#!/usr/bin/env nextflow

nextflow.enable.dsl=2

// Pipeline parameters
params.pdb = 's3://omic/eureka/Pocketminer/1HSG.pdb'
params.outdir = 's3://omic/eureka/Pocketminer/output'
params.model_path = '/workspace/gvp/models/pocketminer'
params.debug = false

process POCKETMINER {
    container 'harbor.cluster.omic.ai/omic/pocketminer:latest'
    publishDir params.outdir, mode: 'copy'

    input:
        val pdb_path

    output:
        path "*-preds.npy", emit: predictions_npy
        path "*-predictions.txt", emit: predictions_txt
        path "*-summary.json", emit: summary
        path "*_X.npy", optional: true, emit: features_debug
        path "*_S.npy", optional: true, emit: sequence_debug
        path "*_mask.npy", optional: true, emit: mask_debug
        path "run.log", emit: log

    script:
        def debug_flag = params.debug ? '--debug' : ''
        """
        touch run.log

        echo "=== Environment Debug ===" | tee -a run.log
        echo "Input path: ${pdb_path}" | tee -a run.log
        echo "AWS_ENDPOINT_URL: \${AWS_ENDPOINT_URL:-not set}" | tee -a run.log
        echo "AWS_ACCESS_KEY_ID set: \${AWS_ACCESS_KEY_ID:+yes}" | tee -a run.log

        # Try multiple methods to get the PDB file
        PDB_FILE=""

        # Method 1: Direct PVC path (if mounted)
        if [ -f "${pdb_path}" ]; then
            echo "Found file at PVC path: ${pdb_path}" | tee -a run.log
            cp "${pdb_path}" input.pdb
            PDB_FILE="input.pdb"
        fi

        # Method 2: Convert s3://omic/eureka/... to /omic/eureka/... (PVC mount)
        if [ -z "\$PDB_FILE" ]; then
            PVC_PATH="\$(echo '${pdb_path}' | sed 's|^s3://|/|')"
            echo "Trying PVC path: \$PVC_PATH" | tee -a run.log
            if [ -f "\$PVC_PATH" ]; then
                echo "Found file at PVC mount: \$PVC_PATH" | tee -a run.log
                cp "\$PVC_PATH" input.pdb
                PDB_FILE="input.pdb"
            fi
        fi

        # Method 3: Download from MinIO via S3 API
        if [ -z "\$PDB_FILE" ]; then
            echo "PVC not available, downloading from MinIO S3..." | tee -a run.log
            pip install awscli 2>/dev/null || conda install -n base -c conda-forge awscli -y 2>/dev/null || true

            if command -v aws &> /dev/null; then
                aws --endpoint-url \${AWS_ENDPOINT_URL:-http://datalake-hl.datalake.svc.cluster.local:9000} \\
                    s3 cp "${pdb_path}" input.pdb 2>&1 | tee -a run.log && PDB_FILE="input.pdb"
            fi
        fi

        # Method 4: Download via curl from MinIO
        if [ -z "\$PDB_FILE" ]; then
            echo "Trying curl download from MinIO..." | tee -a run.log
            S3_PATH="\$(echo '${pdb_path}' | sed 's|^s3://||')"
            curl -sf "http://datalake-hl.datalake.svc.cluster.local:9000/\$S3_PATH" -o input.pdb 2>&1 | tee -a run.log && PDB_FILE="input.pdb"
        fi

        if [ -z "\$PDB_FILE" ] || [ ! -f input.pdb ]; then
            echo "ERROR: Could not retrieve PDB file from any source" | tee -a run.log
            exit 1
        fi

        echo "PDB file ready (\$(wc -c < input.pdb) bytes)" | tee -a run.log
        PDB_BASENAME="\$(basename '${pdb_path}' .pdb)"

        python /workspace/entrypoint.py \\
            --pdb input.pdb \\
            --output-folder . \\
            --output-name "\$PDB_BASENAME" \\
            --model-path ${params.model_path} \\
            ${debug_flag} 2>&1 | tee -a run.log

        echo "Pipeline completed successfully" | tee -a run.log
        """
}

workflow {
    POCKETMINER(params.pdb)
}