Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
json
Languages:
English
Size:
< 1K
ArXiv:
Tags:
benchmark
code-generation
transaction-code-generation
execution-grounded-evaluation
blockchain
evm
License:
| """ | |
| Generic Composite Validator for Multi-Round Interaction Problems | |
| This validator handles composite problems by: | |
| 1. Supporting multi-round interaction results | |
| 2. Validating error reports (if LLM detects issues) | |
| 3. Validating task completion (checking final chain state) | |
| 4. Checking parameter compliance (CRITICAL: no unauthorized modifications) | |
| 5. Modular scoring using atomic validator components | |
| """ | |
| import json | |
| from pathlib import Path | |
| from typing import Dict, List, Any, Optional | |
| from web3 import Web3 | |
| class CompositeValidator: | |
| """ | |
| Generic validator for all composite problems | |
| Validates composite problems by: | |
| - Checking if LLM reported an error (and if it's valid) | |
| - OR checking final chain state and parameter compliance | |
| - Using modular scoring components from atomic validators | |
| """ | |
| def __init__(self, agent_address: str = None, **kwargs): | |
| """ | |
| Initialize composite validator | |
| Args: | |
| agent_address: Agent's blockchain address (optional, can be set later) | |
| **kwargs: Additional parameters (ignored for compatibility) | |
| """ | |
| self.agent_address = Web3.to_checksum_address(agent_address) if agent_address else None | |
| self.composite_def = None | |
| self.scoring_components = [] | |
| def load_composite_definition(self, composite_id: str) -> Dict[str, Any]: | |
| """ | |
| Load composite problem definition from JSON | |
| Args: | |
| composite_id: Composite problem ID | |
| Returns: | |
| Composite problem definition dictionary | |
| """ | |
| # Try multiple possible paths | |
| possible_paths = [ | |
| Path(__file__).parent.parent / 'question_bank' / 'composite_problems' / f'{composite_id}.json', | |
| Path(__file__).parent.parent / 'question_bank' / 'composite_problems' / 'basic_workflows' / f'{composite_id}.json', | |
| ] | |
| for composite_path in possible_paths: | |
| if composite_path.exists(): | |
| with open(composite_path, 'r', encoding='utf-8') as f: | |
| self.composite_def = json.load(f) | |
| self._extract_scoring_components() | |
| return self.composite_def | |
| raise FileNotFoundError(f"Composite problem definition not found: {composite_id}") | |
| def _extract_scoring_components(self): | |
| """Extract scoring strategy from composite definition""" | |
| if not self.composite_def: | |
| return | |
| # scoring_strategy can be at top level OR nested in composite_structure | |
| # Check both locations for backward compatibility | |
| self.scoring_strategy = self.composite_def.get('scoring_strategy', {}) | |
| # If not found at root level, check inside composite_structure | |
| if not self.scoring_strategy: | |
| composite_structure = self.composite_def.get('composite_structure', {}) | |
| self.scoring_strategy = composite_structure.get('scoring_strategy', {}) | |
| def _get_param_value(self, param_name: str, default: Any = ''): | |
| """ | |
| Get parameter value from composite definition | |
| Args: | |
| param_name: Parameter name | |
| default: Default value if not found | |
| Returns: | |
| Parameter value (fixed or default) | |
| """ | |
| if not self.composite_def: | |
| return default | |
| params = self.composite_def.get('parameters', {}) | |
| param_config = params.get(param_name, {}) | |
| generation = param_config.get('generation', {}) | |
| # Check for fixed value | |
| if generation.get('method') == 'fixed': | |
| return generation.get('value', default) | |
| # Check for from_list (use first value) | |
| if generation.get('method') == 'from_list': | |
| addresses = generation.get('addresses', []) | |
| if addresses: | |
| return addresses[0] | |
| return default | |
| def validate( | |
| self, | |
| tx: Dict[str, Any] = None, | |
| receipt: Dict[str, Any] = None, | |
| state_before: Dict[str, Any] = None, | |
| state_after: Dict[str, Any] = None, | |
| # Legacy composite parameters (for future multi-turn support) | |
| final_submission: Dict[str, Any] = None, | |
| chain_state: Dict[str, Any] = None, | |
| task_params: Dict[str, Any] = None, | |
| interaction_history: List[Dict[str, Any]] = None | |
| ) -> Dict[str, Any]: | |
| """ | |
| Main validation entry point | |
| Compatible with both atomic validator interface and composite validator interface. | |
| Atomic interface (current implementation): | |
| - tx: Transaction object | |
| - receipt: Transaction receipt | |
| - state_before: State snapshot before transaction | |
| - state_after: State snapshot after transaction | |
| Composite interface (future multi-turn support): | |
| - final_submission: LLM's final submission (may contain error report) | |
| - chain_state: Final blockchain state | |
| - task_params: Original task parameters | |
| - interaction_history: Full history of all interactions | |
| Returns: | |
| Validation result with score and details | |
| """ | |
| # Set agent_address from transaction if not already set | |
| if not self.agent_address and tx: | |
| from eth_utils import to_checksum_address | |
| self.agent_address = to_checksum_address(tx.get('from', '')) | |
| print(f"\n{'='*70}") | |
| print(f"Validating Composite Problem: {self.composite_def.get('id', 'unknown')}") | |
| print(f"{'='*70}\n") | |
| # Detect which interface is being used | |
| if final_submission is not None: | |
| # Multi-turn interface: use new validation logic | |
| print("📋 Using multi-turn validation mode") | |
| return self._validate_multi_turn(final_submission, chain_state, task_params, interaction_history) | |
| else: | |
| # Atomic interface: use single-turn logic (for backwards compatibility) | |
| print("📋 Using single-turn (atomic) validation mode") | |
| return self._validate_task_completion_atomic(tx, receipt, state_before, state_after) | |
| def _validate_error_report( | |
| self, | |
| error_report: Dict[str, Any], | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Validate error report from LLM | |
| LLM gets 100 points if: | |
| - The reported error type is correct | |
| - The environment really has that error | |
| Args: | |
| error_report: LLM's error report | |
| chain_state: Current blockchain state | |
| task_params: Task parameters | |
| Returns: | |
| Validation result | |
| """ | |
| print("📋 LLM Reported an Error") | |
| print(f"{'-'*70}") | |
| error_type = error_report.get('error_type', 'UNKNOWN') | |
| error_message = error_report.get('error_message', '') | |
| print(f"Error Type: {error_type}") | |
| print(f"Error Message: {error_message}") | |
| print() | |
| # Verify if the error is valid | |
| is_valid_error = self._check_error_validity(error_type, chain_state, task_params) | |
| if is_valid_error: | |
| print("✅ Error Report is VALID") | |
