src/evaluation_config.yaml

# Evaluation Settings
evaluation:
  threshold: 0.8
  verbose_mode: true
  max_retries: 3
  timeout_seconds: 30

# Dataset Settings
dataset:
  input_column: "input"
  output_column: "expected_output"
  context_column: null
  
# Logging Configuration
logging:
  level: "INFO"
  max_file_size_mb: 10
  backup_count: 5
  
# Output Settings
output:
  results_dir: "results"
  logs_dir: "logs"
  save_detailed_results: true
  
# Metrics Configuration
metrics:
  answer_relevancy:
    enabled: true
    threshold: 0.8
    verbose_mode: true
    
  contextual_precision:
    enabled: false
    threshold: 0.8

# RAG System Configuration
rag_system:
  embedding_model: "qwen3-embedding"
  reranker_model: "qwen3-reranker"
  llm_model: "openai/gpt-oss-120b"  # Groq model
  vector_store: "qdrant"
  chunk_size: 512
  chunk_overlap: 50
  max_context_chunks: 5
  similarity_threshold: 0.7
  rerank_top_k: 20
  final_top_k: 5

# API Configuration
apis:
  groq_api_key: "${GROQ_API_KEY}"
  siliconflow_api_key: "${SILICONFLOW_API_KEY}"
  qdrant_url: "${QDRANT_URL}"
  qdrant_api_key: "${QDRANT_API_KEY}"

# Document Processing Configuration
document_processing:
  pdf_engine: "pymupdf"
  excel_engine: "openpyxl"
  ocr_engine: "tesseract"
  image_processing: true
  table_extraction: true
  max_file_size_mb: 100

# Storage Configuration
storage:
  qdrant_collection: "manufacturing_docs"
  metadata_db_path: "./data/metadata.db"
  file_storage_path: "./data/documents"
  enable_compression: true