6 |
+
isMine: boolean,
|
| 7 |
+
isRevealed: boolean,
|
| 8 |
+
isFlagged: boolean,
|
| 9 |
+
neighborMineCount: number
|
| 10 |
+
}
|
| 11 |
+
|
| 12 |
+
Game statuses: "not_started", "playing", "won", "lost"
|
| 13 |
+
"""
|
| 14 |
+
|
| 15 |
+
import logging
|
| 16 |
+
from typing import Any, Dict, List, Tuple
|
| 17 |
+
|
| 18 |
+
logger = logging.getLogger(__name__)
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
# Helper functions for common validation patterns
|
| 22 |
+
|
| 23 |
+
|
| 24 |
+
def _check_cells_flagged(board: List[List[Dict[str, Any]]], positions: List[Tuple[int, int]]) -> Tuple[float, str]:
|
| 25 |
+
"""
|
| 26 |
+
Helper function to check if specific cells are flagged.
|
| 27 |
+
|
| 28 |
+
Args:
|
| 29 |
+
board: The game board (2D array)
|
| 30 |
+
positions: List of (row, col) tuples to check
|
| 31 |
+
|
| 32 |
+
Returns:
|
| 33 |
+
(1.0, success_message) if all positions are flagged
|
| 34 |
+
(0.0, error_message) otherwise
|
| 35 |
+
"""
|
| 36 |
+
flagged = []
|
| 37 |
+
unflagged = []
|
| 38 |
+
|
| 39 |
+
for row, col in positions:
|
| 40 |
+
# Check if position exists
|
| 41 |
+
if row >= len(board) or col >= len(board[row]):
|
| 42 |
+
return 0.0, f"Position [{row}][{col}] does not exist on board"
|
| 43 |
+
|
| 44 |
+
if board[row][col].get("isFlagged"):
|
| 45 |
+
flagged.append(f"[{row}][{col}]")
|
| 46 |
+
else:
|
| 47 |
+
unflagged.append(f"[{row}][{col}]")
|
| 48 |
+
|
| 49 |
+
if len(flagged) == len(positions):
|
| 50 |
+
cells_str = ", ".join(flagged)
|
| 51 |
+
return 1.0, f"All {len(positions)} cell(s) flagged successfully: {cells_str}."
|
| 52 |
+
|
| 53 |
+
return 0.0, f"Not all cells flagged. Flagged: {flagged}, Not flagged: {unflagged}"
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
def _check_cells_revealed(board: List[List[Dict[str, Any]]], min_count: int, check_for_mines: bool = True) -> Tuple[float, str]:
|
| 57 |
+
"""
|
| 58 |
+
Helper function to check if minimum number of cells are revealed.
|
| 59 |
+
|
| 60 |
+
Args:
|
| 61 |
+
board: The game board (2D array)
|
| 62 |
+
min_count: Minimum number of cells that should be revealed
|
| 63 |
+
check_for_mines: If True, fail if any revealed cell is a mine
|
| 64 |
+
|
| 65 |
+
Returns:
|
| 66 |
+
(1.0, success_message) if criteria met
|
| 67 |
+
(0.0, error_message) otherwise
|
| 68 |
+
"""
|
| 69 |
+
revealed_count = 0
|
| 70 |
+
revealed_mine = False
|
| 71 |
+
|
| 72 |
+
for row in board:
|
| 73 |
+
for cell in row:
|
| 74 |
+
if cell.get("isRevealed"):
|
| 75 |
+
revealed_count += 1
|
| 76 |
+
if check_for_mines and cell.get("isMine"):
|
| 77 |
+
revealed_mine = True
|
| 78 |
+
|
| 79 |
+
if check_for_mines and revealed_mine:
|
| 80 |
+
return 0.0, "A mine was revealed - task failed"
|
| 81 |
+
|
| 82 |
+
if revealed_count >= min_count:
|
| 83 |
+
return 1.0, f"Successfully revealed {revealed_count} safe cells (target: {min_count}+)."
|
| 84 |
+
|
| 85 |
+
return 0.0, f"Only {revealed_count} cells revealed, need at least {min_count}"
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
# Task-specific validation functions
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
def _validate_minesweeper_reveal_first_cell(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 92 |
+
"""Validate that first move was made - at least one cell revealed and game is playing."""
|
| 93 |
+
if "gameStatus" not in final_state:
|
| 94 |
+
return 0.0, "No gameStatus in final state"
|
| 95 |
+
|
| 96 |
+
if "board" not in final_state:
|
| 97 |
+
return 0.0, "No board in final state"
|
| 98 |
+
|
| 99 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 100 |
+
|
| 101 |
+
# Check if game status changed to playing
|
| 102 |
+
if final_state["gameStatus"] != "playing":
|
| 103 |
+
return 0.0, f"Game status is '{final_state['gameStatus']}', expected 'playing'"
|
| 104 |
+
|
| 105 |
+
# Check if at least one cell is revealed
|
| 106 |
+
board = final_state["board"]
|
| 107 |
+
revealed_count = 0
|
| 108 |
+
|
| 109 |
+
for row in board:
|
| 110 |
+
for cell in row:
|
| 111 |
+
if cell.get("isRevealed"):
|
| 112 |
+
revealed_count += 1
|
| 113 |
+
|
| 114 |
+
if revealed_count > 0:
|
| 115 |
+
return 1.0, f"First move successful - {revealed_count} cell(s) revealed and game is playing."