| print(f" The environment indeed has this error.") | |
| score = 100.0 | |
| passed = True | |
| status = "error_correctly_reported" | |
| else: | |
| print("❌ Error Report is INVALID") | |
| print(f" The environment does NOT have this error (false alarm).") | |
| score = 0.0 | |
| passed = False | |
| status = "error_falsely_reported" | |
| print(f"\n{'='*70}") | |
| print(f"Final Score: {score:.2f}/100") | |
| print(f"Status: {status}") | |
| print(f"{'='*70}\n") | |
| return { | |
| 'score': score, | |
| 'max_score': 100.0, | |
| 'passed': passed, | |
| 'status': status, | |
| 'validation_mode': 'error_report', | |
| 'error_type': error_type, | |
| 'error_message': error_message, | |
| 'is_valid_error': is_valid_error, | |
| 'details': { | |
| 'reported_error': error_report, | |
| 'chain_state': chain_state | |
| } | |
| } | |
| def _check_error_validity( | |
| self, | |
| error_type: str, | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any] | |
| ) -> bool: | |
| """ | |
| Check if the reported error actually exists in the environment | |
| Args: | |
| error_type: Type of error reported | |
| chain_state: Current blockchain state | |
| task_params: Task parameters | |
| Returns: | |
| True if error is valid, False otherwise | |
| """ | |
| # Get required amount and actual balance | |
| required_amount = task_params.get('amount', 0) | |
| actual_balance = chain_state.get('initial_state', {}).get('token_balance', 0) | |
| if error_type == 'TOKEN_INSUFFICIENT_BALANCE': | |
| # Check if balance is really insufficient | |
| return actual_balance < required_amount | |
| elif error_type == 'BNB_INSUFFICIENT_BALANCE': | |
| bnb_balance = chain_state.get('initial_state', {}).get('bnb_balance', 0) | |
| required_bnb = task_params.get('amount', 0) | |
| return bnb_balance < required_bnb | |
| elif error_type in ['ALLOWANCE_INSUFFICIENT', 'NO_APPROVAL']: | |
| allowance = chain_state.get('initial_state', {}).get('allowance', 0) | |
| required_allowance = task_params.get('amount', 0) | |
| return allowance < required_allowance | |
| # Add more error type checks as needed | |
| # Unknown error type or not applicable | |
| return False | |
| def _validate_multi_turn( | |
| self, | |
| final_submission: Dict[str, Any], | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any], | |
| interaction_history: List[Dict[str, Any]] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Validate multi-turn interaction result | |
| Two possible outcomes: | |
| 1. LLM detected and reported an error (error_detected: true) | |
| 2. LLM completed the task successfully | |
| Args: | |
| final_submission: LLM's final submission | |
| chain_state: Final blockchain state | |
| task_params: Original task parameters | |
| interaction_history: Full interaction history | |
| Returns: | |
| Validation result with score | |
| """ | |
| print("🎯 Validating Multi-Turn Composite Task") | |
| print(f"{'-'*70}\n") | |
| # Check if LLM reported an error | |
| if final_submission.get('error_detected'): | |
| return self._validate_error_report(final_submission, chain_state, task_params) | |
| else: | |
| # LLM claims task is complete - validate from interaction history | |
| return self._validate_task_completion(chain_state, task_params, interaction_history) | |
| def _validate_task_completion_atomic( | |
| self, | |
| tx: Dict[str, Any], | |
| receipt: Dict[str, Any], | |
| state_before: Dict[str, Any], | |
| state_after: Dict[str, Any] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Validate task completion using atomic validator interface | |
| This is a simplified version for current single-transaction composite problems. | |
| It directly calls the key operation's atomic validator. | |
| Args: | |
| tx: Transaction object | |
| receipt: Transaction receipt | |
| state_before: State before transaction | |
| state_after: State after transaction | |
| Returns: | |
| Validation result | |
| """ | |
| print("🎯 Validating Composite Task (Atomic Interface)") | |
| print(f"{'-'*70}") | |
| # Get scoring strategy | |
| method = self.scoring_strategy.get('method', 'atomic_validator_reuse') | |
| key_op_id = self.scoring_strategy.get('key_operation_id') | |
| print(f"Key Operation: {key_op_id}") | |
| print(f"Strategy: {method}") | |
| print() | |
| # Load and call atomic validator | |
| atomic_validator = self._load_atomic_validator(key_op_id, tx) | |
| if not atomic_validator: | |
| print(f"❌ Failed to load atomic validator for '{key_op_id}'") | |
| return { | |
| 'score': 0.0, | |
| 'max_score': 100.0, | |
| 'passed': False, | |
| 'status': 'validator_load_failed', | |
| 'details': {'error': f'Validator not found: {key_op_id}'} | |
| } | |
| print(f"✅ Loaded atomic validator: {atomic_validator.__class__.__name__}") | |
| print() | |
| # Call atomic validator directly | |
| try: | |
| # The atomic validator expects: validate(tx, receipt, state_before, state_after) | |
| atomic_result = atomic_validator.validate( | |
| tx=tx, | |
| receipt=receipt, | |
| state_before=state_before, | |
| state_after=state_after | |
| ) | |
| score = atomic_result.get('score', 0.0) | |
| max_score = atomic_result.get('max_score', 100.0) | |
| passed = atomic_result.get('passed', False) | |
| print(f"📈 Atomic Validator Score: {score:.2f}/{max_score:.2f}") | |
| if 'checks' in atomic_result: | |
| print("\n Check Breakdown:") | |
| checks = atomic_result.get('checks', {}) | |
| # Checks can be either a dict or a list | |
| if isinstance(checks, dict): | |
| for check_name, check_result in checks.items(): | |
| status = '✅' if check_result.get('passed', False) else '❌' | |
| check_score = check_result.get('score', 0) | |
| print(f" {status} {check_name}: {check_score:.2f}") | |
| elif isinstance(checks, list): | |
| for idx, check_result in enumerate(checks): | |
| check_name = check_result.get('name', f'Check {idx+1}') | |
| status = '✅' if check_result.get('passed', False) else '❌' | |
| check_score = check_result.get('score', 0) | |
| print(f" {status} {check_name}: {check_score:.2f}") | |
| print(f"\n{'='*70}") | |
| print(f"Final Score: {score:.2f}/{max_score:.2f}") | |
| print(f"Status: {'✅ PASSED' if passed else '❌ FAILED'}") | |
| print(f"{'='*70}\n") | |
| return { | |
| 'score': score, | |
| 'max_score': max_score, | |
| 'passed': passed, | |
| 'status': 'completed', | |
| 'validation_mode': 'atomic_interface', | |
| 'checks': atomic_result.get('checks', {}), | |
| 'details': { | |
| 'atomic_validator': atomic_validator.__class__.__name__, | |
| 'atomic_result': atomic_result | |
| } | |
| } | |
| except Exception as e: | |
| print(f"❌ Error calling atomic validator: {e}") | |
| import traceback | |
| traceback.print_exc() | |
| return { | |
| 'score': 0.0, | |
| 'max_score': 100.0, | |
| 'passed': False, | |
| 'status': 'validation_error', | |
| 'details': {'error': str(e)} | |
| } | |
| def _validate_task_completion( | |
