schema_version: 1
# Tier 2 Domain Foundry catalogue (spec §5) — pure configuration.
# datasets: HF repo ids (reference only; user picks). benchmarks: eval sets
# with an HF repo or 'upload' (user-provided). Each domain pre-fills Tier-1.

domains:
  - id: finance
    name: Finance
    system_prompt: >
      You are a financial analysis assistant with expertise in markets, filings,
      accounting standards (IFRS/GAAP), and financial regulation. Be precise and
      say so when you are unsure. You do not give personalized investment advice.
    disclaimer: Research/education only — not investment, legal, or accounting advice.
    datasets: [FinGPT/fingpt-fiqa_qa, gbharti/finance-alpaca, FinGPT/fingpt-sentiment-train]
    benchmarks:
      - {id: financial_qa, name: Financial QA (FiQA sample), source: FinGPT/fingpt-fiqa_qa, metric: rougeL}
      - {id: fin_sentiment, name: Financial Sentiment, source: FinGPT/fingpt-sentiment-train, metric: accuracy}
      - {id: sec_filing_qa, name: SEC Filing QA, source: upload, metric: rougeL}
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: law
    name: Law
    system_prompt: You are a legal research assistant. Cite sources cautiously; you do not give legal advice.
    disclaimer: Research/education only — not legal advice.
    datasets: [pile-of-law/pile-of-law, nguha/legalbench]
    benchmarks:
      - {id: legalbench_sample, name: LegalBench (sampled), source: nguha/legalbench, metric: accuracy}
      - {id: citation_accuracy, name: Citation Accuracy, source: upload, metric: accuracy}
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 8.0e-5, epochs: 2}

  - id: medical
    name: Medical
    system_prompt: You are a medical literature assistant. Flag uncertainty; you do not give medical advice.
    disclaimer: Research/education only — not medical advice.
    datasets: [openlifescienceai/medmcqa, qiaojin/PubMedQA]
    benchmarks:
      - {id: medqa_sample, name: MedMCQA (sampled), source: openlifescienceai/medmcqa, metric: accuracy}
      - {id: pubmedqa_sample, name: PubMedQA (sampled), source: qiaojin/PubMedQA, metric: accuracy}
    hyperparameters: {lora_r: 16, lora_alpha: 32, learning_rate: 8.0e-5, epochs: 2}

  - id: regulatory
    name: Regulatory
    system_prompt: You are a regulatory-compliance research assistant (Basel III, MiFID II, FCA, SEC).
    disclaimer: Research/education only — not compliance advice.
    datasets: []
    benchmarks:
      - {id: reg_qa, name: Regulatory QA, source: upload, metric: rougeL}
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: accounting
    name: Accounting
    system_prompt: You are an accounting assistant (IFRS/GAAP reporting and analysis).
    disclaimer: Research/education only.
    datasets: []
    benchmarks: [{id: acc_qa, name: Accounting QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: insurance
    name: Insurance
    system_prompt: You are an insurance-domain assistant (underwriting, claims, policy language).
    disclaimer: Research/education only.
    datasets: []
    benchmarks: [{id: ins_qa, name: Insurance QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: computer_science
    name: Computer Science
    system_prompt: You are a computer-science tutor and research assistant.
    disclaimer: ""
    datasets: [camel-ai/computer_science]
    benchmarks: [{id: cs_qa, name: CS QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: programming
    name: Programming
    system_prompt: You are a coding assistant. Provide runnable, correct code.
    disclaimer: ""
    datasets: [sahil2801/CodeAlpaca-20k, openai/openai_humaneval]
    benchmarks:
      - {id: humaneval_sample, name: HumanEval (sampled), source: openai/openai_humaneval, metric: accuracy}
      - {id: mbpp_sample, name: MBPP (sampled), source: google-research-datasets/mbpp, metric: accuracy}
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.2e-4, epochs: 3}

  - id: artificial_intelligence
    name: Artificial Intelligence
    system_prompt: You are an AI/ML research assistant.
    disclaimer: ""
    datasets: [camel-ai/ai_society]
    benchmarks: [{id: ai_qa, name: AI QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}

  - id: literature
    name: Literature
    system_prompt: You are a literature analysis and writing assistant.
    disclaimer: ""
    datasets: []
    benchmarks: [{id: lit_qa, name: Literature QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 16, lora_alpha: 32, learning_rate: 8.0e-5, epochs: 2}

  - id: general_science
    name: General Science
    system_prompt: You are a general-science explainer grounded in established research.
    disclaimer: ""
    datasets: [camel-ai/physics, camel-ai/chemistry, camel-ai/biology]
    benchmarks: [{id: sci_qa, name: Science QA, source: upload, metric: rougeL}]
    hyperparameters: {lora_r: 32, lora_alpha: 64, learning_rate: 1.0e-4, epochs: 2}