config.py

# config.py
import os

# HuggingFace settings
HF_TOKEN = os.getenv("HF_TOKEN")
LEADERBOARD_DATASET = "Sunbird/salt-translation-leaderboard"
TEST_SET_DATASET = "Sunbird/salt-translation-test-set"
SALT_DATASET = "sunbird/salt"

# Language settings - ALL UG40 LANGUAGES
ALL_UG40_LANGUAGES = ["ach", "eng", "lgg", "lug", "nyn", "rny", "teo", "swa"]

LANGUAGE_NAMES = {
    "ach": "Acholi",
    "eng": "English", 
    "lgg": "Lugbara",
    "lug": "Luganda",
    "nyn": "Runyankole",
    "rny": "Runyoro",
    "teo": "Ateso",
    "swa": "Swahili",
}

# Google Translate supported subset (for fair comparison)
GOOGLE_SUPPORTED_LANGUAGES = ["lug", "ach", "swa", "eng"]

# EVALUATION TRACKS
EVALUATION_TRACKS = {
    "google_comparable": {
        "name": "Google-Comparable Track",
        "description": "Models evaluated on language pairs supported by Google Translate for commercial comparison",
        "languages": GOOGLE_SUPPORTED_LANGUAGES,
        "min_samples_per_pair": 50,
    },
    "ug40_complete": {
        "name": "UG40-Complete Track", 
        "description": "Models evaluated on all UG40 language pairs for comprehensive assessment",
        "languages": ALL_UG40_LANGUAGES,
        "min_samples_per_pair": 30,
    },
}

# MODEL CATEGORIES
MODEL_CATEGORIES = {
    "commercial": {
        "name": "Commercial Systems",
        "description": "Production translation systems",
        "examples": ["google_translate", "azure_translator"],
        "color": "#1f77b4",
    },
    "research": {
        "name": "Research Models", 
        "description": "Academic and research institution models",
        "examples": ["nllb", "m2m100"],
        "color": "#ff7f0e",
    },
    "baseline": {
        "name": "Baseline Models",
        "description": "Simple baseline and reference models",
        "examples": ["word_lookup", "frequency_baseline"],
        "color": "#2ca02c",
    },
    "community": {
        "name": "Community Submissions",
        "description": "User-submitted models and fine-tuned variants", 
        "examples": ["user_submission"],
        "color": "#d62728",
    },
}

# METRICS CONFIGURATION
METRICS_CONFIG = {
    "primary_metrics": ["bleu", "chrf", "quality_score"],
    "secondary_metrics": ["rouge1", "rougeL", "cer", "wer"],
    "display_precision": 4,
    "confidence_level": 0.95,
    "bootstrap_samples": 1000,
    "min_samples_for_ci": 20,
}

# VALIDATION REQUIREMENTS
VALIDATION_CONFIG = {
    "min_samples_per_track": {
        "google_comparable": 200,
        "ug40_complete": 400,
    },
    "max_missing_rate": 0.05,  # 5% missing predictions allowed
    "quality_thresholds": {
        "min_valid_predictions": 0.95,
        "max_duplicate_rate": 0.1,
        "min_avg_length": 3,
        "max_avg_length": 500,
    },
}

# FILE FORMAT SPECIFICATIONS
PREDICTION_FORMAT = {
    "required_columns": ["sample_id", "prediction"],
    "optional_columns": ["model_name", "confidence", "category"],
    "file_types": [".csv", ".tsv", ".json"],
    "category_detection": {
        "google": ["google", "translate"],
        "nllb": ["nllb", "meta"],
        "m2m": ["m2m", "facebook"],
        "baseline": ["baseline", "simple", "lookup"],
    },
}

# EVALUATION SETTINGS
MAX_TEST_SAMPLES = 500  # Per language pair
MIN_SAMPLES_PER_PAIR = 10  # Minimum for basic statistics

# CHART CONFIGURATION
CHART_CONFIG = {
    "category_colors": {cat: info["color"] for cat, info in MODEL_CATEGORIES.items()},
    "height": 600,
    "width": 800,
    "margin": {"l": 100, "r": 50, "t": 50, "b": 100},
}