| self, | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any], | |
| interaction_history: List[Dict[str, Any]] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Validate task completion | |
| Validation logic: | |
| 1. Check if this is a pure query problem | |
| 2. For transaction problems: Check if a transaction was executed (chain state changed) | |
| 3. Base score = 100 if task completed, 0 otherwise | |
| 4. Step decay is applied by controller (optimal_steps / actual_steps) | |
| Args: | |
| chain_state: Final blockchain state | |
| task_params: Original task parameters | |
| interaction_history: All interactions | |
| Returns: | |
| Validation result | |
| """ | |
| print("🎯 LLM Submitted Task Completion") | |
| print(f"{'-'*70}\n") | |
| # Step 1: Check if this is a pure query problem | |
| if self._is_pure_query_problem(): | |
| return self._validate_pure_query_completion(task_params, interaction_history) | |
| # Step 2: Get atomic operations info | |
| composite_structure = self.composite_def.get('composite_structure', {}) | |
| atomic_operations = composite_structure.get('atomic_operations', []) | |
| key_op_id = self.scoring_strategy.get('key_operation_id') | |
| print(f"📋 Atomic Operations: {len(atomic_operations)}") | |
| print(f"🔑 Key Operation: {key_op_id or 'Not specified'}") | |
| print() | |
| # Step 3: Check if final transaction was executed | |
| # This validates that the chain state was modified (task completed) | |
| final_tx_hash = None | |
| if interaction_history: | |
| # Find the last transaction that was executed | |
| for rd in reversed(interaction_history): | |
| ar = rd.get('action_result', {}) | |
| if ar.get('tx_hash'): | |
| final_tx_hash = ar['tx_hash'] | |
| break | |
| # Step 4: Determine pass/fail based on transaction execution | |
| if final_tx_hash: | |
| print(f"✅ Transaction executed: {final_tx_hash[:20]}...") | |
| print(f" Chain state modified - task completed") | |
| base_score = 100.0 | |
| passed = True | |
| status = 'completed' | |
| message = 'Transaction executed successfully' | |
| else: | |
| print(f"❌ No transaction executed") | |
| print(f" Chain state unchanged - task not completed") | |
| base_score = 0.0 | |
| passed = False | |
| status = 'failed' | |
| message = 'No transaction was executed' | |
| print(f"\n{'='*70}") | |
| print(f"Base Score: {base_score:.2f}/100") | |
| print(f"Status: {'✅ PASSED' if passed else '❌ FAILED'}") | |
| print(f"{'='*70}\n") | |
| return { | |
| 'score': base_score, | |
| 'max_score': 100.0, | |
| 'passed': passed, | |
| 'status': status, | |
| 'validation_mode': 'transaction_execution', | |
| 'details': { | |
| 'tx_hash': final_tx_hash, | |
| 'message': message | |
| } | |
| } | |
| def _is_pure_query_problem(self) -> bool: | |
| """ | |
| Check if this composite problem consists only of query operations | |
| Returns: | |
| True if all atomic operations are queries, False otherwise | |
| """ | |
| if not self.composite_def: | |
| print("[DEBUG] _is_pure_query_problem: composite_def not loaded") | |
| return False | |
| composite_structure = self.composite_def.get('composite_structure', {}) | |
| atomic_operations = composite_structure.get('atomic_operations', []) | |
| if not atomic_operations: | |
| print("[DEBUG] _is_pure_query_problem: no atomic_operations found") | |
| return False | |
| # Check if all operations are query operations | |
| for op in atomic_operations: | |
| atomic_id = op.get('atomic_id', '') | |
| # Query operations typically start with 'query_' | |
| if not atomic_id.startswith('query_'): | |
| print(f"[DEBUG] _is_pure_query_problem: found non-query operation: {atomic_id}") | |
| return False | |
| print(f"[DEBUG] _is_pure_query_problem: All {len(atomic_operations)} operations are queries") | |
| return True | |
| def _validate_pure_query_completion( | |
| self, | |
| task_params: Dict[str, Any], | |
| interaction_history: List[Dict[str, Any]] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Validate completion of a pure query problem | |
| For pure query problems, we validate that: | |
| 1. LLM successfully executed all required queries | |
| 2. LLM returned the query results | |
| Note: LLM may batch multiple queries into a single execution. | |
| We check the returned data to count how many queries were completed. | |
| Args: | |
| task_params: Original task parameters | |
| interaction_history: All interactions | |
| Returns: | |
| Validation result | |
| """ | |
| print("📊 Validating Pure Query Problem") | |
| print(f"{'-'*70}\n") | |
| # Get expected number of queries from composite definition | |
| composite_structure = self.composite_def.get('composite_structure', {}) | |
| atomic_operations = composite_structure.get('atomic_operations', []) | |
| expected_queries = len(atomic_operations) | |
| # Count successful query operations | |
| successful_queries = 0 | |
| total_queries = 0 | |
| query_results = [] | |
| batched_query_count = 0 | |
| if interaction_history: | |
| for round_data in interaction_history: | |
| round_num = round_data.get('round', '?') | |
| action_result = round_data.get('action_result', {}) | |
| parsed_response = round_data.get('parsed_response', {}) | |
| action = parsed_response.get('action', '') | |
| # Check if this was a query action or successful execution | |
| if action_result and isinstance(action_result, dict): | |
| is_query = action_result.get('is_query', False) | |
| is_success = action_result.get('success', False) | |
| if is_query or is_success: | |
| total_queries += 1 | |
| if is_success: | |
| successful_queries += 1 | |
| print(f"✅ Round {round_num}: Query succeeded") | |
| query_results.append(action_result) | |
| # Check if this is a batched query result with multiple balances | |
| # Look for 'balances' dict, 'query_result', or 'data.portfolio' | |
| balances = action_result.get('balances', {}) | |
| # Check query_result field | |
| if not balances: | |
| query_result = action_result.get('query_result', {}) | |
| if isinstance(query_result, dict): | |
| balances = query_result.get('balances', {}) | |
| if not balances: | |
| data = query_result.get('data', {}) | |
| if isinstance(data, dict): | |
| balances = data.get('portfolio', data.get('balances', {})) | |
| # Check data.portfolio or data.balances | |
| if not balances: | |
| data = action_result.get('data', {}) | |
| if isinstance(data, dict): | |
| balances = data.get('portfolio', data.get('balances', {})) | |
| if balances and isinstance(balances, dict): | |
| batched_count = len(balances) | |
| if batched_count > 1: | |
| batched_query_count = batched_count | |
| print(f" 📦 Batched query detected: {batched_count} balances returned") | |
| for asset, info in balances.items(): | |
| if isinstance(info, dict): | |
| balance = info.get('formatted', info.get('balanceHuman', '?')) | |
| print(f" - {asset}: {balance}") | |
| else: | |
| print(f"❌ Round {round_num}: Query failed") | |
