Smart EHR MCP Server

#!/bin/bash # --- Configuration --- INPUT_DIR="premera_policies" OUTPUT_DIR="premera_policies_abstracted" PROMPT_FILE="abstract.md" # Prompt defining the abstraction task MODEL="gemini-2.5-flash-preview-04-17" # --- End Configuration --- # --- Validate Prompt File --- if [ ! -f "$PROMPT_FILE" ]; then echo "Error: Prompt file '$PROMPT_FILE' not found." exit 1 fi echo "Using prompt file: $PROMPT_FILE" # Read the prompt content PROMPT_TEXT=$(cat "$PROMPT_FILE") if [ -z "$PROMPT_TEXT" ]; then echo "Error: Prompt file '$PROMPT_FILE' is empty." exit 1 fi # --- Process Files --- # Determine files to process: CLI args or default input directory if [ "$#" -gt 0 ]; then files=( "$@" ) # Validate provided files for md_file in "${files[@]}"; do if [ ! -f "$md_file" ]; then echo "Error: File '$md_file' not found or not a regular file." exit 1 fi done else # Ensure the input directory exists if [ ! -d "$INPUT_DIR" ]; then echo "Error: Input directory '$INPUT_DIR' not found." exit 1 fi files=( "$INPUT_DIR"/*.md ) fi # Create the output directory if it doesn't exist mkdir -p "$OUTPUT_DIR" echo "Starting abstraction process..." if [ "$#" -gt 0 ]; then echo "Files to process: ${files[*]}" else echo "Input Directory: $INPUT_DIR" fi echo "Output Directory: $OUTPUT_DIR" echo "Model: $MODEL" echo "-------------------------------------" # Loop through selected Markdown files for md_file in "${files[@]}"; do # Skip if file doesn't exist (handles glob no-match) [ -e "$md_file" ] || continue # Get the base filename without the extension base_name=$(basename "$md_file" .md) output_json_file="$OUTPUT_DIR/$base_name.json" # Output as JSON echo "--- Processing '$md_file' -> '$output_json_file' ---" # Read the markdown file content MARKDOWN_CONTENT=$(cat "$md_file") if [ -z "$MARKDOWN_CONTENT" ]; then echo "Warning: Skipping empty file '$md_file'." continue fi # Combine prompt and markdown content for the LLM input # Use $'\n' for portability across shells for newline COMBINED_INPUT="$PROMPT_TEXT"$'\n\n---\n\n'"$MARKDOWN_CONTENT" echo "$COMBINED_INPUT" echo "(Running LLM command... Output will be displayed below and saved to file)" # Run the LLM command, capture raw output then strip code fences before saving temp_output=$(mktemp) || { echo "Error: Unable to create temporary file"; exit 1; } llm -m "$MODEL" "$COMBINED_INPUT" | tee "$temp_output" pipeline_status=${PIPESTATUS[0]} sed '/^```/d' "$temp_output" > "$output_json_file" rm "$temp_output" if [ $pipeline_status -eq 0 ]; then # Basic JSON validation (optional but helpful) if jq . "$output_json_file" > /dev/null 2>&1; then echo -e "\n--- Finished processing '$md_file'. JSON output saved to '$output_json_file' (Valid JSON) ---" else echo -e "\n--- Warning: Finished processing '$md_file', but output in '$output_json_file' is NOT valid JSON. ---" fi else echo -e "\n--- Error processing '$md_file' (LLM exit code: $pipeline_status). Check output above. '$output_json_file' may be incomplete or invalid. ---" fi # Optional: Add a small delay to avoid hitting rate limits # sleep 1 done echo "-------------------------------------" echo "Abstraction process finished." echo "JSON files are located in '$OUTPUT_DIR'." exit 0