|
| 116 |
+
|
| 117 |
+
return 0.0, "No cells revealed yet."
|
| 118 |
+
|
| 119 |
+
|
| 120 |
+
def _validate_minesweeper_reveal_numbered_cell(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 121 |
+
"""Validate that a numbered cell (with neighborMineCount > 0) was revealed at position [1][1]."""
|
| 122 |
+
if "board" not in final_state:
|
| 123 |
+
return 0.0, "No board in final state"
|
| 124 |
+
|
| 125 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 126 |
+
|
| 127 |
+
board = final_state["board"]
|
| 128 |
+
|
| 129 |
+
# Check if position [1][1] exists
|
| 130 |
+
if len(board) < 2 or len(board[1]) < 2:
|
| 131 |
+
return 0.0, "Board does not have position [1][1]"
|
| 132 |
+
|
| 133 |
+
target_cell = board[1][1]
|
| 134 |
+
|
| 135 |
+
# Check if cell is revealed
|
| 136 |
+
if not target_cell.get("isRevealed"):
|
| 137 |
+
return 0.0, "Cell at [1][1] is not revealed"
|
| 138 |
+
|
| 139 |
+
# Check if cell shows a number (neighborMineCount > 0)
|
| 140 |
+
neighbor_count = target_cell.get("neighborMineCount", 0)
|
| 141 |
+
|
| 142 |
+
if neighbor_count > 0:
|
| 143 |
+
return 1.0, f"Cell at [1][1] revealed successfully showing {neighbor_count} neighboring mine(s)."
|
| 144 |
+
|
| 145 |
+
return 0.0, f"Cell at [1][1] has neighborMineCount of {neighbor_count}, expected > 0"
|
| 146 |
+
|
| 147 |
+
|
| 148 |
+
def _validate_minesweeper_reveal_10_cells(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 149 |
+
"""Validate that at least 10 cells are revealed, none are mines, and game is still playing."""
|
| 150 |
+
if "board" not in final_state:
|
| 151 |
+
return 0.0, "No board in final state"
|
| 152 |
+
|
| 153 |
+
if "gameStatus" not in final_state:
|
| 154 |
+
return 0.0, "No gameStatus in final state"
|
| 155 |
+
|
| 156 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 157 |
+
|
| 158 |
+
# Check if game is still playing
|
| 159 |
+
if final_state["gameStatus"] != "playing":
|
| 160 |
+
return 0.0, f"Game status is '{final_state['gameStatus']}', expected 'playing'"
|
| 161 |
+
|
| 162 |
+
return _check_cells_revealed(final_state["board"], 10, check_for_mines=True)
|
| 163 |
+
|
| 164 |
+
|
| 165 |
+
def _validate_minesweeper_reveal_20_cells(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 166 |
+
"""Validate that at least 20 cells are revealed and game is still playing."""
|
| 167 |
+
if "board" not in final_state:
|
| 168 |
+
return 0.0, "No board in final state"
|
| 169 |
+
|
| 170 |
+
if "gameStatus" not in final_state:
|
| 171 |
+
return 0.0, "No gameStatus in final state"
|
| 172 |
+
|
| 173 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 174 |
+
|
| 175 |
+
# Check if game is still playing
|
| 176 |
+
if final_state["gameStatus"] != "playing":
|
| 177 |
+
return 0.0, f"Game status is '{final_state['gameStatus']}', expected 'playing' (game should not be won or lost yet)"
|
| 178 |
+
|
| 179 |
+
return _check_cells_revealed(final_state["board"], 20, check_for_mines=True)
|
| 180 |
+
|
| 181 |
+
|
| 182 |
+
def _validate_minesweeper_flag_single_mine(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 183 |
+
"""Validate that cell at position [0][0] is flagged."""
|
| 184 |
+
if "board" not in final_state:
|
| 185 |
+
return 0.0, "No board in final state"
|
| 186 |
+
|
| 187 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 188 |
+
|
| 189 |
+
return _check_cells_flagged(final_state["board"], [(0, 0)])
|
| 190 |
+
|
| 191 |
+
|
| 192 |
+
def _validate_minesweeper_flag_three_mines(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 193 |
+
"""Validate that cells at [0][0], [0][8], and [8][0] are all flagged."""
|
| 194 |
+
if "board" not in final_state:
|
| 195 |
+
return 0.0, "No board in final state"
|
| 196 |
+
|
| 197 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 198 |
+
|
| 199 |
+
target_positions = [(0, 0), (0, 8), (8, 0)]
|
| 200 |
+
return _check_cells_flagged(final_state["board"], target_positions)
|
| 201 |
+
|
| 202 |
+
|
| 203 |
+
def _validate_minesweeper_flag_corner_mines(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 204 |
+
"""Validate that all four corner cells [0][0], [0][8], [8][0], [8][8] are flagged."""
