{
  "context": "This evaluation assesses the engineer's ability to use deeplake's Tensor Query Language (TQL) to perform filtering, aggregation, and similarity search operations on datasets. The focus is exclusively on proper usage of TQL query methods and syntax.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "TQL query execution",
      "description": "Uses deeplake.query() or deeplake.prepare_query() to execute TQL queries rather than iterating through dataset manually or using other non-TQL approaches",
      "max_score": 25
    },
    {
      "name": "WHERE clause filtering",
      "description": "Correctly uses TQL WHERE clause syntax to filter by label (e.g., SELECT * FROM dataset WHERE label = 'cat') in the filter_by_label function",
      "max_score": 15
    },
    {
      "name": "GROUP BY aggregation",
      "description": "Uses TQL GROUP BY and COUNT() aggregation (e.g., SELECT label, COUNT(*) FROM dataset GROUP BY label) in get_label_statistics function",
      "max_score": 20
    },
    {
      "name": "COSINE_SIMILARITY function",
      "description": "Correctly uses TQL's COSINE_SIMILARITY function with ORDER BY and LIMIT to find similar images (e.g., ORDER BY COSINE_SIMILARITY(embedding, [vector]) DESC LIMIT k)",
      "max_score": 25
    },
    {
      "name": "Multi-condition filtering",
      "description": "Uses TQL AND operator to combine multiple WHERE conditions (e.g., WHERE id > min_id AND label = 'dog') in filter_by_multiple_conditions function",
      "max_score": 10
    },
    {
      "name": "Query result handling",
      "description": "Correctly accesses and processes TQL query results to extract required data (IDs, counts, labels) matching the specified return types",
      "max_score": 5
    }
  ]
}