CBCFacil v8.0 - Refactored with AMD GPU support

This commit is contained in:
2026-01-09 13:05:46 -03:00
parent cb17136f21
commit b017504c52
54 changed files with 7251 additions and 3670 deletions

15
services/ai/__init__.py Normal file
View File

@@ -0,0 +1,15 @@
"""
AI Providers package for CBCFacil
"""
from .base_provider import AIProvider
from .claude_provider import ClaudeProvider
from .gemini_provider import GeminiProvider
from .provider_factory import AIProviderFactory
__all__ = [
'AIProvider',
'ClaudeProvider',
'GeminiProvider',
'AIProviderFactory'
]

View File

@@ -0,0 +1,35 @@
"""
Base AI Provider interface (Strategy pattern)
"""
from abc import ABC, abstractmethod
from typing import Optional, Dict, Any
class AIProvider(ABC):
"""Abstract base class for AI providers"""
@abstractmethod
def summarize(self, text: str, **kwargs) -> str:
"""Generate summary of text"""
pass
@abstractmethod
def correct_text(self, text: str, **kwargs) -> str:
"""Correct grammar and spelling in text"""
pass
@abstractmethod
def classify_content(self, text: str, **kwargs) -> Dict[str, Any]:
"""Classify content into categories"""
pass
@abstractmethod
def is_available(self) -> bool:
"""Check if provider is available and configured"""
pass
@property
@abstractmethod
def name(self) -> str:
"""Provider name"""
pass

View File

@@ -0,0 +1,108 @@
"""
Claude AI Provider implementation
"""
import logging
import subprocess
import shutil
from typing import Dict, Any, Optional
from ..config import settings
from ..core import AIProcessingError
from .base_provider import AIProvider
class ClaudeProvider(AIProvider):
"""Claude AI provider using CLI"""
def __init__(self):
self.logger = logging.getLogger(__name__)
self._cli_path = settings.CLAUDE_CLI_PATH or shutil.which("claude")
self._token = settings.ZAI_AUTH_TOKEN
self._base_url = settings.ZAI_BASE_URL
@property
def name(self) -> str:
return "Claude"
def is_available(self) -> bool:
"""Check if Claude CLI is available"""
return bool(self._cli_path and self._token)
def _get_env(self) -> Dict[str, str]:
"""Get environment variables for Claude"""
env = {
'ANTHROPIC_AUTH_TOKEN': self._token,
'ANTHROPIC_BASE_URL': self._base_url,
'PYTHONUNBUFFERED': '1'
}
return env
def _run_cli(self, prompt: str, timeout: int = 300) -> str:
"""Run Claude CLI with prompt"""
if not self.is_available():
raise AIProcessingError("Claude CLI not available or not configured")
try:
cmd = [self._cli_path]
process = subprocess.run(
cmd,
input=prompt,
env=self._get_env(),
text=True,
capture_output=True,
timeout=timeout,
shell=False
)
if process.returncode != 0:
error_msg = process.stderr or "Unknown error"
raise AIProcessingError(f"Claude CLI failed: {error_msg}")
return process.stdout.strip()
except subprocess.TimeoutExpired:
raise AIProcessingError(f"Claude CLI timed out after {timeout}s")
except Exception as e:
raise AIProcessingError(f"Claude CLI error: {e}")
def summarize(self, text: str, **kwargs) -> str:
"""Generate summary using Claude"""
prompt = f"""Summarize the following text:
{text}
Provide a clear, concise summary in Spanish."""
return self._run_cli(prompt)
def correct_text(self, text: str, **kwargs) -> str:
"""Correct text using Claude"""
prompt = f"""Correct the following text for grammar, spelling, and clarity:
{text}
Return only the corrected text, nothing else."""
return self._run_cli(prompt)
def classify_content(self, text: str, **kwargs) -> Dict[str, Any]:
"""Classify content using Claude"""
categories = ["historia", "analisis_contable", "instituciones_gobierno", "otras_clases"]
prompt = f"""Classify the following text into one of these categories:
- historia
- analisis_contable
- instituciones_gobierno
- otras_clases
Text: {text}
Return only the category name, nothing else."""
result = self._run_cli(prompt).lower()
# Validate result
if result not in categories:
result = "otras_clases"
return {
"category": result,
"confidence": 0.9,
"provider": self.name
}