|
| 205 |
+
if "board" not in final_state:
|
| 206 |
+
return 0.0, "No board in final state"
|
| 207 |
+
|
| 208 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 209 |
+
|
| 210 |
+
corner_positions = [(0, 0), (0, 8), (8, 0), (8, 8)]
|
| 211 |
+
return _check_cells_flagged(final_state["board"], corner_positions)
|
| 212 |
+
|
| 213 |
+
|
| 214 |
+
def _validate_minesweeper_reveal_safe_cell(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 215 |
+
"""Validate that cell at position [4][4] is revealed and has 0 neighboring mines."""
|
| 216 |
+
if "board" not in final_state:
|
| 217 |
+
return 0.0, "No board in final state"
|
| 218 |
+
|
| 219 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 220 |
+
|
| 221 |
+
board = final_state["board"]
|
| 222 |
+
|
| 223 |
+
# Check if position [4][4] exists
|
| 224 |
+
if len(board) < 5 or len(board[4]) < 5:
|
| 225 |
+
return 0.0, "Board does not have position [4][4]"
|
| 226 |
+
|
| 227 |
+
target_cell = board[4][4]
|
| 228 |
+
|
| 229 |
+
# Check if cell is revealed
|
| 230 |
+
if not target_cell.get("isRevealed"):
|
| 231 |
+
return 0.0, "Cell at [4][4] is not revealed"
|
| 232 |
+
|
| 233 |
+
# Check if it's not a mine
|
| 234 |
+
if target_cell.get("isMine"):
|
| 235 |
+
return 0.0, "Cell at [4][4] is a mine - task failed!"
|
| 236 |
+
|
| 237 |
+
# Check if it has 0 neighboring mines
|
| 238 |
+
neighbor_count = target_cell.get("neighborMineCount", -1)
|
| 239 |
+
if neighbor_count == 0:
|
| 240 |
+
return 1.0, "Cell at [4][4] revealed successfully with 0 neighboring mines."
|
| 241 |
+
|
| 242 |
+
return 0.0, f"Cell at [4][4] has {neighbor_count} neighboring mines, expected 0"
|
| 243 |
+
|
| 244 |
+
|
| 245 |
+
# Registry of all minesweeper reward functions
|
| 246 |
+
REWARD_FUNCTIONS_MINESWEEPER = {
|
| 247 |
+
"_validate_minesweeper_reveal_first_cell": _validate_minesweeper_reveal_first_cell,
|
| 248 |
+
"_validate_minesweeper_reveal_numbered_cell": _validate_minesweeper_reveal_numbered_cell,
|
| 249 |
+
"_validate_minesweeper_reveal_10_cells": _validate_minesweeper_reveal_10_cells,
|
| 250 |
+
"_validate_minesweeper_reveal_20_cells": _validate_minesweeper_reveal_20_cells,
|
| 251 |
+
"_validate_minesweeper_flag_single_mine": _validate_minesweeper_flag_single_mine,
|
| 252 |
+
"_validate_minesweeper_flag_three_mines": _validate_minesweeper_flag_three_mines,
|
| 253 |
+
"_validate_minesweeper_flag_corner_mines": _validate_minesweeper_flag_corner_mines,
|
| 254 |
+
"_validate_minesweeper_reveal_safe_cell": _validate_minesweeper_reveal_safe_cell,
|
| 255 |
+
}
|
rewards/salesforce.py
ADDED
|
@@ -0,0 +1,423 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Salesforce reward functions for CRM task validation.
|
| 3 |
+
|
| 4 |
+
This module contains reward validation functions for Salesforce CRM tasks,
|
| 5 |
+
particularly for complex workflows like lead conversion.
|
| 6 |
+
"""
|
| 7 |
+
|
| 8 |
+
from typing import Any, Dict, Tuple
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
def _validate_salesforce_convert_lead(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 12 |
+
"""
|
| 13 |
+
Validate that a lead has been successfully converted to a contact.
