synth-ai 0.2.16__py3-none-any.whl → 0.2.19__py3-none-any.whl

synth_ai/cli/train/__init__.py

@@ -0,0 +1,12 @@
+from __future__ import annotations
+
+from .core import register, train_command
+from .errors import TrainCliError
+from .validation import validate_train_environment
+
+__all__ = [
+    "register",
+    "train_command",
+    "TrainCliError",
+    "validate_train_environment",
+]

synth_ai/cli/train/core.py

@@ -0,0 +1,21 @@
+from __future__ import annotations
+
+import click
+from synth_ai.api.train.cli import (
+    register as _register_with_cli,
+)
+from synth_ai.api.train.cli import (
+    train_command as _train_command,
+)
+
+__all__ = ["register", "train_command"]
+
+
+def register(cli: click.Group) -> None:
+    """Attach the train command to the root CLI."""
+    _register_with_cli(cli)
+
+
+def train_command(*args, **kwargs):
+    """Entrypoint used by the train CLI command."""
+    return _train_command(*args, **kwargs)

synth_ai/cli/train/errors.py

@@ -0,0 +1,8 @@
+from __future__ import annotations
+
+
+class TrainCliError(RuntimeError):
+    """Base exception for train CLI failures."""
+
+
+__all__ = ["TrainCliError"]

synth_ai/cli/train/validation.py

@@ -0,0 +1,24 @@
+from __future__ import annotations
+
+from collections.abc import Iterable
+from pathlib import Path
+from typing import Dict, Tuple
+
+from synth_ai.api.train.env_resolver import KeySpec, resolve_env
+
+__all__ = ["validate_train_environment"]
+
+
+def validate_train_environment(
+    *,
+    config_path: Path | None,
+    explicit_env_paths: Iterable[str],
+    required_keys: list[KeySpec],
+) -> Tuple[Path, Dict[str, str]]:
+    """Validate and resolve environment secrets used by the train command."""
+    resolved_path, resolved_keys = resolve_env(
+        config_path=config_path,
+        explicit_env_paths=explicit_env_paths,
+        required_keys=required_keys,
+    )
+    return resolved_path, resolved_keys

synth_ai/cli/train.py

@@ -1,18 +1,5 @@
 from __future__ import annotations
 
-from typing import Any
-
-from synth_ai.api.train.cli import register as _register
-from synth_ai.api.train.cli import train_command as _train_command
+from synth_ai.cli.commands.train.core import register, train_command
 
 __all__ = ["register", "train_command"]
-
-
-def register(cli: Any) -> None:
-    """Compatibility wrapper for the legacy train CLI location."""
-
-    _register(cli)
-
-
-def train_command(*args: Any, **kwargs: Any) -> Any:
-    return _train_command(*args, **kwargs)

synth_ai/demos/crafter/grpo_crafter_task_app.py

@@ -3,7 +3,7 @@
 This module now delegates to the TaskAppConfig defined in the local example at
 `examples/warming_up_to_rl/task_app/grpo_crafter.py`. It is kept for legacy usage
 (running the file directly or targeting `fastapi_app` from external tooling).
-Prefer using `uvx synth-ai serve grpo-crafter` for local development and testing.
+Prefer using `uvx synth-ai deploy --runtime uvicorn grpo-crafter` for local development and testing.
 """
 
 from __future__ import annotations

synth_ai/demos/demo_task_apps/crafter/grpo_crafter_task_app.py

@@ -3,7 +3,7 @@
 This module now delegates to the TaskAppConfig defined in the local example at
 `examples/task_apps/crafter/task_app/grpo_crafter.py`. It is kept for legacy usage
 (running the file directly or targeting `fastapi_app` from external tooling).
-Prefer using `uvx synth-ai serve grpo-crafter` for local development and testing.
+Prefer using `uvx synth-ai deploy --runtime uvicorn grpo-crafter` for local development and testing.
 """
 
 from __future__ import annotations

synth_ai/environments/examples/crafter_classic/engine_deterministic_patch.py

@@ -6,16 +6,18 @@ the hash-based set-iteration nondeterminism that caused the drift.
 """
 
 import collections
+import os
 
 import crafter
 
-print("[PATCH] Attempting to apply Crafter deterministic patch...")
+# Patch messages permanently disabled to reduce noise
+# print("[PATCH] Attempting to apply Crafter deterministic patch...")
 
 # -----------------------------------------------------------------------------
 # 1.  Make per–chunk object order stable
 # -----------------------------------------------------------------------------
 if not hasattr(crafter.Env, "_orig_balance_object"):
-    print("[PATCH] Patching crafter.Env._balance_object...")
+    # print("[PATCH] Patching crafter.Env._balance_object...")
     crafter.Env._orig_balance_object = crafter.Env._balance_object
 
     def _balance_object_det(self, chunk, objs, *args, **kwargs):
@@ -25,9 +27,10 @@ if not hasattr(crafter.Env, "_orig_balance_object"):
         return crafter.Env._orig_balance_object(self, chunk, objs, *args, **kwargs)
 
     crafter.Env._balance_object = _balance_object_det
-    print("[PATCH] crafter.Env._balance_object patched.")
+    # print("[PATCH] crafter.Env._balance_object patched.")
 else:
-    print("[PATCH] crafter.Env._balance_object already patched or _orig_balance_object exists.")
+    pass
+    # print("[PATCH] crafter.Env._balance_object already patched or _orig_balance_object exists.")
 
