# if using anaconda, before running this, remember to # conda activate gemini_env import time import re import sys from datetime import date from google import genai from google.genai import types # 1. Setup MY_KEY = "youNeedToPutYourOwnAPIkeyHereAndUncommentTheNextLube" #client = genai.Client(api_key=MY_KEY) today = str(date.today()) thePlay = sys.argv[1] def discover_metadata(text_sample): """Retrieves Title and Author directly from the first 2000 chars of text.""" prompt = "Analyze this title page. Return ONLY: Title | Author. Example: A Christmas Carol | C. Z. Barnett" response = client.models.generate_content( model='gemini-2.5-flash', contents=f"{prompt}\n\n{text_sample[:2000]}" ) parts = response.text.strip().split('|') title = parts[0].strip() if len(parts) > 0 else "Unknown Title" author = parts[1].strip() if len(parts) > 1 else "Unknown Author" return title, author def encode_front_matter(text_chunk): """Encodes preamble: titlePage, castList (with ), and .""" system_instruction = ( "You are a TEI P5 expert. Convert this preamble into a element.\n" "1. Wrap title/author in .\n" "2. For 'Dramatis Personae', use .\n" "3. Each entry must be a containing , , and .\n" "4. Use for time/place descriptions.\n" "5. Output ONLY the XML fragment." ) response = client.models.generate_content( model='gemini-2.5-flash', contents=text_chunk, config=types.GenerateContentConfig(system_instruction=system_instruction, temperature=0.1) ) return response.text.strip().replace('```xml', '').replace('```', ''), response.usage_metadata def encode_act(chunk_text): """Encodes an individual Act into a
.""" system_instruction = ( "You are a TEI P5 expert. Convert this Act into a
.\n" "1. Use , ,

, , , and

.\n" "2. Convert ALL CAPS names in stage directions to Title Case.\n" "3. Keep speaker labels literal.\n" "4. Output ONLY the XML
fragment." ) response = client.models.generate_content( model='gemini-2.5-flash', contents=chunk_text, config=types.GenerateContentConfig(system_instruction=system_instruction, temperature=0.1) ) return response.text.strip().replace('```xml', '').replace('```', ''), response.usage_metadata def main(): theFile=thePlay+".txt" print("Processing "+theFile+" on "+today) with open(theFile, "r", encoding="utf-8") as f: full_text = f.read() # Split logic (Case-sensitive ACT) split_pattern = r'(?=ACT\s+(?:THE\s+)?(?:FIRST|SECOND|THIRD|FOURTH|FIFTH|[IVXLCDM]+))' parts = re.split(split_pattern, full_text) front_text = parts[0].strip() acts_text = parts[1:] # Step 1: Discover Metadata for the Header print("Discovering manuscript metadata...") title, author = discover_metadata(full_text) # Step 2: Build the XML Skeleton using discovered data hd1=f""" {title} {author}

To be supplied

To be supplied

 TEI autotagging by Gemini Pro 2.5 """ hdr=hd1+thePlay+hd2+today+hd3 final_xml = [hdr] total_in = 0 total_out = 0 # Step 3: Encode Front Matter if front_text: print(f"Encoding Front Matter for '{title}'...") xml, usage = encode_front_matter(front_text) final_xml.append(xml) total_in += usage.prompt_token_count total_out += usage.candidates_token_count final_xml.append("\n \n") # Step 4: Encode Acts for i, act_text in enumerate(acts_text): print(f"Encoding Act {i+1} of {len(acts_text)}...") xml, usage = encode_act(act_text) final_xml.append(xml) total_in += usage.prompt_token_count total_out += usage.candidates_token_count time.sleep(1) final_xml.append("\n \n \n") # Write and Report with open(thePlay+"-final.xml", "w", encoding="utf-8") as f: f.writelines(final_xml) cost = (total_in/1_000_000 * 0.075) + (total_out/1_000_000 * 0.30) print(f"\nDone. Processed {len(acts_text)} acts from"+thePlay) print(f"Estimated Cost: ${cost:.4f}") if __name__ == "__main__": main()