| print() | |
| # If batched queries detected, use that count instead | |
| if batched_query_count >= expected_queries: | |
| successful_queries = batched_query_count | |
| total_queries = batched_query_count | |
| print(f"📦 Batch query mode: {batched_query_count} queries completed in 1 round") | |
| print(f"📈 Query Statistics:") | |
| print(f" Expected queries: {expected_queries}") | |
| print(f" Executed queries: {total_queries}") | |
| print(f" Successful queries: {successful_queries}") | |
| print() | |
| # Score based on successful queries | |
| if expected_queries > 0: | |
| # Give full score if LLM completed the required queries | |
| # Note: LLM might batch them into fewer rounds | |
| if successful_queries >= expected_queries: | |
| score = 100.0 | |
| passed = True | |
| status = 'all_queries_completed' | |
| message = f'All {expected_queries} queries completed successfully' | |
| elif successful_queries > 0: | |
| # Partial credit based on completed queries | |
| score = (successful_queries / expected_queries) * 100.0 | |
| passed = successful_queries >= expected_queries * 0.5 # Pass if >= 50% queries done | |
| status = 'partial_queries_completed' | |
| message = f'{successful_queries}/{expected_queries} queries completed' | |
| else: | |
| score = 0.0 | |
| passed = False | |
| status = 'no_queries_completed' | |
| message = 'No queries were successfully completed' | |
| else: | |
| # No expected queries defined, just check if any were successful | |
| if successful_queries > 0: | |
| score = 100.0 | |
| passed = True | |
| status = 'queries_completed' | |
| message = f'{successful_queries} queries completed successfully' | |
| else: | |
| score = 0.0 | |
| passed = False | |
| status = 'no_queries_completed' | |
| message = 'No queries were completed' | |
| print(f"{'='*70}") | |
| print(f"Final Score: {score:.2f}/100") | |
| print(f"Status: {'✅ PASSED' if passed else '❌ FAILED'}") | |
| print(f"{'='*70}\n") | |
| return { | |
| 'score': score, | |
| 'max_score': 100.0, | |
| 'passed': passed, | |
| 'status': status, | |
| 'validation_mode': 'pure_query', | |
| 'details': { | |
| 'expected_queries': expected_queries, | |
| 'total_queries': total_queries, | |
| 'successful_queries': successful_queries, | |
| 'query_results': query_results, | |
| 'message': message | |
| } | |
| } | |
| def _check_parameter_compliance( | |
| self, | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Check if LLM strictly followed the required parameters | |
| This is CRITICAL: any parameter modification results in 0 score | |
| Args: | |
| chain_state: Final blockchain state | |
| task_params: Original task parameters | |
| Returns: | |
| Compliance check result | |
| """ | |
| print("🔍 Checking Parameter Compliance...") | |
| print(f"{'-'*70}") | |
| # Find the transfer transaction | |
| transfer_txs = [ | |
| tx for tx in chain_state.get('transactions', []) | |
| if tx.get('function_name') == 'transfer' | |
| ] | |
| if not transfer_txs: | |
| return { | |
| 'compliant': False, | |
| 'violation_reason': 'No transfer transaction found', | |
| 'details': {'expected': 'transfer transaction', 'actual': 'none'} | |
| } | |
| # Get the last transfer transaction (in case of multiple attempts) | |
| transfer_tx = transfer_txs[-1] | |
| # Check transfer amount | |
| required_amount = task_params.get('amount') | |
| actual_amount = transfer_tx.get('decoded_input', {}).get('amount') | |
| # Convert to same units for comparison (handle decimals) | |
| token_decimals = task_params.get('token_decimals', 18) | |
| required_amount_wei = int(required_amount * (10 ** token_decimals)) | |
| print(f"Required Amount: {required_amount} tokens ({required_amount_wei} wei)") | |
| print(f"Actual Amount: {actual_amount} wei") | |
| if actual_amount != required_amount_wei: | |
| return { | |
| 'compliant': False, | |
| 'violation_reason': f'Transfer amount mismatch: required {required_amount}, got {actual_amount / (10**token_decimals)}', | |
| 'details': { | |
| 'parameter': 'amount', | |
| 'required': required_amount_wei, | |
| 'actual': actual_amount | |
| } | |
| } | |
| # Check recipient address | |
| required_address = Web3.to_checksum_address(task_params.get('to_address')) | |
| actual_address = Web3.to_checksum_address(transfer_tx.get('decoded_input', {}).get('to', '')) | |
| print(f"Required Address: {required_address}") | |
| print(f"Actual Address: {actual_address}") | |
| if actual_address != required_address: | |
| return { | |
| 'compliant': False, | |
| 'violation_reason': f'Recipient address mismatch', | |
| 'details': { | |
| 'parameter': 'to_address', | |
| 'required': required_address, | |
| 'actual': actual_address | |
| } | |
| } | |
| return { | |
| 'compliant': True, | |
| 'details': { | |
| 'amount_check': 'passed', | |
| 'address_check': 'passed' | |
| } | |
| } | |
| def _score_key_operation( | |
| self, | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Score based on key operation by DIRECTLY REUSING atomic validator | |
| This method loads the atomic validator for the key operation and calls it directly, | |
| eliminating code duplication and ensuring consistency. | |
| Args: | |
| chain_state: Final blockchain state | |
| task_params: Task parameters | |
| Returns: | |
| Scoring result with breakdown | |
| """ | |
| print("\n📊 Scoring Key Operation...") | |
| print(f"{'-'*70}") | |
| # Get scoring strategy | |
| method = self.scoring_strategy.get('method', 'atomic_validator_reuse') | |
| if method != 'atomic_validator_reuse': | |
| print(f"⚠️ Warning: Unsupported scoring method '{method}', using atomic_validator_reuse") | |
| # Get key operation info | |
| key_op_id = self.scoring_strategy.get('key_operation_id') | |
| key_op_step = self.scoring_strategy.get('key_operation_step', 2) | |
| print(f"Key Operation: {key_op_id} (Step {key_op_step})") | |
| print(f"Strategy: Directly reuse atomic validator (zero code duplication)") | |
| print() | |
| # Load atomic validator | |
| atomic_validator = self._load_atomic_validator(key_op_id) | |
| if not atomic_validator: | |
| print(f"❌ Failed to load atomic validator for '{key_op_id}'") | |
| return { | |
| 'score': 0.0, | |
| 'breakdown': [], | |
| 'details': {'error': f'Validator not found: {key_op_id}'} | |
| } | |
| print(f"✅ Loaded atomic validator: {atomic_validator.__class__.__name__}") | |
| print() | |
| # Call atomic validator directly | |
| try: | |
| # The atomic validator expects: validate(transaction, initial_state, final_state) | |
| # We need to extract these from chain_state | |
| atomic_result = atomic_validator.validate( | |
| chain_state.get('key_operation_transaction', {}), | |
| chain_state.get('initial_state', {}), | |
| chain_state.get('final_state', {}) | |
| ) | |
| score = atomic_result.get('score', 0.0) | |
| details = atomic_result.get('details', {}) | |
| print(f"📈 Atomic Validator Score: {score:.2f}/100") | |
