Spaces:
Running
Running
File size: 9,686 Bytes
226ac39 |
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 |
"""
Reasoning Module - Core Abstraction
Provides clean separation between:
- Deterministic data processing (tools)
- Non-deterministic reasoning (LLM)
Design Principles:
- NO RAW DATA ACCESS - Only summaries/metadata
- NO TRAINING DECISIONS - Only explanations
- STRUCTURED I/O - JSON in, JSON + text out
- CACHEABLE - Deterministic enough to cache
- REASONING ONLY - No execution, no side effects
Architecture:
Tool → Generates Summary → Reasoning Module → Returns Explanation
Tool: "Here's what I found: {stats}"
Reasoning: "Based on these stats, this means..."
Usage:
from reasoning import get_reasoner
reasoner = get_reasoner()
result = reasoner.explain_data(
summary={"rows": 1000, "columns": 20, "missing": 50}
)
"""
import os
from typing import Dict, Any, Optional, Union
from abc import ABC, abstractmethod
class ReasoningBackend(ABC):
"""Abstract base class for reasoning backends."""
@abstractmethod
def generate(
self,
prompt: str,
system_prompt: Optional[str] = None,
temperature: float = 0.1,
max_tokens: int = 2048
) -> str:
"""Generate reasoning response."""
pass
@abstractmethod
def generate_structured(
self,
prompt: str,
schema: Dict[str, Any],
system_prompt: Optional[str] = None
) -> Dict[str, Any]:
"""Generate structured JSON response."""
pass
class GeminiBackend(ReasoningBackend):
"""Gemini reasoning backend."""
def __init__(self, api_key: Optional[str] = None, model: str = "gemini-2.0-flash-exp"):
try:
import google.generativeai as genai
except ImportError:
raise ImportError(
"google-generativeai not installed. "
"Install with: pip install google-generativeai"
)
api_key = api_key or os.getenv("GOOGLE_API_KEY")
if not api_key:
raise ValueError(
"Google API key required. Set GOOGLE_API_KEY env var or pass api_key"
)
genai.configure(api_key=api_key)
self.model = genai.GenerativeModel(
model,
generation_config={"temperature": 0.1}
)
self.model_name = model
def generate(
self,
prompt: str,
system_prompt: Optional[str] = None,
temperature: float = 0.1,
max_tokens: int = 2048
) -> str:
"""Generate reasoning response."""
# Combine system and user prompts
full_prompt = prompt
if system_prompt:
full_prompt = f"{system_prompt}\n\n{prompt}"
response = self.model.generate_content(
full_prompt,
generation_config={
"temperature": temperature,
"max_output_tokens": max_tokens
}
)
return response.text
def generate_structured(
self,
prompt: str,
schema: Dict[str, Any],
system_prompt: Optional[str] = None
) -> Dict[str, Any]:
"""Generate structured JSON response."""
import json
# Add schema instruction
schema_str = json.dumps(schema, indent=2)
structured_prompt = f"""{prompt}
Respond with valid JSON matching this schema:
{schema_str}
Your response must be valid JSON only, no other text."""
response_text = self.generate(structured_prompt, system_prompt)
# Extract JSON from response
try:
# Try direct parse
return json.loads(response_text)
except json.JSONDecodeError:
# Try to extract JSON from markdown code blocks
import re
json_match = re.search(r'```json\s*\n(.*?)\n```', response_text, re.DOTALL)
if json_match:
return json.loads(json_match.group(1))
# Try to extract any JSON object
json_match = re.search(r'\{.*\}', response_text, re.DOTALL)
if json_match:
return json.loads(json_match.group(0))
raise ValueError(f"Failed to extract JSON from response: {response_text[:200]}...")
class GroqBackend(ReasoningBackend):
"""Groq reasoning backend."""