|
| 14 |
+
|
| 15 |
+
This function checks:
|
| 16 |
+
1. The active tab is the leads list page
|
| 17 |
+
2. The lead has status 'Converted' and a convertedToContactId
|
| 18 |
+
3. A new contact exists with the lead's information (name: Michael Brown)
|
| 19 |
+
|
| 20 |
+
Args:
|
| 21 |
+
initial_state: The initial state before the conversion
|
| 22 |
+
final_state: The final state after the conversion
|
| 23 |
+
|
| 24 |
+
Returns:
|
| 25 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 26 |
+
"""
|
| 27 |
+
try:
|
| 28 |
+
# Check 1: Active tab should be the leads list page
|
| 29 |
+
active_tab_id = final_state.get("activeTabId")
|
| 30 |
+
if active_tab_id != "home-listLeads":
|
| 31 |
+
return 0.0, f"Expected activeTabId to be 'home-listLeads', got '{active_tab_id}'"
|
| 32 |
+
|
| 33 |
+
# Check 2: Lead should be marked as converted
|
| 34 |
+
leads = final_state.get("leads", [])
|
| 35 |
+
if not leads:
|
| 36 |
+
return 0.0, "No leads found in final state"
|
| 37 |
+
|
| 38 |
+
# Find the lead (should be lead-001)
|
| 39 |
+
lead = None
|
| 40 |
+
for temp_lead in leads:
|
| 41 |
+
if temp_lead.get("id") == "lead-001":
|
| 42 |
+
lead = temp_lead
|
| 43 |
+
break
|
| 44 |
+
|
| 45 |
+
if not lead:
|
| 46 |
+
return 0.0, "Lead with id 'lead-001' not found"
|
| 47 |
+
|
| 48 |
+
# Check lead status and convertedToContactId
|
| 49 |
+
lead_status = lead.get("leadStatus")
|
| 50 |
+
converted_to_contact_id = lead.get("convertedToContactId")
|
| 51 |
+
|
| 52 |
+
if lead_status != "Converted":
|
| 53 |
+
return 0.0, f"Expected leadStatus to be 'Converted', got '{lead_status}'"
|
| 54 |
+
|
| 55 |
+
if not converted_to_contact_id:
|
| 56 |
+
return 0.0, "Expected convertedToContactId to be set, but it's missing"
|
| 57 |
+
|
| 58 |
+
# Check 3: Contact should exist with the lead's information
|
| 59 |
+
contacts = final_state.get("contacts", [])
|
| 60 |
+
if not contacts:
|
| 61 |
+
return 0.0, "No contacts found in final state"
|
| 62 |
+
|
| 63 |
+
# Find the contact by the convertedToContactId
|
| 64 |
+
contact = None
|
| 65 |
+
for c in contacts:
|
| 66 |
+
if c.get("id") == converted_to_contact_id:
|
| 67 |
+
contact = c
|
| 68 |
+
break
|
| 69 |
+
|
| 70 |
+
if not contact:
|
| 71 |
+
return 0.0, f"Contact with id '{converted_to_contact_id}' not found"
|
| 72 |
+
|
| 73 |
+
# Verify the contact has the correct information (name should be Michael Brown)
|
| 74 |
+
contact_name = contact.get("name")
|
| 75 |
+
if contact_name != "Michael Brown":
|
| 76 |
+
return 0.0, f"Expected contact name to be 'Michael Brown', got '{contact_name}'"
|
| 77 |
+
|
| 78 |
+
# Additional validation: check that contact has the lead's company
|
| 79 |
+
contact_company = contact.get("company")
|
| 80 |
+
lead_company = lead.get("company")
|
| 81 |
+
if contact_company != lead_company:
|
| 82 |
+
return 0.0, f"Contact company '{contact_company}' doesn't match lead company '{lead_company}'"
|
| 83 |
+
|
| 84 |
+
# All checks passed
|
| 85 |
+
return 1.0, "Lead successfully converted to contact with correct information"
|
| 86 |
+
|
| 87 |
+
except Exception as e:
|
| 88 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
def _validate_salesforce_create_and_convert_lead(
|
| 92 |
+
initial_state: Dict[str, Any], final_state: Dict[str, Any]
|
| 93 |
+
) -> Tuple[float, str]:
|
| 94 |
+
"""
|
| 95 |
+
Validate that a lead has been created and successfully converted to a contact.
|
| 96 |
+
|
| 97 |
+
This function checks:
|
| 98 |
+
1. The active tab is the leads list page
|
| 99 |
+
2. A lead exists with name 'Robert Taylor' and status 'Converted'
|
| 100 |
+
3. The lead has a convertedToContactId set
|
| 101 |
+
4. A new contact exists with the lead's information
|
| 102 |
+
5. The contact has the correct name and company
|
| 103 |
+
|
| 104 |
+
Args:
|
| 105 |
+
initial_state: The initial state before the creation and conversion
|
| 106 |
+
final_state: The final state after the creation and conversion
|
| 107 |
+
|
| 108 |
+
Returns:
|
| 109 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 110 |
+
"""
|
| 111 |
+
try:
|
| 112 |
+
# Check 1: Active tab should be the leads list page
|
| 113 |
+
active_tab_id = final_state.get("activeTabId")
|
| 114 |
+
if active_tab_id != "home-listLeads":
|
| 115 |
+
return 0.0, f"Expected activeTabId to be 'home-listLeads', got '{active_tab_id}'"
|
| 116 |
+
|
| 117 |
+
# Check 2: Lead should exist and be marked as converted
|
| 118 |
+
leads = final_state.get("leads", [])
|
| 119 |
+
if not leads:
|
| 120 |
+
return 0.0, "No leads found in final state"
|
| 121 |
+
|
| 122 |
+
# Find the lead with name 'Robert Taylor'
|
| 123 |
+
robert_lead = None
|
| 124 |
+
for lead in leads:
|
| 125 |
+
if lead.get("name") == "Robert Taylor":
|
| 126 |
+
robert_lead = lead
|
| 127 |
+
break
|
| 128 |
+
|
| 129 |
+
if not robert_lead:
|
| 130 |
+
return 0.0, "Lead with name 'Robert Taylor' not found"
|
| 131 |
+
|
| 132 |
+
# Check lead status and convertedToContactId
|
| 133 |
+
lead_status = robert_lead.get("leadStatus")
|
| 134 |
+