| if 'checks' in atomic_result: | |
| print("\n Check Breakdown:") | |
| for check_name, check_result in atomic_result.get('checks', {}).items(): | |
| status = '✅' if check_result.get('passed', False) else '❌' | |
| print(f" {status} {check_name}: {check_result.get('score', 0):.2f}") | |
| return { | |
| 'score': score, | |
| 'breakdown': atomic_result.get('checks', {}), | |
| 'details': { | |
| 'atomic_validator': atomic_validator.__class__.__name__, | |
| 'atomic_result': atomic_result | |
| } | |
| } | |
| except Exception as e: | |
| print(f"❌ Error calling atomic validator: {e}") | |
| import traceback | |
| traceback.print_exc() | |
| return { | |
| 'score': 0.0, | |
| 'breakdown': [], | |
| 'details': {'error': str(e)} | |
| } | |
| def _load_atomic_validator(self, atomic_id: str, tx: Dict[str, Any] = None): | |
| """ | |
| Load atomic validator by atomic problem ID | |
| Args: | |
| atomic_id: Atomic problem ID (e.g., 'erc20_transfer_fixed') | |
| tx: Transaction object (for extracting parameters) | |
| Returns: | |
| Atomic validator instance, or None if not found | |
| """ | |
| # Map atomic_id to validator class name | |
| # Format: atomic_id_to_class_name (e.g., erc20_transfer_fixed -> ERC20TransferValidator) | |
| validator_map = { | |
| 'erc20_transfer_fixed': 'ERC20TransferValidator', | |
| 'erc20_transfer_percentage': 'ERC20TransferPercentageValidator', | |
| 'erc20_transfer_max_amount': 'ERC20TransferMaxAmountValidator', | |
| 'bnb_transfer_basic': 'BNBTransferValidator', | |
| 'bnb_transfer_percentage': 'BNBTransferPercentageValidator', | |
| 'bnb_transfer_max_amount': 'BNBTransferMaxAmountValidator', | |
| 'erc20_approve': 'ERC20ApproveValidator', | |
| 'erc20_transferfrom_basic': 'ERC20TransferFromBasicValidator', | |
| 'erc721_transfer': 'ERC721TransferValidator', | |
| 'swap_exact_tokens_for_tokens': 'SwapExactTokensForTokensValidator', | |
| 'swap_exact_bnb_for_tokens': 'SwapExactBNBForTokensValidator', | |
| 'swap_exact_tokens_for_bnb': 'SwapExactTokensForBNBValidator', | |
| 'add_liquidity_bnb_token': 'AddLiquidityBNBTokenValidator', | |
| 'add_liquidity_tokens': 'AddLiquidityTokensValidator', | |
| 'stake_single_token': 'StakeSingleTokenValidator', | |
| 'stake_lp_tokens': 'StakeLPTokensValidator', | |
| 'unstake_lp_tokens': 'UnstakeLPTokensValidator', | |
| 'harvest_rewards': 'HarvestRewardsValidator', | |
| 'wbnb_deposit': 'WBNBDepositValidator', | |
| 'wbnb_withdraw': 'WBNBWithdrawValidator', | |
| 'remove_liquidity_tokens': 'RemoveLiquidityTokensValidator', | |
| 'remove_liquidity_bnb_token': 'RemoveLiquidityBNBTokenValidator', | |
| 'swap_exact_tokens_for_bnb': 'SwapExactTokensForBNBValidator', | |
| # Add more mappings as needed | |
| } | |
| validator_class_name = validator_map.get(atomic_id) | |
| if not validator_class_name: | |
| print(f"⚠️ Warning: No validator mapping for '{atomic_id}'") | |
| return None | |
| try: | |
| # Import from validators module | |
| from bsc_quest_bench import validators | |
| validator_class = getattr(validators, validator_class_name, None) | |
| if validator_class is None: | |
| print(f"⚠️ Warning: Validator class '{validator_class_name}' not found") | |
| return None | |
| # Instantiate validator with appropriate parameters | |
| # Different validators have different constructor signatures | |
| # For ERC20 transfers, we need to extract token_address, to_address and amount from tx | |
| if atomic_id == 'erc20_transfer_fixed': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Token address is the target of the transaction | |
| token_address = tx.get('to', '') | |
| # Decode transfer parameters from transaction data | |
| # transfer(address to, uint256 amount) - selector: 0xa9059cbb | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # 0x (2) + selector (8) + to (64) + amount (64) | |
| print(f"❌ Invalid transaction data length") | |
| return None | |
| # Extract to_address (bytes 4-36 after selector) | |
| to_address_hex = '0x' + tx_data[34:74] # Remove leading zeros | |
| from eth_utils import to_checksum_address | |
| to_address = to_checksum_address(to_address_hex) | |
| # Extract amount (bytes 36-68 after selector) | |
| amount_hex = tx_data[74:138] | |
| amount_wei = int(amount_hex, 16) | |
| amount_ether = amount_wei / (10**18) | |
| print(f" Decoded parameters:") | |
| print(f" - token_address: {token_address}") | |
| print(f" - to_address: {to_address}") | |
| print(f" - amount: {amount_ether} tokens ({amount_wei} wei)") | |
| return validator_class( | |
| token_address=token_address, | |
| to_address=to_address, | |
| amount=amount_ether, | |
| token_decimals=18 # Default to 18, can be made configurable later | |
| ) | |
| elif atomic_id == 'erc20_transfer_percentage': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Token address is the target of the transaction | |
| token_address = tx.get('to', '') | |
| # Decode transfer parameters from transaction data | |
| # transfer(address to, uint256 amount) - selector: 0xa9059cbb | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # 0x (2) + selector (8) + to (64) + amount (64) | |
| print(f"❌ Invalid transaction data length") | |
| return None | |
| # Extract to_address (bytes 4-36 after selector) | |
| to_address_hex = '0x' + tx_data[34:74] # Remove leading zeros | |
| from eth_utils import to_checksum_address | |
| to_address = to_checksum_address(to_address_hex) | |
| # Get percentage from task_params or default to 100% | |
| percentage = self._get_param_value('percentage') or 100 | |
| print(f" Decoded percentage transfer parameters:") | |
| print(f" - token_address: {token_address}") | |
| print(f" - to_address: {to_address}") | |
| print(f" - percentage: {percentage}%") | |
| return validator_class( | |
| token_address=token_address, | |
| to_address=to_address, | |
| percentage=int(percentage), | |
| token_decimals=18 | |
| ) | |
| elif atomic_id == 'erc20_transfer_max_amount': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Token address is the target of the transaction | |
| token_address = tx.get('to', '') | |
| # Decode transfer parameters from transaction data | |
| # transfer(address to, uint256 amount) - selector: 0xa9059cbb | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # 0x (2) + selector (8) + to (64) + amount (64) | |
| print(f"❌ Invalid transaction data length") | |
| return None | |
| # Extract to_address (bytes 4-36 after selector) | |
| to_address_hex = '0x' + tx_data[34:74] # Remove leading zeros | |
| from eth_utils import to_checksum_address | |
| to_address = to_checksum_address(to_address_hex) | |
| # Extract amount (bytes 36-68 after selector) | |
| amount_hex = tx_data[74:138] | |
| amount_wei = int(amount_hex, 16) | |
| amount_ether = amount_wei / (10**18) | |
| print(f" Decoded max amount transfer parameters:") | |
| print(f" - token_address: {token_address}") | |
| print(f" - to_address: {to_address}") | |
