|
| 1 | +""" |
| 2 | +Module for analyzing transcripts using Gemini API and generating DOCX documents. |
| 3 | +""" |
| 4 | + |
| 5 | +import os |
| 6 | +import re |
| 7 | +from docx import Document |
| 8 | +from docx.shared import Pt, RGBColor |
| 9 | +from docx.enum.text import WD_ALIGN_PARAGRAPH |
| 10 | +import google.genai as genai |
| 11 | +from google.genai import types |
| 12 | +from generate_podcast import setup_logging |
| 13 | +from utils import get_asset_path |
| 14 | + |
| 15 | +logger = setup_logging() |
| 16 | + |
| 17 | + |
| 18 | +def get_analysis_prompt_path(): |
| 19 | + """ |
| 20 | + Returns the path to the analysis prompt configuration file. |
| 21 | + Checks in order: |
| 22 | + 1. ./config/analysis_prompt.txt (for Docker and local config directory) |
| 23 | + 2. app data directory (user-editable location) |
| 24 | + 3. asset path (for bundled apps) |
| 25 | + """ |
| 26 | + from utils import get_app_data_dir |
| 27 | + |
| 28 | + # First check in config directory (Docker and local development) |
| 29 | + config_dir_prompt = os.path.join(os.getcwd(), "config", "analysis_prompt.txt") |
| 30 | + if os.path.exists(config_dir_prompt): |
| 31 | + return config_dir_prompt |
| 32 | + |
| 33 | + # Then check in app data directory (user-editable location) |
| 34 | + app_data_prompt = os.path.join(get_app_data_dir(), "analysis_prompt.txt") |
| 35 | + if os.path.exists(app_data_prompt): |
| 36 | + return app_data_prompt |
| 37 | + |
| 38 | + # Finally check in asset path (bundled with app) |
| 39 | + asset_prompt = get_asset_path("analysis_prompt.txt") |
| 40 | + if asset_prompt and os.path.exists(asset_prompt): |
| 41 | + return asset_prompt |
| 42 | + |
| 43 | + return None |
| 44 | + |
| 45 | + |
| 46 | +def load_analysis_prompt(): |
| 47 | + """ |
| 48 | + Loads the analysis prompt from the configuration file. |
| 49 | + Returns None if the file doesn't exist. |
| 50 | + """ |
| 51 | + prompt_path = get_analysis_prompt_path() |
| 52 | + if not prompt_path: |
| 53 | + return None |
| 54 | + |
| 55 | + try: |
| 56 | + with open(prompt_path, 'r', encoding='utf-8') as f: |
| 57 | + return f.read().strip() |
| 58 | + except Exception as e: |
| 59 | + logger.error(f"Error reading analysis prompt file: {e}") |
| 60 | + return None |
| 61 | + |
| 62 | + |
| 63 | +def generate_prompt(text: str) -> str: |
| 64 | + """ |
| 65 | + Génère le prompt pour l'API Gemini. |
| 66 | + Loads the prompt template from the configuration file and appends the transcript. |
| 67 | +
|
| 68 | + Raises: |
| 69 | + ValueError: If the prompt configuration file is not found or cannot be read. |
| 70 | + """ |
| 71 | + prompt_template = load_analysis_prompt() |
| 72 | + if not prompt_template: |
| 73 | + raise ValueError("Analysis prompt configuration file not found. Please create 'analysis_prompt.txt' in the application directory.") |
| 74 | + |
| 75 | + return f"{prompt_template}\n\nTranscript:\n{text}" |
| 76 | + |
| 77 | + |
| 78 | +def analyze_transcript_with_gemini(transcript: str, api_key: str = None) -> str: |
| 79 | + """ |
| 80 | + Analyzes a transcript using the Gemini API. |
| 81 | +
|
| 82 | + Args: |
| 83 | + transcript: The transcript text to analyze |
| 84 | + api_key: Gemini API key (will use environment variable if not provided) |
| 85 | +
|
| 86 | + Returns: |
| 87 | + The analysis response from Gemini |
| 88 | +
|
| 89 | + Raises: |
| 90 | + ValueError: If no API key is found |
| 91 | + Exception: If the API call fails |
| 92 | + """ |
| 93 | + if not api_key: |
| 94 | + api_key = os.environ.get("GEMINI_API_KEY") |
| 95 | + |
| 96 | + if not api_key: |
| 97 | + raise ValueError("GEMINI_API_KEY not found in environment variables") |
| 98 | + |
| 99 | + # Get the model name from environment variable or use default |