converted_to_contact_id = robert_lead.get("convertedToContactId")
|
| 135 |
+
|
| 136 |
+
if lead_status != "Converted":
|
| 137 |
+
return 0.0, f"Expected leadStatus to be 'Converted', got '{lead_status}'"
|
| 138 |
+
|
| 139 |
+
if not converted_to_contact_id:
|
| 140 |
+
return 0.0, "Expected convertedToContactId to be set, but it's missing"
|
| 141 |
+
|
| 142 |
+
# Check 3: Contact should exist with the lead's information
|
| 143 |
+
contacts = final_state.get("contacts", [])
|
| 144 |
+
if not contacts:
|
| 145 |
+
return 0.0, "No contacts found in final state"
|
| 146 |
+
|
| 147 |
+
# Find the contact by the convertedToContactId
|
| 148 |
+
contact = None
|
| 149 |
+
for c in contacts:
|
| 150 |
+
if c.get("id") == converted_to_contact_id:
|
| 151 |
+
contact = c
|
| 152 |
+
break
|
| 153 |
+
|
| 154 |
+
if not contact:
|
| 155 |
+
return 0.0, f"Contact with id '{converted_to_contact_id}' not found"
|
| 156 |
+
|
| 157 |
+
# Verify the contact has the correct information
|
| 158 |
+
contact_name = contact.get("name")
|
| 159 |
+
if contact_name != "Robert Taylor":
|
| 160 |
+
return 0.0, f"Expected contact name to be 'Robert Taylor', got '{contact_name}'"
|
| 161 |
+
|
| 162 |
+
# Check that contact has the lead's company
|
| 163 |
+
contact_company = contact.get("company")
|
| 164 |
+
lead_company = robert_lead.get("company")
|
| 165 |
+
if contact_company != lead_company:
|
| 166 |
+
return 0.0, f"Contact company '{contact_company}' doesn't match lead company '{lead_company}'"
|
| 167 |
+
|
| 168 |
+
# Additional validation: check that contact has the lead's email
|
| 169 |
+
contact_email = contact.get("email")
|
| 170 |
+
lead_email = robert_lead.get("email")
|
| 171 |
+
if contact_email != lead_email:
|
| 172 |
+
return 0.0, f"Contact email '{contact_email}' doesn't match lead email '{lead_email}'"
|
| 173 |
+
|
| 174 |
+
# Verify the lead was created from scratch (not in initial state)
|
| 175 |
+
initial_leads = initial_state.get("leads", [])
|
| 176 |
+
initial_robert_exists = any(lead.get("name") == "Robert Taylor" for lead in initial_leads)
|
| 177 |
+
if initial_robert_exists:
|
| 178 |
+
return 0.0, "Lead 'Robert Taylor' already existed in initial state - should be newly created"
|
| 179 |
+
|
| 180 |
+
# All checks passed
|
| 181 |
+
return 1.0, "Lead 'Robert Taylor' successfully created and converted to contact with correct information"
|
| 182 |
+
|
| 183 |
+
except Exception as e:
|
| 184 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 185 |
+
|
| 186 |
+
|
| 187 |
+
def _validate_salesforce_create_new_lead(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 188 |
+
"""
|
| 189 |
+
Validate that a new lead has been successfully created.
|
| 190 |
+
|
| 191 |
+
This function checks:
|
| 192 |
+
1. The active tab is a lead details page
|
| 193 |
+
2. A lead exists with name 'Jane Smith' and company 'Innovation Labs'
|
| 194 |
+
3. The lead was not in the initial state (proves it was created)
|
| 195 |
+
|
| 196 |
+
Args:
|
| 197 |
+
initial_state: The initial state before the lead creation
|
| 198 |
+
final_state: The final state after the lead creation
|
| 199 |
+
|
| 200 |
+
Returns:
|
| 201 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 202 |
+
"""
|
| 203 |
+
try:
|
| 204 |
+
# Check 1: Active tab should be a lead details page
|
| 205 |
+
active_tab_id = final_state.get("activeTabId")
|
| 206 |
+
if not active_tab_id or not active_tab_id.startswith("lead-"):
|
| 207 |
+
return 0.0, f"Expected activeTabId to start with 'lead-', got '{active_tab_id}'"
|
| 208 |
+
|
| 209 |
+
# Check 2: Lead should exist with correct information
|
| 210 |
+
leads = final_state.get("leads", [])
|
| 211 |
+
if not leads:
|
| 212 |
+
return 0.0, "No leads found in final state"
|
| 213 |
+
|
| 214 |
+
# Find the lead with name 'Jane Smith'
|
| 215 |
+
jane_lead = None
|
| 216 |
+
for lead in leads:
|
| 217 |
+
if lead.get("name") == "Jane Smith":
|
| 218 |
+
jane_lead = lead
|
| 219 |
+
break
|
| 220 |
+
|
| 221 |
+
if not jane_lead:
|
| 222 |
+
return 0.0, "Lead with name 'Jane Smith' not found"
|
| 223 |
+
|
| 224 |
+
# Check company
|
| 225 |
+
if jane_lead.get("company") != "Innovation Labs":
|
| 226 |
+
return 0.0, f"Expected company to be 'Innovation Labs', got '{jane_lead.get('company')}'"
|
| 227 |
+
|
| 228 |
+
# Check 3: Verify the lead was created from scratch
|
| 229 |
+
initial_leads = initial_state.get("leads", [])
|
| 230 |
+
initial_jane_exists = any(lead.get("name") == "Jane Smith" for lead in initial_leads)
|
| 231 |
+
if initial_jane_exists:
|
| 232 |
+
return 0.0, "Lead 'Jane Smith' already existed in initial state - should be newly created"
|
| 233 |
+
|
| 234 |
+
# All checks passed
|
| 235 |
+
return 1.0, "Lead 'Jane Smith' successfully created with correct information"
|
| 236 |
+
|
| 237 |
+
except Exception as e:
|
| 238 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 239 |
+
|
| 240 |
+
|
| 241 |
+
def _validate_salesforce_create_new_case(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 242 |
+
"""
|
| 243 |
+
Validate that a new case has been successfully created.