| print(f" - amount: {amount_ether} tokens ({amount_wei} wei)") | |
| return validator_class( | |
| token_address=token_address, | |
| to_address=to_address, | |
| amount=amount_ether, | |
| token_decimals=18 | |
| ) | |
| elif atomic_id == 'bnb_transfer_basic': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Extract to_address and amount from transaction | |
| to_address = tx.get('to', '') | |
| # Value is in Wei | |
| value_wei = tx.get('value', 0) | |
| if isinstance(value_wei, str): | |
| value_wei = int(value_wei, 16) if value_wei.startswith('0x') else int(value_wei) | |
| amount_bnb = value_wei / (10**18) | |
| print(f" Decoded BNB transfer parameters:") | |
| print(f" - to_address: {to_address}") | |
| print(f" - amount: {amount_bnb} BNB ({value_wei} wei)") | |
| return validator_class( | |
| to_address=to_address, | |
| amount=amount_bnb | |
| ) | |
| elif atomic_id == 'erc20_approve': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Token address is the target of the transaction | |
| token_address = tx.get('to', '') | |
| # Decode approve parameters from transaction data | |
| # approve(address spender, uint256 amount) - selector: 0x095ea7b3 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # 0x (2) + selector (8) + spender (64) + amount (64) | |
| print(f"❌ Invalid transaction data length") | |
| return None | |
| # Extract spender_address (bytes 4-36 after selector) | |
| spender_address_hex = '0x' + tx_data[34:74] # Remove leading zeros | |
| from eth_utils import to_checksum_address | |
| spender_address = to_checksum_address(spender_address_hex) | |
| # Extract amount (bytes 36-68 after selector) | |
| amount_hex = tx_data[74:138] | |
| amount_wei = int(amount_hex, 16) | |
| amount_ether = amount_wei / (10**18) | |
| print(f" Decoded approve parameters:") | |
| print(f" - token_address: {token_address}") | |
| print(f" - spender_address: {spender_address}") | |
| print(f" - amount: {amount_ether} tokens ({amount_wei} wei)") | |
| return validator_class( | |
| token_address=token_address, | |
| spender_address=spender_address, | |
| amount=amount_ether, | |
| agent_address=self.agent_address | |
| ) | |
| elif atomic_id == 'erc20_transferfrom_basic': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Token address is the target of the transaction | |
| token_address = tx.get('to', '') | |
| # Decode transferFrom parameters from transaction data | |
| # transferFrom(address from, address to, uint256 amount) - selector: 0x23b872dd | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 202: # 0x (2) + selector (8) + from (64) + to (64) + amount (64) | |
| print(f"❌ Invalid transaction data length for ERC20 transferFrom") | |
| return None | |
| # Extract from_address (bytes 4-36 after selector) | |
| from_address_hex = '0x' + tx_data[34:74] | |
| from eth_utils import to_checksum_address | |
| from_address = to_checksum_address(from_address_hex) | |
| # Extract to_address (bytes 36-68 after selector) | |
| to_address_hex = '0x' + tx_data[98:138] | |
| to_address = to_checksum_address(to_address_hex) | |
| # Extract amount (bytes 68-100 after selector) | |
| amount_hex = tx_data[138:202] | |
| amount_wei = int(amount_hex, 16) | |
| amount_tokens = amount_wei / (10**18) | |
| print(f" Decoded ERC20 transferFrom parameters:") | |
| print(f" - token_address: {token_address}") | |
| print(f" - from_address: {from_address}") | |
| print(f" - to_address: {to_address}") | |
| print(f" - amount: {amount_tokens} tokens ({amount_wei} wei)") | |
| print(f" - agent_address (spender): {self.agent_address}") | |
| return validator_class( | |
| token_address=token_address, | |
| from_address=from_address, | |
| to_address=to_address, | |
| amount=amount_tokens, | |
| agent_address=self.agent_address, | |
| token_decimals=18 | |
| ) | |
| elif atomic_id == 'erc721_transfer': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # NFT contract address is the target of the transaction | |
| nft_address = tx.get('to', '') | |
| # Decode transferFrom parameters from transaction data | |
| # transferFrom(address from, address to, uint256 tokenId) - selector: 0x23b872dd | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 202: # 0x (2) + selector (8) + from (64) + to (64) + tokenId (64) | |
| print(f"❌ Invalid transaction data length for transferFrom") | |
| return None | |
| # Extract from_address (bytes 4-36 after selector) - not needed, but for logging | |
| from_address_hex = '0x' + tx_data[34:74] | |
| from eth_utils import to_checksum_address | |
| from_address = to_checksum_address(from_address_hex) | |
| # Extract to_address (bytes 36-68 after selector) | |
| to_address_hex = '0x' + tx_data[98:138] | |
| to_address = to_checksum_address(to_address_hex) | |
| # Extract tokenId (bytes 68-100 after selector) | |
| token_id_hex = tx_data[138:202] | |
| token_id = int(token_id_hex, 16) | |
| print(f" Decoded NFT transfer parameters:") | |
| print(f" - nft_address: {nft_address}") | |
| print(f" - from_address: {from_address}") | |
| print(f" - to_address: {to_address}") | |
| print(f" - token_id: {token_id}") | |
| return validator_class( | |
| nft_address=nft_address, | |
| to_address=to_address, | |
| token_id=token_id | |
| ) | |
| elif atomic_id == 'stake_single_token': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Pool address is the target of the transaction | |
| pool_address = tx.get('to', '') | |
| # Decode deposit parameters from transaction data | |
| # deposit(uint256 _amount) - selector: 0xb6b55f25 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # 0x (2) + selector (8) + amount (64) | |
| print(f"❌ Invalid transaction data length for deposit") | |
| return None | |
| # Extract amount (bytes 4-36 after selector) | |
| amount_hex = tx_data[10:74] # Skip '0x' and 8 char selector | |
| amount_wei = int(amount_hex, 16) | |
| amount_tokens = amount_wei / (10**18) | |
| # Get token_address from composite definition | |
| # Try multiple parameter names for compatibility | |
| token_address = self._get_param_value('token_address') | |
| if not token_address: | |
| # Fallback to cake_address for CAKE staking scenarios | |
| token_address = self._get_param_value('cake_address') | |
| if not token_address: | |
| # Fallback to staking_token_address | |
| token_address = self._get_param_value('staking_token_address') | |
| print(f" Decoded staking parameters:") | |
| print(f" - pool_address: {pool_address}") | |
| print(f" - token_address: {token_address}") | |
| print(f" - stake_amount: {amount_tokens} tokens ({amount_wei} wei)") | |
| print(f" - user_address: {self.agent_address}") | |
| return validator_class( | |
| stake_amount=amount_tokens, | |
| token_address=token_address, | |
| pool_address=pool_address, | |
| user_address=self.agent_address | |
| ) | |
| elif atomic_id == 'stake_lp_tokens': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Pool address is the target of the transaction | |
