# Database Schema Documentation Based on the evidence gathered, I can provide information about the columns across multiple tables in this database. The data shows **50 rows were returned** from what appears to be a larger schema (indicated by "showing a bounded subset"). ## Tables and Columns Identified ### 1. **analysis_run_llm_usage** Tracks LLM (Large Language Model) usage metrics for analysis runs. | Column | Data Type | Semantic Role | Meaning | |--------|-----------|---------------|---------| | `calls` | BIGINT | metric | Number of LLM API calls made | | `cost_usd` | DOUBLE | measure | Cost in USD for LLM usage | | `input_tokens` | BIGINT | measure | Number of tokens sent to the LLM | | `jobs_id` | BIGINT | identifier | Foreign key linking to a job | | `model_name` | VARCHAR | descriptive | Name of the LLM model used | | `output_tokens` | BIGINT | measure | Number of tokens generated by the LLM | ### 2. **analysis_runs** Contains information about analysis job executions. | Column | Data Type | Semantic Role | Meaning | |--------|-----------|---------------|---------| | `analysis_config_json` | VARCHAR | classifier | JSON configuration for the analysis | | `code_version` | VARCHAR | identifier | Version of code used for the analysis | | `cost_compute_seconds` | DOUBLE | measure | Computational cost in seconds | | `cost_storage_bytes` | BIGINT | measure | Storage cost in bytes | | `dataset_version_hash` | VARCHAR | identifier | Hash identifying the dataset version | | `error_text` | VARCHAR | descriptive | Error message if the run failed | | `intent` | VARCHAR | identifier | Purpose or goal of the analysis | | `internal_run_id` | VARCHAR | identifier | Internal identifier for the run | | `jobs_id` | BIGINT | identifier | Job identifier | | `narrative_summary` | VARCHAR | measure | Text summary of the analysis | | `queries_executed` | BIGINT | measure | Number of queries executed | | `report_id` | VARCHAR | identifier | Identifier for the generated report | | `style` | VARCHAR | classifier | Style or type of analysis | | `total_cost_usd` | DOUBLE | measure | Total cost in USD | | `total_input_tokens` | BIGINT | measure | Total input tokens across all LLM calls | | `total_llm_calls` | BIGINT | measure | Total number of LLM calls | | `total_output_tokens` | BIGINT | measure | Total output tokens across all LLM calls | | `total_tool_calls` | BIGINT | measure | Total number of tool invocations | ### 3. **batches** Tracks batch processing jobs. | Column | Data Type | Semantic Role | Meaning | |--------|-----------|---------------|---------| | `batch_id` | VARCHAR | identifier | Unique identifier for the batch | | `completed_at` | VARCHAR | temporal | Timestamp when batch completed | | `content_hash` | VARCHAR | descriptive | Hash of the batch content | | `dataset_id` | VARCHAR | identifier | Identifier for the dataset | | `elapsed_ms` | BIGINT | identifier | Elapsed time in milliseconds | | `engine_versions_json` | VARCHAR | classifier | JSON of engine versions used | | `error_text` | VARCHAR | descriptive | Error message if batch failed | | `job_id` | VARCHAR | identifier | Job identifier | | `nettle_version` | VARCHAR | classifier | Version of Nettle system | | `started_at` | VARCHAR | temporal | Timestamp when batch started | | `status` | VARCHAR | classifier | Current status of the batch | | `tenant_id` | VARCHAR | identifier | Identifier for the tenant/customer | | `total_bytes` | BIGINT | measure | Total size in bytes | | `total_rows` | BIGINT | measure | Total number of rows processed | ### 4. **classification_candidates** Stores potential classification schemes for data organization. | Column | Data Type | Semantic Role | Meaning | |--------|-----------|---------------|---------| | `clustering_key` | VARCHAR | identifier | Key used for clustering data | | `clustering_key_label` | VARCHAR | identifier | Human-readable label for clustering key | | `clustering_key_reasoning` | VARCHAR | identifier | Explanation for clustering key choice | | `confidence` | VARCHAR | classifier | Confidence level of the classification | | `id` | BIGINT | identifier | Unique identifier | | `notes` | VARCHAR | descriptive | Additional notes or comments | | `partition_key` | VARCHAR | identifier | Key used for partitioning data | | `partition_key_label` | VARCHAR | identifier | Human-readable label for partition key | | `partition_key_reasoning` | VARCHAR | identifier | Explanation for partition key choice | | `rank` | BIGINT | classifier | Ranking of this candidate | | `session_column` | VARCHAR | classifier | Column identifying sessions | | `table_classification_id` | BIGINT | identifier | Foreign key to table classification | ## Important Notes - **Incomplete data**: Only 50 rows were returned from what appears to be a larger schema - **Additional tables likely exist** beyond these four tables - The schema appears to be for a data analysis platform (Nettle) that uses LLMs and tracks costs, performance metrics, and data processing batches