| 100 | + model_name = os.environ.get("GEMINI_ANALYSIS_MODEL", "gemini-2.5-flash") |
| 101 | + |
| 102 | + try: |
| 103 | + client = genai.Client(api_key=api_key) |
| 104 | + prompt = generate_prompt(transcript) |
| 105 | + |
| 106 | + response = client.models.generate_content( |
| 107 | + model=model_name, |
| 108 | + contents=prompt, |
| 109 | + config=types.GenerateContentConfig( |
| 110 | + temperature=0.7, |
| 111 | + max_output_tokens=2048, |
| 112 | + ) |
| 113 | + ) |
| 114 | + |
| 115 | + if response.text: |
| 116 | + return response.text |
| 117 | + else: |
| 118 | + raise Exception("Empty response from Gemini API") |
| 119 | + |
| 120 | + except Exception as e: |
| 121 | + logger.error(f"Error analyzing transcript with Gemini: {e}", exc_info=True) |
| 122 | + raise |
| 123 | + |
| 124 | + |
| 125 | +def create_docx_from_analysis(analysis_text: str, output_path: str): |
| 126 | + """ |
| 127 | + Creates a well-formatted DOCX document from the Gemini analysis. |
| 128 | +
|
| 129 | + Args: |
| 130 | + analysis_text: The analysis text from Gemini (with markdown-style formatting) |
| 131 | + output_path: Path where to save the DOCX file |
| 132 | + """ |
| 133 | + doc = Document() |
| 134 | + |
| 135 | + # Set document styles |
| 136 | + style = doc.styles['Normal'] |
| 137 | + style.font.name = 'Calibri' |
| 138 | + style.font.size = Pt(11) |
| 139 | + |
| 140 | + # Process the text line by line |
| 141 | + lines = analysis_text.split('\n') |
| 142 | + |
| 143 | + for line in lines: |
| 144 | + line = line.strip() |
| 145 | + if not line: |
| 146 | + continue |
| 147 | + |
| 148 | + # Check if the line contains bold text (**text**) |
| 149 | + bold_pattern = r'\*\*(.+?)\*\*' |
| 150 | + |
| 151 | + if '**' in line: |
| 152 | + # Create a paragraph |
| 153 | + p = doc.add_paragraph() |
| 154 | + |
| 155 | + # Split the line by bold markers |
| 156 | + parts = re.split(bold_pattern, line) |
| 157 | + |
| 158 | + is_bold = False |
| 159 | + for i, part in enumerate(parts): |
| 160 | + if not part: |
| 161 | + continue |
| 162 | + |
| 163 | + # Alternate between normal and bold |
| 164 | + if i % 2 == 1: # Odd indices are bold |
| 165 | + run = p.add_run(part) |
| 166 | + run.bold = True |
| 167 | + run.font.size = Pt(12) |
| 168 | + else: # Even indices are normal |
| 169 | + run = p.add_run(part) |
| 170 | + run.font.size = Pt(11) |
| 171 | + else: |
| 172 | + # Regular paragraph |
| 173 | + p = doc.add_paragraph(line) |
| 174 | + p.style = 'Normal' |
| 175 | + |
| 176 | + # Save the document |
| 177 | + doc.save(output_path) |
| 178 | + logger.info(f"DOCX document saved to: {output_path}") |
| 179 | + |
| 180 | + |
| 181 | +def generate_analysis_docx(transcript: str, output_path: str = None, api_key: str = None) -> str: |
| 182 | + """ |
| 183 | + Complete workflow: analyze transcript with Gemini and create a DOCX document. |
| 184 | +
|
| 185 | + Args: |
| 186 | + transcript: The transcript text to analyze |
| 187 | + output_path: Path where to save the DOCX file (auto-generated if not provided) |
| 188 | + api_key: Gemini API key (will use environment variable if not provided) |
| 189 | +
|
| 190 | + Returns: |
| 191 | + Path to the generated DOCX file |
| 192 | + """ |
| 193 | + # Analyze the transcript |
| 194 | + analysis = analyze_transcript_with_gemini(transcript, api_key) |
| 195 | + |
| 196 | + # Generate output path if not provided |
| 197 | + if not output_path: |
| 198 | + import tempfile |
| 199 | + output_path = os.path.join(tempfile.gettempdir(), f"script_analysis_{os.urandom(4).hex()}.docx") |
| 200 | + |
| 201 | + # Create the DOCX |
| 202 | + create_docx_from_analysis(analysis, output_path) |
| 203 | + |
| 204 | + return output_path |
0 commit comments