View File

@@ -0,0 +1,297 @@
"""
Gemini AI Provider - Optimized version with rate limiting and retry
"""
import logging
import subprocess
import shutil
import requests
import time
from typing import Dict, Any, Optional
from datetime import datetime, timedelta
from ..config import settings
from ..core import AIProcessingError
from .base_provider import AIProvider
class TokenBucket:
"""Token bucket rate limiter"""
def __init__(self, rate: float = 10, capacity: int = 20):
self.rate = rate # tokens per second
self.capacity = capacity
self.tokens = capacity
self.last_update = time.time()
self._lock = None # Lazy initialization
def _get_lock(self):
if self._lock is None:
import threading
self._lock = threading.Lock()
return self._lock
def acquire(self, tokens: int = 1) -> float:
with self._get_lock():
now = time.time()
elapsed = now - self.last_update
self.last_update = now
self.tokens = min(self.capacity, self.tokens + elapsed * self.rate)
if self.tokens >= tokens:
self.tokens -= tokens
return 0.0
wait_time = (tokens - self.tokens) / self.rate
self.tokens = 0
return wait_time
class CircuitBreaker:
"""Circuit breaker for API calls"""
def __init__(self, failure_threshold: int = 5, recovery_timeout: int = 60):
self.failure_threshold = failure_threshold
self.recovery_timeout = recovery_timeout
self.failures = 0
self.last_failure: Optional[datetime] = None
self.state = "closed" # closed, open, half-open
self._lock = None
def _get_lock(self):
if self._lock is None:
import threading
self._lock = threading.Lock()
return self._lock
def call(self, func, *args, **kwargs):
with self._get_lock():
if self.state == "open":
if self.last_failure and (datetime.utcnow() - self.last_failure).total_seconds() > self.recovery_timeout:
self.state = "half-open"
else:
raise AIProcessingError("Circuit breaker is open")
try:
result = func(*args, **kwargs)
if self.state == "half-open":
self.state = "closed"
self.failures = 0
return result
except Exception as e:
self.failures += 1
self.last_failure = datetime.utcnow()
if self.failures >= self.failure_threshold:
self.state = "open"
raise
class GeminiProvider(AIProvider):
"""Gemini AI provider with rate limiting and retry"""
def __init__(self):
super().__init__()
self._cli_path = settings.GEMINI_CLI_PATH or shutil.which("gemini")
self._api_key = settings.GEMINI_API_KEY
self._flash_model = settings.GEMINI_FLASH_MODEL
self._pro_model = settings.GEMINI_PRO_MODEL
self._session = None
self._rate_limiter = TokenBucket(rate=15, capacity=30)
self._circuit_breaker = CircuitBreaker(failure_threshold=5, recovery_timeout=60)
self._retry_config = {
"max_attempts": 3,
"base_delay": 1.0,
"max_delay": 30.0,
"exponential_base": 2
}
def _init_session(self) -> None:
"""Initialize HTTP session with connection pooling"""
if self._session is None:
self._session = requests.Session()
adapter = requests.adapters.HTTPAdapter(
pool_connections=10,
pool_maxsize=20,
max_retries=0 # We handle retries manually
)
self._session.mount('https://', adapter)
def _run_with_retry(self, func, *args, **kwargs):
"""Execute function with exponential backoff retry"""
max_attempts = self._retry_config["max_attempts"]
base_delay = self._retry_config["base_delay"]
last_exception = None
for attempt in range(max_attempts):
try:
return self._circuit_breaker.call(func, *args, **kwargs)
except requests.exceptions.RequestException as e:
last_exception = e
if attempt < max_attempts - 1:
delay = min(
base_delay * (2 ** attempt),
self._retry_config["max_delay"]
)
# Add jitter
delay += delay * 0.1 * (time.time() % 1)
self.logger.warning(f"Attempt {attempt + 1} failed: {e}, retrying in {delay:.2f}s")
time.sleep(delay)
raise AIProcessingError(f"Max retries exceeded: {last_exception}")
def _run_cli(self, prompt: str, use_flash: bool = True, timeout: int = 300) -> str:
"""Run Gemini CLI with prompt"""
if not self._cli_path:
raise AIProcessingError("Gemini CLI not available")
model = self._flash_model if use_flash else self._pro_model
cmd = [self._cli_path, model, prompt]
try:
# Apply rate limiting
wait_time = self._rate_limiter.acquire()
if wait_time > 0:
time.sleep(wait_time)
process = subprocess.run(
cmd,
text=True,
capture_output=True,
timeout=timeout,
shell=False
)
if process.returncode != 0:
error_msg = process.stderr or "Unknown error"
raise AIProcessingError(f"Gemini CLI failed: {error_msg}")
return process.stdout.strip()
except subprocess.TimeoutExpired:
raise AIProcessingError(f"Gemini CLI timed out after {timeout}s")
except Exception as e:
raise AIProcessingError(f"Gemini CLI error: {e}")
def _call_api(self, prompt: str, use_flash: bool = True, timeout: int = 180) -> str:
"""Call Gemini API with rate limiting and retry"""
if not self._api_key:
raise AIProcessingError("Gemini API key not configured")
self._init_session()
model = self._flash_model if use_flash else self._pro_model
url = f"https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent"
payload = {
"contents": [{
"parts": [{"text": prompt}]
}]
}
params = {"key": self._api_key}
def api_call():
# Apply rate limiting
wait_time = self._rate_limiter.acquire()
if wait_time > 0:
time.sleep(wait_time)
response = self._session.post(
url,
json=payload,
params=params,
timeout=timeout
)
response.raise_for_status()
return response
response = self._run_with_retry(api_call)
data = response.json()
if "candidates" not in data or not data["candidates"]:
raise AIProcessingError("Empty response from Gemini API")
candidate = data["candidates"][0]
if "content" not in candidate or "parts" not in candidate["content"]:
raise AIProcessingError("Invalid response format from Gemini API")
result = candidate["content"]["parts"][0]["text"]
return result.strip()
def _run(self, prompt: str, use_flash: bool = True, timeout: int = 300) -> str:
"""Run Gemini with fallback between CLI and API"""
# Try CLI first if available
if self._cli_path:
try:
return self._run_cli(prompt, use_flash, timeout)
except Exception as e:
self.logger.warning(f"Gemini CLI failed, trying API: {e}")
# Fallback to API
if self._api_key:
api_timeout = min(timeout, 180)
return self._call_api(prompt, use_flash, api_timeout)
raise AIProcessingError("No Gemini provider available (CLI or API)")
def summarize(self, text: str, **kwargs) -> str:
"""Generate summary using Gemini"""
prompt = f"""Summarize the following text:
{text}
Provide a clear, concise summary in Spanish."""
return self._run(prompt, use_flash=True)
def correct_text(self, text: str, **kwargs) -> str:
"""Correct text using Gemini"""
prompt = f"""Correct the following text for grammar, spelling, and clarity:
{text}
Return only the corrected text, nothing else."""
return self._run(prompt, use_flash=True)
def classify_content(self, text: str, **kwargs) -> Dict[str, Any]:
"""Classify content using Gemini"""
categories = ["historia", "analisis_contable", "instituciones_gobierno", "otras_clases"]
prompt = f"""Classify the following text into one of these categories:
- historia
- analisis_contable
- instituciones_gobierno
- otras_clases
Text: {text}
Return only the category name, nothing else."""
result = self._run(prompt, use_flash=True).lower()
# Validate result
if result not in categories:
result = "otras_clases"
return {
"category": result,
"confidence": 0.9,
"provider": self.name
}
def get_stats(self) -> Dict[str, Any]:
"""Get provider statistics"""
return {
"rate_limiter": {
"tokens": round(self._rate_limiter.tokens, 2),
"capacity": self._rate_limiter.capacity,
"rate": self._rate_limiter.rate
},
"circuit_breaker": {
"state": self._circuit_breaker.state,
"failures": self._circuit_breaker.failures,
"failure_threshold": self._circuit_breaker.failure_threshold
},
"cli_available": bool(self._cli_path),
"api_available": bool(self._api_key)
}
# Global instance is created in __init__.py