| pool_address = tx.get('to', '') | |
| # Decode deposit parameters from transaction data | |
| # deposit(uint256 _amount) - selector: 0xb6b55f25 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # 0x (2) + selector (8) + amount (64) | |
| print(f"❌ Invalid transaction data length for LP staking deposit") | |
| return None | |
| # Extract amount (bytes 4-36 after selector) | |
| amount_hex = tx_data[10:74] # Skip '0x' and 8 char selector | |
| amount_wei = int(amount_hex, 16) | |
| amount_tokens = amount_wei / (10**18) | |
| # Get lp_token_address from composite definition | |
| lp_token_address = self._get_param_value('lp_token_address') | |
| print(f" Decoded LP staking parameters:") | |
| print(f" - pool_address: {pool_address}") | |
| print(f" - lp_token_address: {lp_token_address}") | |
| print(f" - stake_amount: {amount_tokens} LP tokens ({amount_wei} wei)") | |
| print(f" - user_address: {self.agent_address}") | |
| return validator_class( | |
| stake_amount=amount_tokens, | |
| lp_token_address=lp_token_address, | |
| pool_address=pool_address, | |
| user_address=self.agent_address | |
| ) | |
| elif atomic_id == 'swap_exact_bnb_for_tokens': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Get BNB amount from transaction value | |
| value_wei = tx.get('value', 0) | |
| if isinstance(value_wei, str): | |
| value_wei = int(value_wei, 16) if value_wei.startswith('0x') else int(value_wei) | |
| amount_in_bnb = value_wei / (10**18) | |
| # Get token_address and slippage from composite definition | |
| token_address = self._get_param_value('token_address') | |
| slippage = self._get_param_value('slippage', 5.0) | |
| print(f" Decoded swap parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_address: {token_address}") | |
| print(f" - amount_in: {amount_in_bnb} BNB ({value_wei} wei)") | |
| print(f" - slippage: {slippage}%") | |
| return validator_class( | |
| router_address=router_address, | |
| token_address=token_address, | |
| amount_in=amount_in_bnb, | |
| token_decimals=18, | |
| slippage=slippage | |
| ) | |
| elif atomic_id == 'wbnb_deposit': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # WBNB address is the target of the transaction | |
| wbnb_address = tx.get('to', '') | |
| # Get BNB amount from transaction value | |
| value_wei = tx.get('value', 0) | |
| if isinstance(value_wei, str): | |
| value_wei = int(value_wei, 16) if value_wei.startswith('0x') else int(value_wei) | |
| amount_bnb = value_wei / (10**18) | |
| print(f" Decoded WBNB deposit parameters:") | |
| print(f" - wbnb_address: {wbnb_address}") | |
| print(f" - amount: {amount_bnb} BNB ({value_wei} wei)") | |
| return validator_class( | |
| wbnb_address=wbnb_address, | |
| amount=amount_bnb | |
| ) | |
| elif atomic_id == 'wbnb_withdraw': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # WBNB address is the target of the transaction | |
| wbnb_address = tx.get('to', '') | |
| # Decode withdraw parameters from transaction data | |
| # withdraw(uint256 wad) - selector: 0x2e1a7d4d | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # 0x (2) + selector (8) + amount (64) | |
| print(f"❌ Invalid transaction data length for WBNB withdraw") | |
| return None | |
| # Extract amount (bytes 4-36 after selector) | |
| amount_hex = tx_data[10:74] # Skip '0x' and 8 char selector | |
| amount_wei = int(amount_hex, 16) | |
| amount_wbnb = amount_wei / (10**18) | |
| print(f" Decoded WBNB withdraw parameters:") | |
| print(f" - wbnb_address: {wbnb_address}") | |
| print(f" - amount: {amount_wbnb} WBNB ({amount_wei} wei)") | |
| return validator_class( | |
| wbnb_address=wbnb_address, | |
| amount=amount_wbnb | |
| ) | |
| elif atomic_id == 'harvest_rewards': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Pool address is the target of the transaction | |
| pool_address = tx.get('to', '') | |
| # Get reward_token_address from composite definition | |
| reward_token_address = self._get_param_value('reward_token_address') | |
| print(f" Decoded harvest parameters:") | |
| print(f" - pool_address: {pool_address}") | |
| print(f" - reward_token_address: {reward_token_address}") | |
| print(f" - user_address: {self.agent_address}") | |
| return validator_class( | |
| reward_token_address=reward_token_address, | |
| pool_address=pool_address, | |
| user_address=self.agent_address | |
| ) | |
| elif atomic_id == 'swap_exact_tokens_for_tokens': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Decode swapExactTokensForTokens parameters from transaction data | |
| # swapExactTokensForTokens(uint amountIn, uint amountOutMin, address[] calldata path, address to, uint deadline) | |
| # selector: 0x38ed1739 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # At least selector + amountIn | |
| print(f"❌ Invalid transaction data length for swap") | |
| return None | |
| # Extract amountIn (bytes 4-36 after selector) | |
| amount_in_hex = tx_data[10:74] # Skip '0x' and 8 char selector | |
| amount_in_wei = int(amount_in_hex, 16) | |
| # Get token addresses and other params from composite definition | |
| token_in_address = self._get_param_value('token_in_address') | |
| token_out_address = self._get_param_value('token_out_address') | |
| token_in_decimals = self._get_param_value('token_in_decimals', 18) | |
| token_out_decimals = self._get_param_value('token_out_decimals', 18) | |
| slippage = self._get_param_value('slippage', 5.0) | |
| amount_in_tokens = amount_in_wei / (10**token_in_decimals) | |
| print(f" Decoded swap parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_in_address: {token_in_address}") | |
| print(f" - token_out_address: {token_out_address}") | |
| print(f" - amount_in: {amount_in_tokens} tokens ({amount_in_wei} wei)") | |
| print(f" - slippage: {slippage}%") | |
| return validator_class( | |
| router_address=router_address, | |
| token_in_address=token_in_address, | |
| token_out_address=token_out_address, | |
| amount_in=amount_in_tokens, | |
| token_in_decimals=token_in_decimals, | |
| token_out_decimals=token_out_decimals, | |
| slippage=slippage | |
| ) | |
| elif atomic_id == 'add_liquidity_tokens': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Decode addLiquidity parameters from transaction data | |
| # addLiquidity(address tokenA, address tokenB, uint amountADesired, uint amountBDesired, uint amountAMin, uint amountBMin, address to, uint deadline) | |
| # selector: 0xe8e33700 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # At least selector + tokenA + tokenB | |
| print(f"❌ Invalid transaction data length for addLiquidity") | |
| return None | |
| # Extract amountADesired (bytes 68-100 after selector, 3rd parameter) | |
| # Skip: 0x (2) + selector (8) + tokenA (64) + tokenB (64) = 138 | |
| amount_a_hex = tx_data[138:202] # 64 chars for amountADesired | |
| amount_a_wei = int(amount_a_hex, 16) | |
| # Extract amountBDesired (bytes 100-132 after selector, 4th parameter) | |