|
| 244 |
+
|
| 245 |
+
This function checks:
|
| 246 |
+
1. The active tab is a case details page
|
| 247 |
+
2. A case exists with subject 'Technical support request'
|
| 248 |
+
3. The case was not in the initial state (proves it was created)
|
| 249 |
+
|
| 250 |
+
Args:
|
| 251 |
+
initial_state: The initial state before the case creation
|
| 252 |
+
final_state: The final state after the case creation
|
| 253 |
+
|
| 254 |
+
Returns:
|
| 255 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 256 |
+
"""
|
| 257 |
+
try:
|
| 258 |
+
# Check 1: Active tab should be a case details page
|
| 259 |
+
active_tab_id = final_state.get("activeTabId")
|
| 260 |
+
if not active_tab_id or not active_tab_id.startswith("case-"):
|
| 261 |
+
return 0.0, f"Expected activeTabId to start with 'case-', got '{active_tab_id}'"
|
| 262 |
+
|
| 263 |
+
# Check 2: Case should exist with correct subject
|
| 264 |
+
cases = final_state.get("cases", [])
|
| 265 |
+
if not cases:
|
| 266 |
+
return 0.0, "No cases found in final state"
|
| 267 |
+
|
| 268 |
+
# Find the case with subject 'Technical support request'
|
| 269 |
+
support_case = None
|
| 270 |
+
for case in cases:
|
| 271 |
+
if case.get("subject") == "Technical support request":
|
| 272 |
+
support_case = case
|
| 273 |
+
break
|
| 274 |
+
|
| 275 |
+
if not support_case:
|
| 276 |
+
return 0.0, "Case with subject 'Technical support request' not found"
|
| 277 |
+
|
| 278 |
+
# Check 3: Verify the case was created from scratch
|
| 279 |
+
initial_cases = initial_state.get("cases", [])
|
| 280 |
+
initial_case_exists = any(case.get("subject") == "Technical support request" for case in initial_cases)
|
| 281 |
+
if initial_case_exists:
|
| 282 |
+
return 0.0, "Case 'Technical support request' already existed in initial state - should be newly created"
|
| 283 |
+
|
| 284 |
+
# All checks passed
|
| 285 |
+
return 1.0, "Case 'Technical support request' successfully created"
|
| 286 |
+
|
| 287 |
+
except Exception as e:
|
| 288 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 289 |
+
|
| 290 |
+
|
| 291 |
+
def _validate_salesforce_edit_lead_inline(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 292 |
+
"""
|
| 293 |
+
Validate that a lead's title has been successfully edited inline.
