File size: 14,510 Bytes
c922f8b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 |
"""
Reasoning tools for the GAIA agent.
This module provides tools for complex reasoning tasks, including:
- Step-by-step reasoning
- Mathematical calculations
- Fact verification
All tools handle errors gracefully and provide detailed error messages.
"""
import logging
import traceback
import re
import math
import json
from typing import Dict, Any, List, Optional, Union, Tuple, Callable
from datetime import datetime
import operator
from sympy import symbols, sympify, solve, simplify
import numpy as np
from src.gaia.agent.config import get_model_config, get_tool_config
from langchain_openai import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain_core.output_parsers import StrOutputParser
logger = logging.getLogger("gaia_agent.tools.reasoning")
class StepByStepReasoner:
"""Tool for performing step-by-step reasoning."""
def __init__(self, config: Optional[Dict[str, Any]] = None):
"""
Initialize the step-by-step reasoner.
Args:
config: Optional configuration dictionary
"""
self.model_config = config or get_model_config()
self.model = ChatOpenAI(
model=self.model_config.get("reasoning_model", "gpt-4o"),
temperature=self.model_config.get("temperature", 0.1),
max_tokens=self.model_config.get("max_tokens", 4096)
)
def reason(self, question: str, context: Optional[str] = None) -> Dict[str, Any]:
"""
Perform step-by-step reasoning on a question.
Args:
question: The question to reason about
context: Optional context information
Returns:
Dictionary containing the reasoning steps and conclusion
Raises:
Exception: If an error occurs during reasoning
"""
try:
prompt_template = """You are an expert at step-by-step reasoning.
Your task is to solve the following problem by breaking it down into clear, logical steps.
{context_text}
Question: {question}
Think through this step-by-step and provide your reasoning in the following JSON format:
{{
"steps": [
{{
"step_number": 1,
"reasoning": "First step of reasoning",
"intermediate_conclusion": "Conclusion from this step"
}},
...
],
"final_conclusion": "The final answer based on all steps",
"confidence": 0.95 // A number between 0 and 1 indicating your confidence
}}
JSON Response:"""
context_text = f"Context: {context}" if context else "No additional context provided."
prompt = PromptTemplate.from_template(prompt_template)
chain = prompt | self.model | StrOutputParser()
result = chain.invoke({
"question": question,
"context_text": context_text
})
parsed_result = json.loads(result)
return parsed_result
except Exception as e:
logger.error(f"Error during step-by-step reasoning: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Step-by-step reasoning failed: {str(e)}")
def chain_of_thought(self, question: str, context: Optional[str] = None) -> str:
"""
Perform chain-of-thought reasoning and return a textual response.
Args:
question: The question to reason about
context: Optional context information
Returns:
String containing the reasoning and conclusion
Raises:
Exception: If an error occurs during reasoning
"""
try:
prompt_template = """You are an expert at chain-of-thought reasoning.
Your task is to solve the following problem by thinking step-by-step.
{context_text}
Question: {question}
Let's think through this step-by-step:"""
context_text = f"Context: {context}" if context else "No additional context provided."
prompt = PromptTemplate.from_template(prompt_template)
chain = prompt | self.model | StrOutputParser()
result = chain.invoke({
"question": question,
"context_text": context_text
})
return result
except Exception as e:
logger.error(f"Error during chain-of-thought reasoning: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Chain-of-thought reasoning failed: {str(e)}")
class MathCalculator:
"""Tool for performing mathematical calculations."""
def __init__(self):
"""Initialize the math calculator."""
self.operations = {
'+': operator.add,
'-': operator.sub,
'*': operator.mul,
'/': operator.truediv,
'^': operator.pow,
'**': operator.pow,
'%': operator.mod,
'//': operator.floordiv
}
def calculate(self, expression: str) -> Dict[str, Any]:
"""
Evaluate a mathematical expression.
Args:
expression: The mathematical expression to evaluate
Returns:
Dictionary containing the result and steps
Raises:
Exception: If an error occurs during calculation
"""
try:
cleaned_expr = self._clean_expression(expression)
result = float(sympify(cleaned_expr).evalf())
if result.is_integer():
result = int(result)
return {
"expression": expression,
"result": result,
"steps": [f"Evaluated expression: {cleaned_expr}", f"Result: {result}"]
}
except Exception as e:
logger.error(f"Error calculating expression: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Mathematical calculation failed: {str(e)}")
def solve_equation(self, equation: str, variable: str = 'x') -> Dict[str, Any]:
"""
Solve a mathematical equation for a variable.
