mindnet/app/services/semantic_analyzer.py

"""
app/services/semantic_analyzer.py — Edge Validation & Filtering
Version: 1.1 (Robust JSON Parsing)
"""

import json
import logging
from typing import List, Optional
from dataclasses import dataclass

# Importe
from app.services.llm_service import LLMService

logger = logging.getLogger(__name__)

class SemanticAnalyzer:
    def __init__(self):
        self.llm = LLMService()

    async def assign_edges_to_chunk(self, chunk_text: str, all_edges: List[str], note_type: str) -> List[str]:
        """
        Sendet einen Chunk und eine Liste potenzieller Kanten an das LLM.
        Das LLM filtert heraus, welche Kanten für diesen Chunk relevant sind.
        """
        if not all_edges:
            return []

        # 1. Prompt laden
        prompt_template = self.llm.prompts.get("edge_allocation_template")
        
        # Fallback, falls Prompt nicht in YAML definiert ist (für Tests ohne volle Config)
        if not prompt_template:
            prompt_template = (
                "TASK: Wähle aus den Kandidaten die relevanten Kanten für den Text.\n"
                "TEXT: {chunk_text}\n"
                "KANDIDATEN: {edge_list}\n"
                "OUTPUT: JSON Liste von Strings [\"kind:target\"]."
            )

        # 2. Kandidaten-Liste formatieren
        edges_str = "\n".join([f"- {e}" for e in all_edges])

        # 3. Prompt füllen
        final_prompt = prompt_template.format(
            chunk_text=chunk_text[:3000], 
            edge_list=edges_str
        )

        try:
            # 4. LLM Call mit JSON Erzwingung
            response_json = await self.llm.generate_raw_response(
                prompt=final_prompt,
                force_json=True
            )

            # 5. Parsing & Cleaning
            clean_json = response_json.replace("```json", "").replace("```", "").strip()
            if not clean_json: return []

            data = json.loads(clean_json)
            valid_edges = []

            # 6. Robuste Validierung (List vs Dict)
            if isinstance(data, list):
                # Standardfall: ["kind:target", ...]
                valid_edges = [str(e) for e in data if isinstance(e, str) and ":" in e]
            
            elif isinstance(data, dict):
                # Abweichende Formate behandeln
                for key, val in data.items():
                    # Fall A: {"edges": ["kind:target"]}
                    if key.lower() in ["edges", "results", "kanten"] and isinstance(val, list):
                         valid_edges.extend([str(e) for e in val if isinstance(e, str) and ":" in e])
                    
                    # Fall B: {"kind": "target"} (Das beobachtete Format im Log)
                    elif isinstance(val, str):
                        # Wir rekonstruieren "kind:target"
                        valid_edges.append(f"{key}:{val}")
                    
                    # Fall C: {"kind": ["target1", "target2"]}
                    elif isinstance(val, list):
                        for target in val:
                            if isinstance(target, str):
                                valid_edges.append(f"{key}:{target}")

            # Safety: Filtere nur Kanten, die halbwegs valide aussehen
            return [e for e in valid_edges if ":" in e]

        except json.JSONDecodeError:
            logger.warning("SemanticAnalyzer: LLM lieferte kein valides JSON. Ignoriere Zuweisung.")
            return []
        except Exception as e:
            logger.error(f"SemanticAnalyzer Error: {e}")
            return []

    async def close(self):
        if self.llm:
            await self.llm.close()

# Singleton Helper
_analyzer_instance = None
def get_semantic_analyzer():
    global _analyzer_instance
    if _analyzer_instance is None:
        _analyzer_instance = SemanticAnalyzer()
    return _analyzer_instance