|
| 294 |
+
|
| 295 |
+
This function checks:
|
| 296 |
+
1. Lead with id 'lead-001' exists
|
| 297 |
+
2. The lead's title has been changed from 'VP of Sales' to 'Senior VP of Sales'
|
| 298 |
+
3. Other lead fields remain unchanged
|
| 299 |
+
|
| 300 |
+
Args:
|
| 301 |
+
initial_state: The initial state before the edit
|
| 302 |
+
final_state: The final state after the edit
|
| 303 |
+
|
| 304 |
+
Returns:
|
| 305 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 306 |
+
"""
|
| 307 |
+
try:
|
| 308 |
+
# Get the lead from both states
|
| 309 |
+
initial_leads = initial_state.get("leads", [])
|
| 310 |
+
final_leads = final_state.get("leads", [])
|
| 311 |
+
|
| 312 |
+
initial_lead = None
|
| 313 |
+
for lead in initial_leads:
|
| 314 |
+
if lead.get("id") == "lead-001":
|
| 315 |
+
initial_lead = lead
|
| 316 |
+
break
|
| 317 |
+
|
| 318 |
+
final_lead = None
|
| 319 |
+
for lead in final_leads:
|
| 320 |
+
if lead.get("id") == "lead-001":
|
| 321 |
+
final_lead = lead
|
| 322 |
+
break
|
| 323 |
+
|
| 324 |
+
if not initial_lead:
|
| 325 |
+
return 0.0, "Lead 'lead-001' not found in initial state"
|
| 326 |
+
|
| 327 |
+
if not final_lead:
|
| 328 |
+
return 0.0, "Lead 'lead-001' not found in final state"
|
| 329 |
+
|
| 330 |
+
# Check that title was updated correctly
|
| 331 |
+
if initial_lead.get("title") != "VP of Sales":
|
| 332 |
+
return 0.0, f"Initial title should be 'VP of Sales', got '{initial_lead.get('title')}'"
|
| 333 |
+
|
| 334 |
+
if final_lead.get("title") != "Senior VP of Sales":
|
| 335 |
+
return 0.0, f"Expected title to be 'Senior VP of Sales', got '{final_lead.get('title')}'"
|
| 336 |
+
|
| 337 |
+
# Check that other important fields remain unchanged
|
| 338 |
+
fields_to_check = ["name", "company", "email", "phone", "leadStatus"]
|
| 339 |
+
for field in fields_to_check:
|
| 340 |
+
if initial_lead.get(field) != final_lead.get(field):
|
| 341 |
+
return 0.0, f"Field '{field}' should not have changed, but it did"
|
| 342 |
+
|
| 343 |
+
# All checks passed
|
| 344 |
+
return 1.0, "Lead title successfully updated to 'Senior VP of Sales'"
|
| 345 |
+
|
| 346 |
+
except Exception as e:
|
| 347 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 348 |
+
|
| 349 |
+
|
| 350 |
+
def _validate_salesforce_create_and_close_case(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 351 |
+
"""
|
| 352 |
+
Validate that a case has been created and successfully closed.
|
| 353 |
+
|
| 354 |
+
This function checks:
|
| 355 |
+
1. The active tab is still a case details page
|
| 356 |
+
2. A case exists with subject 'Customer complaint resolved' and status 'Closed'
|
| 357 |
+
3. The case was not in the initial state (proves it was created)
|
| 358 |
+
|
| 359 |
+
Args:
|
| 360 |
+
initial_state: The initial state before the case creation and closure
|
| 361 |
+
final_state: The final state after the case creation and closure
|
| 362 |
+
|
| 363 |
+
Returns:
|
| 364 |
+
Tuple of (score, reason) where score is 1.0 for success, 0.0 for failure
|
| 365 |
+
"""
|
| 366 |
+
try:
|
| 367 |
+
# Check 1: Active tab should still be a case details page
|
| 368 |
+
active_tab_id = final_state.get("activeTabId")
|
| 369 |
+
if not active_tab_id or not active_tab_id.startswith("case-"):
|
| 370 |
+
return 0.0, f"Expected activeTabId to start with 'case-', got '{active_tab_id}'"
|
| 371 |
+
|
| 372 |
+
# Check 2: Case should exist with correct subject and be closed
|
| 373 |
+
cases = final_state.get("cases", [])
|
| 374 |
+
if not cases:
|
| 375 |
+
return 0.0, "No cases found in final state"
|
| 376 |
+
|
| 377 |
+
# Find the case with subject 'Customer complaint resolved'
|
| 378 |
+
complaint_case = None
|
| 379 |
+
for case in cases:
|
| 380 |
+
if case.get("subject") == "Customer complaint resolved":
|
| 381 |
+
complaint_case = case
|
| 382 |
+
break
|
| 383 |
+
|
| 384 |
+
if not complaint_case:
|
| 385 |
+
return 0.0, "Case with subject 'Customer complaint resolved' not found"
|
| 386 |
+
|
| 387 |
+
# Check case status
|
| 388 |
+
case_status = complaint_case.get("status")
|
| 389 |
+
if case_status != "Closed":
|
| 390 |
+
return 0.0, f"Expected case status to be 'Closed', got '{case_status}'"
|
| 391 |
+
|
| 392 |
+
# Check 3: Verify the case was created from scratch
|
| 393 |
+
initial_cases = initial_state.get("cases", [])
|
| 394 |
+
initial_case_exists = any(case.get("subject") == "Customer complaint resolved" for case in initial_cases)
|
| 395 |
+
if initial_case_exists:
|
| 396 |
+
return 0.0, "Case 'Customer complaint resolved' already existed in initial state - should be newly created"
|
| 397 |
+
|
| 398 |
+
# All checks passed
|
| 399 |
+
return 1.0, "Case 'Customer complaint resolved' successfully created and closed"
|
| 400 |
+
|
| 401 |
+
except Exception as e:
|
| 402 |
+
return 0.0, f"Error during validation: {str(e)}"
|
| 403 |
+
|
| 404 |
+
|
| 405 |
+
# Registry of Salesforce reward functions
|
| 406 |
+
REWARD_FUNCTIONS_SALESFORCE = {
|
| 407 |
+
"_validate_salesforce_convert_lead": _validate_salesforce_convert_lead,
|
| 408 |
+
"_validate_salesforce_create_and_convert_lead": _validate_salesforce_create_and_convert_lead,
|
| 409 |
+
"_validate_salesforce_create_new_lead": _validate_salesforce_create_new_lead,
|
| 410 |
+
"_validate_salesforce_create_new_case": _validate_salesforce_create_new_case,
|
| 411 |
+
"_validate_salesforce_edit_lead_inline": _validate_salesforce_edit_lead_inline,
|
| 412 |
+
"_validate_salesforce_create_and_close_case": _validate_salesforce_create_and_close_case,
|
| 413 |
+
}
|
| 414 |
+
|
| 415 |
+
__all__ = [
|
| 416 |
+
"REWARD_FUNCTIONS_SALESFORCE",
|
| 417 |
+
"_validate_salesforce_convert_lead",
|
| 418 |
+
"_validate_salesforce_create_and_convert_lead",
|
| 419 |
+
"_validate_salesforce_create_new_lead",
|
| 420 |
+
"_validate_salesforce_create_new_case",
|
| 421 |
+
"_validate_salesforce_edit_lead_inline",
|
| 422 |
+
"_validate_salesforce_create_and_close_case",
|
| 423 |
+
]
|
rewards/tic_tac_toe.py
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Reward functions for tic-tac-toe game tasks.