| amount_b_hex = tx_data[202:266] # 64 chars for amountBDesired | |
| amount_b_wei = int(amount_b_hex, 16) | |
| # Get token addresses and decimals from composite definition | |
| token_a_address = self._get_param_value('token_a_address') | |
| token_b_address = self._get_param_value('token_b_address') | |
| token_a_decimals = self._get_param_value('token_a_decimals', 18) | |
| token_b_decimals = self._get_param_value('token_b_decimals', 18) | |
| amount_token_a = amount_a_wei / (10**token_a_decimals) | |
| amount_token_b = amount_b_wei / (10**token_b_decimals) | |
| print(f" Decoded add liquidity parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_a_address: {token_a_address}") | |
| print(f" - token_b_address: {token_b_address}") | |
| print(f" - amount_token_a: {amount_token_a} tokens ({amount_a_wei} wei)") | |
| print(f" - amount_token_b: {amount_token_b} tokens ({amount_b_wei} wei)") | |
| return validator_class( | |
| router_address=router_address, | |
| token_a_address=token_a_address, | |
| token_b_address=token_b_address, | |
| amount_token_a=amount_token_a, | |
| amount_token_b=amount_token_b, | |
| token_a_decimals=token_a_decimals, | |
| token_b_decimals=token_b_decimals | |
| ) | |
| elif atomic_id == 'remove_liquidity_tokens': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Decode removeLiquidity parameters from transaction data | |
| # removeLiquidity(address tokenA, address tokenB, uint liquidity, uint amountAMin, uint amountBMin, address to, uint deadline) | |
| # selector: 0xbaa2abde | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 138: # At least selector + tokenA + tokenB | |
| print(f"❌ Invalid transaction data length for removeLiquidity") | |
| return None | |
| # Get token addresses and percentage from composite definition | |
| token_a_address = self._get_param_value('token_a_address') | |
| token_b_address = self._get_param_value('token_b_address') | |
| liquidity_percentage = self._get_param_value('liquidity_percentage', 50.0) | |
| print(f" Decoded remove liquidity parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_a_address: {token_a_address}") | |
| print(f" - token_b_address: {token_b_address}") | |
| print(f" - liquidity_percentage: {liquidity_percentage}%") | |
| return validator_class( | |
| router_address=router_address, | |
| token_a_address=token_a_address, | |
| token_b_address=token_b_address, | |
| liquidity_percentage=liquidity_percentage | |
| ) | |
| elif atomic_id == 'remove_liquidity_bnb_token': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Decode removeLiquidityETH parameters from transaction data | |
| # removeLiquidityETH(address token, uint liquidity, uint amountTokenMin, uint amountETHMin, address to, uint deadline) | |
| # selector: 0x02751cec | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # At least selector + token | |
| print(f"❌ Invalid transaction data length for removeLiquidityETH") | |
| return None | |
| # Get token address and other params from composite definition | |
| token_address = self._get_param_value('token_address') | |
| token_decimals = self._get_param_value('token_decimals', 18) | |
| liquidity_percentage = self._get_param_value('liquidity_percentage', 50.0) | |
| slippage = self._get_param_value('slippage', 5.0) | |
| print(f" Decoded remove liquidity BNB parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_address: {token_address}") | |
| print(f" - liquidity_percentage: {liquidity_percentage}%") | |
| print(f" - slippage: {slippage}%") | |
| return validator_class( | |
| router_address=router_address, | |
| token_address=token_address, | |
| liquidity_percentage=liquidity_percentage, | |
| token_decimals=token_decimals, | |
| slippage=slippage | |
| ) | |
| elif atomic_id == 'swap_exact_tokens_for_bnb': | |
| if not tx: | |
| print(f"❌ Transaction object required for {atomic_id}") | |
| return None | |
| # Router address is the target of the transaction | |
| router_address = tx.get('to', '') | |
| # Decode swapExactTokensForETH parameters from transaction data | |
| # swapExactTokensForETH(uint amountIn, uint amountOutMin, address[] calldata path, address to, uint deadline) | |
| # selector: 0x18cbafe5 | |
| tx_data = tx.get('data', '') | |
| if len(tx_data) < 74: # At least selector + amountIn | |
| print(f"❌ Invalid transaction data length for swapExactTokensForETH") | |
| return None | |
| # Extract amountIn (bytes 4-36 after selector) | |
| amount_in_hex = tx_data[10:74] # Skip '0x' and 8 char selector | |
| amount_in_wei = int(amount_in_hex, 16) | |
| # Get token address and other params from composite definition | |
| token_address = self._get_param_value('token_address') | |
| token_decimals = self._get_param_value('token_decimals', 18) | |
| slippage = self._get_param_value('slippage', 5.0) | |
| amount_in_tokens = amount_in_wei / (10**token_decimals) | |
| print(f" Decoded swap to BNB parameters:") | |
| print(f" - router_address: {router_address}") | |
| print(f" - token_address: {token_address}") | |
| print(f" - amount_in: {amount_in_tokens} tokens ({amount_in_wei} wei)") | |
| print(f" - slippage: {slippage}%") | |
| return validator_class( | |
| router_address=router_address, | |
| token_address=token_address, | |
| amount_in=amount_in_tokens, | |
| token_decimals=token_decimals, | |
| slippage=slippage | |
| ) | |
| else: | |
| # Default: only pass agent_address | |
| return validator_class(self.agent_address) | |
| except Exception as e: | |
| print(f"❌ Error loading validator '{validator_class_name}': {e}") | |
| import traceback | |
| traceback.print_exc() | |
| return None | |
| def validate_composite( | |
| composite_id: str, | |
| agent_address: str, | |
| final_submission: Dict[str, Any], | |
| chain_state: Dict[str, Any], | |
| task_params: Dict[str, Any], | |
| interaction_history: List[Dict[str, Any]] | |
| ) -> Dict[str, Any]: | |
| """ | |
| Convenience function to validate a composite problem | |
| Args: | |
| composite_id: Composite problem ID | |
| agent_address: Agent's address | |
| final_submission: LLM's final submission | |
| chain_state: Final blockchain state | |
| task_params: Original task parameters | |
| interaction_history: Full interaction history | |
| Returns: | |
| Validation result | |
| """ | |
| validator = CompositeValidator(agent_address) | |
| validator.load_composite_definition(composite_id) | |
| return validator.validate(final_submission, chain_state, task_params, interaction_history) | |
| # Example usage | |
| if __name__ == "__main__": | |
| print("Composite Validator - Generic validator for multi-round interaction problems") | |
| print("\nFeatures:") | |
| print(" ✅ Multi-round interaction support") | |
| print(" ✅ Error report validation") | |
| print(" ✅ Parameter compliance checking (CRITICAL)") | |
| print(" ✅ Modular scoring using atomic components") | |
| print(" ✅ Normalized 100-point scoring") | |