Args:
equation: The equation to solve (e.g., "x + 5 = 10")
variable: The variable to solve for (default: 'x')
Returns:
Dictionary containing the solution and steps
Raises:
Exception: If an error occurs during solving
"""
try:
if '=' in equation:
left_side, right_side = equation.split('=', 1)
equation_sympy = f"({left_side.strip()}) - ({right_side.strip()})"
else:
equation_sympy = equation
var = symbols(variable)
expr = sympify(equation_sympy)
solutions = solve(expr, var)
formatted_solutions = []
for sol in solutions:
if sol.is_real:
try:
float_sol = float(sol)
if float_sol.is_integer():
formatted_solutions.append(int(float_sol))
else:
formatted_solutions.append(float_sol)
except:
formatted_solutions.append(str(sol))
else:
formatted_solutions.append(str(sol))
return {
"equation": equation,
"variable": variable,
"solutions": formatted_solutions,
"steps": [
f"Parsed equation: {equation}",
f"Rearranged to: {equation_sympy} = 0",
f"Solved for {variable}",
f"Solutions: {', '.join(map(str, formatted_solutions))}"
]
}
except Exception as e:
logger.error(f"Error solving equation: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Equation solving failed: {str(e)}")
def _clean_expression(self, expression: str) -> str:
"""
Clean a mathematical expression for evaluation.
Args:
expression: The expression to clean
Returns:
Cleaned expression
"""
expression = expression.strip()
replacements = {
'sin': 'math.sin',
'cos': 'math.cos',
'tan': 'math.tan',
'log': 'math.log',
'exp': 'math.exp',
'sqrt': 'math.sqrt',
'pi': 'math.pi',
'e': 'math.e'
}
for old, new in replacements.items():
expression = re.sub(r'\b' + old + r'\b', new, expression)
return expression
class FactVerifier:
"""Tool for verifying facts and claims."""
def __init__(self, config: Optional[Dict[str, Any]] = None):
"""
Initialize the fact verifier.
Args:
config: Optional configuration dictionary
"""
self.model_config = config or get_model_config()
self.model = ChatOpenAI(
model=self.model_config.get("reasoning_model", "gpt-4o"),
temperature=self.model_config.get("temperature", 0.1),
max_tokens=self.model_config.get("max_tokens", 4096)
)
def verify(self, claim: str, evidence: Optional[str] = None) -> Dict[str, Any]:
"""
Verify a claim against evidence.
Args:
claim: The claim to verify
evidence: Optional evidence to use for verification
Returns:
Dictionary containing verification results
Raises:
Exception: If an error occurs during verification
"""
try:
prompt_template = """You are an expert fact checker.
Your task is to verify the following claim based on the provided evidence.
Claim: {claim}
Evidence: {evidence}
Analyze the claim and evidence carefully. Provide your verification in the following JSON format:
{{
"is_verifiable": true/false, // Whether the claim can be verified with the given evidence
"verification_result": "supported"/"contradicted"/"insufficient_evidence",
"confidence": 0.95, // A number between 0 and 1 indicating your confidence
"reasoning": "Your step-by-step reasoning process",
"missing_information": "What additional information would help verify this claim"
}}
JSON Response:"""
evidence_text = evidence if evidence else "No evidence provided."
prompt = PromptTemplate.from_template(prompt_template)
chain = prompt | self.model | StrOutputParser()
result = chain.invoke({
"claim": claim,
"evidence": evidence_text
})
parsed_result = json.loads(result)
return parsed_result
except Exception as e:
logger.error(f"Error during fact verification: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Fact verification failed: {str(e)}")
def compare_claims(self, claim1: str, claim2: str) -> Dict[str, Any]:
"""
Compare two claims for consistency.
Args:
claim1: The first claim
claim2: The second claim
Returns:
Dictionary containing comparison results
Raises:
Exception: If an error occurs during comparison
"""
try:
prompt_template = """You are an expert at analyzing logical consistency.
Your task is to compare the following two claims and determine if they are consistent, contradictory, or unrelated.
Claim 1: {claim1}
Claim 2: {claim2}
Analyze the claims carefully. Provide your analysis in the following JSON format:
{{
"relationship": "consistent"/"contradictory"/"unrelated",
"confidence": 0.95, // A number between 0 and 1 indicating your confidence
"reasoning": "Your step-by-step reasoning process",
"implications": "What the relationship between these claims implies"
}}
JSON Response:"""
prompt = PromptTemplate.from_template(prompt_template)
chain = prompt | self.model | StrOutputParser()
result = chain.invoke({
"claim1": claim1,
"claim2": claim2
})
parsed_result = json.loads(result)
return parsed_result
except Exception as e:
logger.error(f"Error during claim comparison: {str(e)}")
logger.error(traceback.format_exc())
raise Exception(f"Claim comparison failed: {str(e)}")
def create_step_by_step_reasoner() -> StepByStepReasoner:
"""Create a step-by-step reasoner instance."""
return StepByStepReasoner()
def create_math_calculator() -> MathCalculator:
"""Create a math calculator instance."""
return MathCalculator()
def create_fact_verifier() -> FactVerifier:
"""Create a fact verifier instance."""
return FactVerifier() |