|
| 3 |
+
|
| 4 |
+
Board is a 9-element array representing a 3x3 grid:
|
| 5 |
+
[0, 1, 2,
|
| 6 |
+
3, 4, 5,
|
| 7 |
+
6, 7, 8]
|
| 8 |
+
|
| 9 |
+
Players: "X" (human), "O" (computer), or null (empty)
|
| 10 |
+
"""
|
| 11 |
+
|
| 12 |
+
import logging
|
| 13 |
+
from typing import Any, Dict, Tuple
|
| 14 |
+
|
| 15 |
+
logger = logging.getLogger(__name__)
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
def _validate_tic_tac_toe_make_first_move(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 19 |
+
"""Validate that a move was made on previously empty board."""
|
| 20 |
+
if "board" not in final_state:
|
| 21 |
+
return 0.0, "No board in final state"
|
| 22 |
+
|
| 23 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 24 |
+
|
| 25 |
+
board = final_state["board"]
|
| 26 |
+
|
| 27 |
+
# Check if board has any non-null values
|
| 28 |
+
if all(cell is None for cell in board):
|
| 29 |
+
return 0.0, "Board is still empty - no move was made."
|
| 30 |
+
|
| 31 |
+
# Check if board has at least one X
|
| 32 |
+
if "X" in board:
|
| 33 |
+
return 1.0, "First move made successfully - X placed on board."
|
| 34 |
+
|
| 35 |
+
return 0.0, "No X found on board."
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
def _validate_take_center(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 39 |
+
"""Validate that X was placed in center position (index 4)."""
|
| 40 |
+
if "board" not in final_state:
|
| 41 |
+
return 0.0, "No board in final state"
|
| 42 |
+
|
| 43 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 44 |
+
|
| 45 |
+
board = final_state["board"]
|
| 46 |
+
|
| 47 |
+
# Check if center position (index 4) contains X
|
| 48 |
+
if board[4] == "X":
|
| 49 |
+
return 1.0, "Center square successfully taken by X."
|
| 50 |
+
|
| 51 |
+
center_value = board[4] if board[4] is not None else "empty"
|
| 52 |
+
return 0.0, f"Center square not taken by X. Center is: {center_value}"
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
def _validate_block_computer_win(initial_state: Dict[str, Any], final_state: Dict[str, Any]) -> Tuple[float, str]:
|
| 56 |
+
"""Validate that X blocked computer's win at position 2."""
|
| 57 |
+
if "board" not in final_state:
|
| 58 |
+
return 0.0, "No board in final state"
|
| 59 |
+
|
| 60 |
+
logger.debug(f"running reward function on state: {final_state}")
|
| 61 |
+
|
| 62 |
+
board = final_state["board"]
|
| 63 |
+
|
| 64 |
+
# Check if position 2 (top-right) contains X
|
| 65 |
+
if board[2] != "X":
|
| 66 |
+
position_value = board[2] if board[2] is not None else "empty"
|
| 67 |
+
return 0.0, f"Position 2 not blocked by X. Position 2 is: {position_value}"
|
| 68 |
+
|
| 69 |
+
# Check that game is still ongoing (player hasn't lost)
|
| 70 |
+
winner = final_state.get("winner")
|
| 71 |
+
if winner == "O":
|
| 72 |
+
return 0.0, "Computer (O) won - blocking was unsuccessful."
|
| 73 |
+
|
| 74 |
+
return 1.0, "Successfully blocked computer's winning move at position 2."
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
# Registry of all tic-tac-toe reward functions
|
| 78 |
+
REWARD_FUNCTIONS_TIC_TAC_TOE = {
|
| 79 |
+
"_validate_tic_tac_toe_make_first_move": _validate_tic_tac_toe_make_first_move,
|
| 80 |
+
"_validate_take_center": _validate_take_center,
|
| 81 |
+
"_validate_block_computer_win": _validate_block_computer_win,
|
| 82 |
+
}
|