chore: Remove obsolete build and GUI files, and introduce a new Docker Compose setup for microservices architecture. This change streamlines the project structure by eliminating legacy scripts and integrating a PostgreSQL database service alongside extraction and cleaning services.

services/extraction/domain/strategies.py

@@ -0,0 +1,131 @@
+"""Estrategias de extracción de información desde imágenes de documentos.
+
+Define la interfaz `InvoiceExtractorStrategy` y una implementación basada en
+un endpoint compatible con la API de OpenAI para modelos multimodales.
+"""
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import List, Dict, Any, Optional
+
+from .config import extractor_llm_config
+from .llm_client import LLMClient
+import logging
+import time
+
+
+class InvoiceExtractorStrategy(ABC):
+  """Interfaz para extraer datos a partir de páginas en base64 e instrucción."""
+  @abstractmethod
+  def extract(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]] = None) -> Dict[str, Any]:
+    """Devuelve un dict con al menos la clave "raw" (texto de salida del modelo)."""
+    raise NotImplementedError
+
+  async def extract_async(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]] = None) -> Dict[str, Any]:
+    """Versión asíncrona de extract. Por defecto lanza NotImplementedError."""
+    raise NotImplementedError
+
+
+_LOG_EXTRA_BASE = {"dpi": '-', "pdf": '-', "radicado": '-'}
+
+
+class OpenAICompatibleExtractor(InvoiceExtractorStrategy):
+  """Extractor contra un backend compatible con Chat Completions.
+
+  Parámetros configurables: `api_key`, `base_url` y `model`.
+  """
+  def __init__(self, api_key: Optional[str] = None, base_url: Optional[str] = None, model: Optional[str] = None, llm_client: Optional[LLMClient] = None) -> None:
+    config = extractor_llm_config()
+    self.api_key = api_key if api_key is not None else str(config["api_key"])
+    self.base_url = base_url if base_url is not None else str(config["base_url"])
+    self.model = model if model is not None else str(config["model"])
+    self.request_timeout_s = float(config["timeout_s"])
+    self.llm = llm_client or LLMClient(
+      api_key=self.api_key,
+      base_url=self.base_url,
+      default_model=self.model,
+      request_timeout_s=self.request_timeout_s,
+    )
+    self._logger = logging.getLogger(__name__)
+
+  def _build_messages(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]]) -> List[Dict[str, Any]]:
+    content: List[Dict[str, Any]] = [
+      {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{img}"}}
+      for img in images_b64
+    ]
+    content.append({"type": "text", "text": instruction})
+    messages = list(history) if history else []
+    messages.append({"role": "user", "content": content})
+    return messages
+
+  def _log_call_start(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]]) -> None:
+    prompt_len = len(instruction) if isinstance(instruction, str) else 0
+    self._logger.info(
+      "backend_call_start",
+      extra={**_LOG_EXTRA_BASE, "base_url": self.base_url, "model": self.model,
+             "timeout_s": self.request_timeout_s, "images": len(images_b64),
+             "prompt_len": prompt_len, "history_len": len(history) if history else 0},
+    )
+    self._logger.debug(
+      f"Llamada backend: images={len(images_b64)}, prompt_len={prompt_len}, history_len={len(history) if history else 0}",
+      extra=_LOG_EXTRA_BASE,
+    )
+
+  def _log_call_ok(self, elapsed: float) -> None:
+    self._logger.info(
+      "Extracción completada",
+      extra={**_LOG_EXTRA_BASE, "base_url": self.base_url, "model": self.model,
+             "timeout_s": self.request_timeout_s, "elapsed_ms": int(elapsed * 1000)},
+    )
+    self._logger.debug(f"Backend OK (elapsed_ms={int(elapsed*1000)})", extra=_LOG_EXTRA_BASE)
+
+  def _log_call_error(self, elapsed: float) -> None:
+    self._logger.exception(
+      "Error en llamada al backend",
+      extra={**_LOG_EXTRA_BASE, "base_url": self.base_url, "model": self.model,
+             "timeout_s": self.request_timeout_s, "elapsed_ms": int(elapsed * 1000)},
+    )
+    self._logger.debug(f"Backend error tras {int(elapsed*1000)} ms", extra=_LOG_EXTRA_BASE)
+
+  def extract(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]] = None) -> Dict[str, Any]:
+    """Construye el prompt multimodal y devuelve el texto crudo del modelo."""
+    messages = self._build_messages(images_b64, instruction, history)
+    self._log_call_start(images_b64, instruction, history)
+    start = time.time()
+    try:
+      text, usage_dict = self.llm.chat_text(
+        messages=messages, model=self.model,
+        timeout_s=self.request_timeout_s, component="extractor",
+      )
+      self._log_call_ok(time.time() - start)
+      return {"raw": text, "usage": usage_dict}
+    except Exception:
+      self._log_call_error(time.time() - start)
+      raise
+
+  async def extract_async(self, images_b64: List[str], instruction: str, history: Optional[List[Dict[str, Any]]] = None) -> Dict[str, Any]:
+    """Versión asíncrona de extract."""
+    messages = self._build_messages(images_b64, instruction, history)
+    self._log_call_start(images_b64, instruction, history)
+    start = time.time()
+    try:
+      text, usage_dict = await self.llm.chat_text_async(
+        messages=messages, model=self.model,
+        timeout_s=self.request_timeout_s, component="extractor",
+      )
+      self._log_call_ok(time.time() - start)
+      return {"raw": text, "usage": usage_dict}
+    except Exception:
+      self._log_call_error(time.time() - start)
+      raise
+
+  def update_config(self, *, api_key: str | None = None, base_url: str | None = None, model: str | None = None) -> None:
+    """Actualiza la configuración del cliente y recrea la instancia interna."""
+    if api_key is not None:
+      self.api_key = api_key
+    if base_url is not None:
+      self.base_url = base_url
+    if model is not None:
+      self.model = model
+    # actualizar cliente LLM con config actual
+    self.llm.update_config(api_key=self.api_key, base_url=self.base_url, default_model=self.model, request_timeout_s=self.request_timeout_s)