View File

@@ -0,0 +1,54 @@
"""
AI Provider Factory (Factory Pattern)
"""
import logging
from typing import Dict, Type
from ..core import AIProcessingError
from .base_provider import AIProvider
from .claude_provider import ClaudeProvider
from .gemini_provider import GeminiProvider
class AIProviderFactory:
"""Factory for creating AI providers with fallback"""
def __init__(self):
self.logger = logging.getLogger(__name__)
self._providers: Dict[str, AIProvider] = {
'claude': ClaudeProvider(),
'gemini': GeminiProvider()
}
def get_provider(self, preferred: str = 'gemini') -> AIProvider:
"""Get available provider with fallback"""
# Try preferred provider first
if preferred in self._providers:
provider = self._providers[preferred]
if provider.is_available():
self.logger.info(f"Using {preferred} provider")
return provider
# Fallback to any available provider
for name, provider in self._providers.items():
if provider.is_available():
self.logger.info(f"Falling back to {name} provider")
return provider
raise AIProcessingError("No AI providers available")
def get_all_available(self) -> Dict[str, AIProvider]:
"""Get all available providers"""
return {
name: provider
for name, provider in self._providers.items()
if provider.is_available()
}
def get_best_provider(self) -> AIProvider:
"""Get the best available provider (Gemini > Claude)"""
return self.get_provider('gemini')
# Global instance
ai_provider_factory = AIProviderFactory()