migrations.sql

-- prompts
create table if not exists public.prompts (
  id uuid primary key default gen_random_uuid(),
  prompt_text text not null,
  created_at timestamptz default now()
);

-- model_responses: one row per prompt, columns for each model
create table if not exists public.model_responses (
  id uuid primary key default gen_random_uuid(),
  prompt_id uuid references public.prompts(id) on delete cascade,
  -- store model outputs separately; add/remove columns to match model_list in code
  openai_gpt_5 text,
  anthropic_claude_sonnet_4_5 text,
  google_gemini_2_5_pro text,
  x_ai_grok_4_fast text,
  deepseek_v3_2 text,
  z_ai_glm_4_5 text,
  openai_gpt_4_1 text,
  moonshot_kimi_k2 text,
  qwen_qwen3_235b text,
  created_at timestamptz default now()
);

-- judge evaluations (parsed scores)
create table if not exists public.llm_evaluations (
  id uuid primary key default gen_random_uuid(),
  prompt_id uuid references public.prompts(id) on delete cascade,
  model_name text,
  scores jsonb,   -- parsed per-metric scores
  evaluation_text text,
  judge_model text,
  created_at timestamptz default now()
);

-- prompt_workflow store raw aggregated response + judge output (for debugging)
create table if not exists public.prompt_workflow (
  id uuid primary key default gen_random_uuid(),
  prompt_id uuid,
  test_prompt text,
  all_models_response text,
  judge_response text,
  created_at timestamptz default now()
);