Storage

Storage Backends

ArtemisKit supports multiple storage backends for persisting evaluation results. Choose the one that best fits your workflow.

Available Backends

Backend	Best For	Persistence
Local	Development, single-user	Filesystem
Supabase	Teams, CI/CD, cloud	PostgreSQL + Storage

Configuration

Set the storage backend in your config file:

storage:
  type: local  # or "supabase"
  basePath: ./artemis-runs

Or use environment variables:

export ARTEMIS_STORAGE_PATH=./artemis-runs

Storage Features

All storage backends support:

Save - Store run manifests with full results
Load - Retrieve a specific run by ID
List - Query runs with filtering and pagination
Delete - Remove runs
Compare - Diff two runs for regression detection

Run Manifest

Each evaluation run produces a manifest containing:

interface RunManifest {
  run_id: string;           // Unique identifier (e.g., "ar-20260115-abc123")
  project: string;          // Project name
  config: {
    scenario: string;       // Scenario filename
    provider: string;       // LLM provider used
    model: string;          // Model used
  };
  metrics: {
    success_rate: number;   // Pass rate (0-1)
    total_cases: number;
    passed_cases: number;
    failed_cases: number;
    median_latency_ms: number;
    p95_latency_ms: number;
    total_tokens: number;
  };
  git: {
    commit: string;
    branch: string;
    dirty: boolean;
  };
  provenance: {
    run_by: string;
    run_reason: string;
  };
  start_time: string;
  end_time: string;
  results: CaseResult[];
}

Choosing a Backend

Use Local Storage When:

Developing and testing scenarios
Running evaluations on a single machine
You don’t need to share results across team members
Quick setup with zero configuration

Use Supabase Storage When:

Running evaluations in CI/CD pipelines
Collaborating with a team
Need historical trending and analytics
Want SQL-queryable results
Building dashboards or integrations