def __init__(self, api_key: Optional[str] = None, model: str = "llama-3.3-70b-versatile"):
try:
from groq import Groq
except ImportError:
raise ImportError(
"groq not installed. "
"Install with: pip install groq"
)
api_key = api_key or os.getenv("GROQ_API_KEY")
if not api_key:
raise ValueError(
"Groq API key required. Set GROQ_API_KEY env var or pass api_key"
)
self.client = Groq(api_key=api_key)
self.model_name = model
def generate(
self,
prompt: str,
system_prompt: Optional[str] = None,
temperature: float = 0.1,
max_tokens: int = 2048
) -> str:
"""Generate reasoning response."""
messages = []
if system_prompt:
messages.append({"role": "system", "content": system_prompt})
messages.append({"role": "user", "content": prompt})
response = self.client.chat.completions.create(
model=self.model_name,
messages=messages,
temperature=temperature,
max_tokens=max_tokens
)
return response.choices[0].message.content
def generate_structured(
self,
prompt: str,
schema: Dict[str, Any],
system_prompt: Optional[str] = None
) -> Dict[str, Any]:
"""Generate structured JSON response."""
import json
# Add schema instruction
schema_str = json.dumps(schema, indent=2)
structured_prompt = f"""{prompt}
Respond with valid JSON matching this schema:
{schema_str}
Your response must be valid JSON only, no other text."""
response_text = self.generate(structured_prompt, system_prompt)
# Extract JSON from response
try:
return json.loads(response_text)
except json.JSONDecodeError:
# Try to extract JSON from markdown code blocks
import re
json_match = re.search(r'```json\s*\n(.*?)\n```', response_text, re.DOTALL)
if json_match:
return json.loads(json_match.group(1))
# Try to extract any JSON object
json_match = re.search(r'\{.*\}', response_text, re.DOTALL)
if json_match:
return json.loads(json_match.group(0))
raise ValueError(f"Failed to extract JSON from response: {response_text[:200]}...")
class ReasoningEngine:
"""
Main reasoning engine.
Delegates to appropriate backend (Gemini, Groq, etc).
Provides high-level reasoning capabilities.
"""
def __init__(
self,
backend: Optional[ReasoningBackend] = None,
provider: str = "gemini"
):
"""
Initialize reasoning engine.
Args:
backend: Custom backend instance
provider: 'gemini' or 'groq' (if backend not provided)
"""
if backend:
self.backend = backend
else:
provider = provider or os.getenv("LLM_PROVIDER", "gemini")
if provider == "gemini":
self.backend = GeminiBackend()
elif provider == "groq":
self.backend = GroqBackend()
else:
raise ValueError(f"Unsupported provider: {provider}")
self.provider = provider
def reason(
self,
prompt: str,
system_prompt: Optional[str] = None,
temperature: float = 0.1
) -> str:
"""
General-purpose reasoning.
Args:
prompt: User prompt
system_prompt: Optional system context
temperature: Creativity (0.0 = deterministic, 1.0 = creative)
Returns:
Natural language response
"""
return self.backend.generate(prompt, system_prompt, temperature)
def reason_structured(
self,
prompt: str,
schema: Dict[str, Any],
system_prompt: Optional[str] = None
) -> Dict[str, Any]:
"""
Structured reasoning with JSON output.
Args:
prompt: User prompt
schema: Expected JSON schema
system_prompt: Optional system context
Returns:
Parsed JSON response
"""
return self.backend.generate_structured(prompt, schema, system_prompt)
# Singleton instance
_reasoning_engine: Optional[ReasoningEngine] = None
def get_reasoner(
backend: Optional[ReasoningBackend] = None,
provider: Optional[str] = None
) -> ReasoningEngine:
"""
Get singleton reasoning engine.
Args:
backend: Custom backend instance
provider: 'gemini' or 'groq'
Returns:
ReasoningEngine instance
"""
global _reasoning_engine
if _reasoning_engine is None or backend is not None:
_reasoning_engine = ReasoningEngine(backend=backend, provider=provider)
return _reasoning_engine
def reset_reasoner():
"""Reset singleton (for testing)."""
global _reasoning_engine
_reasoning_engine = None
|