 # -----------------------------------------------------------------------------
 # 2.  Make *chunk* iteration order stable

synth_ai/environments/examples/crafter_classic/engine_serialization_patch_v3.py

@@ -4,6 +4,7 @@ This version handles player references for Zombie and Skeleton objects.
 """
 
 import collections
+import os
 import pickle
 from typing import Any, Dict, Optional, Set
 
@@ -11,11 +12,12 @@ import crafter
 import numpy as np
 from crafter import objects
 
-print("[PATCH] Attempting to apply Crafter serialization patch v3...")
+# Patch messages permanently disabled
+# print("[PATCH] Attempting to apply Crafter serialization patch v3...")
 
 # Check if already patched
 if not hasattr(crafter.Env, "save"):
-    print("[PATCH] Adding enhanced save/load methods to crafter.Env...")
+    # print("[PATCH] Adding enhanced save/load methods to crafter.Env...")
 
     def _save(self) -> Dict[str, Any]:
         """Save complete environment state including all details."""
@@ -260,8 +262,10 @@ if not hasattr(crafter.Env, "save"):
     crafter.Env.save = _save
     crafter.Env.load = _load
 
-    print("[PATCH] crafter.Env.save() and load() methods added (v3).")
+    pass
+    # print("[PATCH] crafter.Env.save() and load() methods added (v3).")
 else:
-    print("[PATCH] crafter.Env already has save/load methods.")
+    pass
+    # print("[PATCH] crafter.Env already has save/load methods.")
 
-print("[PATCH] Crafter serialization patch v3 complete.")
+# print("[PATCH] Crafter serialization patch v3 complete.")

synth_ai/environments/examples/crafter_classic/world_config_patch_simple.py

@@ -9,7 +9,8 @@ from typing import Any, Dict, Optional
 
 import crafter
 
-print("[PATCH] Attempting to apply simplified Crafter world configuration patch...")
+# Patch messages permanently disabled
+# print("[PATCH] Attempting to apply simplified Crafter world configuration patch...")
 
 # World configuration presets
 WORLD_CONFIGS = {
@@ -279,8 +280,8 @@ def patched_env_init(
 
 crafter.Env.__init__ = patched_env_init
 
-print("[PATCH] Simplified Crafter world configuration patch complete.")
-print("[PATCH] Available configs: easy, normal, hard, peaceful")
+# print("[PATCH] Simplified Crafter world configuration patch complete.")
+# print("[PATCH] Available configs: easy, normal, hard, peaceful")
 
 # Example custom config
 EXAMPLE_CUSTOM_CONFIG = {

synth_ai/environments/examples/red/engine.py

@@ -14,12 +14,15 @@ from synth_ai.environments.stateful.engine import StatefulEngine, StatefulEngine
 from synth_ai.environments.tasks.core import TaskInstance
 
 from .engine_helpers.reward_components import (
-    BadgeRewardComponent,
-    BattleVictoryComponent,
-    LevelUpComponent,
-    MapTransitionComponent,
+    RouteExplorationReward,
+    StrategicTrainingReward,
+    BattleProgressionReward,
+    GymPreparationReward,
+    ItemCollectionReward,
+    HealingManagementReward,
+    EfficientExplorationReward,
+    BadgeVictoryReward,
     StepPenaltyComponent,
-    XPGainComponent,
 )
 from .engine_helpers.state_extraction import extract_game_state
 
@@ -268,15 +271,27 @@ class PokemonRedEngine(StatefulEngine, IReproducibleEngine):
             # For testing purposes, use None emulator
             self.emulator = None
 
-        # Initialize reward stack with dense components
+        # Initialize reward stack with comprehensive progress-based components
         self.reward_stack = RewardStack(
             components=[
-                BadgeRewardComponent(),
-                MapTransitionComponent(),
-                BattleVictoryComponent(),
-                LevelUpComponent(),
-                XPGainComponent(),
-                StepPenaltyComponent(),
+                # Major progress rewards
+                BadgeVictoryReward(),        # +50.0 for Boulder Badge (main goal)
+                RouteExplorationReward(),    # +1.0-5.0 for reaching key areas
+                GymPreparationReward(),      # +3.0 for being gym-ready
+
+                # Training and battle rewards
+                StrategicTrainingReward(),   # +0.2-3.0 for level ups and milestones
+                BattleProgressionReward(),   # +0.1-1.0 for battles
+
+                # Resource management rewards
+                ItemCollectionReward(),      # +0.1-0.5 for collecting items
+                HealingManagementReward(),   # +0.05-0.8 for healing Pokemon
+
+                # Exploration efficiency
+                EfficientExplorationReward(), # +0.02 for discovering new positions
+
+                        # No penalty for unproductive actions
+                        StepPenaltyComponent(penalty=0.0),        # 0.0 per step
             ]
         )
 
@@ -640,6 +655,12 @@ class PokemonRedEngine(StatefulEngine, IReproducibleEngine):
                         "prev_text_box_active": bool(prev_state.get("text_box_active", False)),
                         "prev_enemy_hp_current": int(prev_state.get("enemy_hp_current", 0)),
                         "prev_enemy_hp_percentage": float(prev_state.get("enemy_hp_percentage", 0.0)),
+                        "prev_player_x": int(prev_state.get("player_x", 0)),
+                        "prev_player_y": int(prev_state.get("player_y", 0)),
+                        "prev_party": prev_state.get("party", []),
+                        "prev_inventory": prev_state.get("inventory", []),
+                        "prev_party_hp_current": int(prev_state.get("party_hp_current", 0)),
+                        "prev_party_hp_max": int(prev_state.get("party_hp_max", 0)),
                     },
                 )
             except Exception as e: