Upload 5 files
Browse files- app.py +455 -0
- config.py +296 -0
- file_handler.py +277 -0
- report_generator.py +399 -0
- transcription.py +287 -0
app.py
ADDED
|
@@ -0,0 +1,455 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import os
|
| 3 |
+
from datetime import datetime
|
| 4 |
+
import time
|
| 5 |
+
import traceback
|
| 6 |
+
|
| 7 |
+
# Import modułów
|
| 8 |
+
from transcription import AudioTranscriber
|
| 9 |
+
from report_generator import ReportGenerator
|
| 10 |
+
from file_handler import FileHandler
|
| 11 |
+
from config import NVIDIA_THEME, DEFAULT_SETTINGS
|
| 12 |
+
|
| 13 |
+
# Konfiguracja strony
|
| 14 |
+
st.set_page_config(
|
| 15 |
+
page_title="FGI/IDI Research Analyzer",
|
| 16 |
+
page_icon="🎙️",
|
| 17 |
+
layout="wide",
|
| 18 |
+
initial_sidebar_state="expanded"
|
| 19 |
+
)
|
| 20 |
+
|
| 21 |
+
# Custom CSS - kolorystyka NVIDIA
|
| 22 |
+
st.markdown(f"""
|
| 23 |
+
<style>
|
| 24 |
+
.main {{
|
| 25 |
+
background-color: {NVIDIA_THEME['background']};
|
| 26 |
+
color: {NVIDIA_THEME['text']};
|
| 27 |
+
}}
|
| 28 |
+
.stButton > button {{
|
| 29 |
+
background-color: {NVIDIA_THEME['primary']};
|
| 30 |
+
color: {NVIDIA_THEME['background']};
|
| 31 |
+
border: none;
|
| 32 |
+
border-radius: 5px;
|
| 33 |
+
font-weight: bold;
|
| 34 |
+
}}
|
| 35 |
+
.stButton > button:hover {{
|
| 36 |
+
background-color: {NVIDIA_THEME['accent']};
|
| 37 |
+
color: {NVIDIA_THEME['background']};
|
| 38 |
+
}}
|
| 39 |
+
.sidebar .sidebar-content {{
|
| 40 |
+
background-color: {NVIDIA_THEME['secondary']};
|
| 41 |
+
}}
|
| 42 |
+
.stProgress > div > div {{
|
| 43 |
+
background-color: {NVIDIA_THEME['primary']};
|
| 44 |
+
}}
|
| 45 |
+
.success-box {{
|
| 46 |
+
background-color: rgba(118, 185, 0, 0.1);
|
| 47 |
+
border: 1px solid {NVIDIA_THEME['primary']};
|
| 48 |
+
border-radius: 5px;
|
| 49 |
+
padding: 10px;
|
| 50 |
+
margin: 10px 0;
|
| 51 |
+
}}
|
| 52 |
+
.error-box {{
|
| 53 |
+
background-color: rgba(255, 0, 0, 0.1);
|
| 54 |
+
border: 1px solid #ff0000;
|
| 55 |
+
border-radius: 5px;
|
| 56 |
+
padding: 10px;
|
| 57 |
+
margin: 10px 0;
|
| 58 |
+
}}
|
| 59 |
+
</style>
|
| 60 |
+
""", unsafe_allow_html=True)
|
| 61 |
+
|
| 62 |
+
class FGIIDIAnalyzer:
|
| 63 |
+
def __init__(self):
|
| 64 |
+
self.transcriber = None
|
| 65 |
+
self.report_generator = None
|
| 66 |
+
self.file_handler = FileHandler()
|
| 67 |
+
self.initialize_session_state()
|
| 68 |
+
|
| 69 |
+
def initialize_session_state(self):
|
| 70 |
+
"""Inicjalizacja session state"""
|
| 71 |
+
if 'transcriptions' not in st.session_state:
|
| 72 |
+
st.session_state.transcriptions = {}
|
| 73 |
+
if 'uploaded_files' not in st.session_state:
|
| 74 |
+
st.session_state.uploaded_files = []
|
| 75 |
+
if 'processing_status' not in st.session_state:
|
| 76 |
+
st.session_state.processing_status = 'ready'
|
| 77 |
+
if 'final_report' not in st.session_state:
|
| 78 |
+
st.session_state.final_report = None
|
| 79 |
+
if 'research_brief' not in st.session_state:
|
| 80 |
+
st.session_state.research_brief = ""
|
| 81 |
+
if 'logs' not in st.session_state:
|
| 82 |
+
st.session_state.logs = []
|
| 83 |
+
|
| 84 |
+
def log_message(self, message, level="INFO"):
|
| 85 |
+
"""Dodaj wiadomość do logów"""
|
| 86 |
+
timestamp = datetime.now().strftime("%H:%M:%S")
|
| 87 |
+
log_entry = f"[{timestamp}] {level}: {message}"
|
| 88 |
+
st.session_state.logs.append(log_entry)
|
| 89 |
+
|
| 90 |
+
# Ograniczenie liczby logów do 100
|
| 91 |
+
if len(st.session_state.logs) > 100:
|
| 92 |
+
st.session_state.logs = st.session_state.logs[-100:]
|
| 93 |
+
|
| 94 |
+
def render_sidebar(self):
|
| 95 |
+
"""Renderuj sidebar z konfiguracją"""
|
| 96 |
+
st.sidebar.title("🎙️ FGI/IDI Analyzer")
|
| 97 |
+
st.sidebar.markdown("---")
|
| 98 |
+
|
| 99 |
+
# API Keys
|
| 100 |
+
st.sidebar.subheader("🔑 Konfiguracja API")
|
| 101 |
+
|
| 102 |
+
openai_key = st.sidebar.text_input(
|
| 103 |
+
"OpenAI API Key:",
|
| 104 |
+
type="password",
|
| 105 |
+
help="Klucz do Whisper (transkrypcja) i GPT-4o-mini (raporty)"
|
| 106 |
+
)
|
| 107 |
+
|
| 108 |
+
if openai_key:
|
| 109 |
+
self.transcriber = AudioTranscriber(openai_key)
|
| 110 |
+
self.report_generator = ReportGenerator(openai_key)
|
| 111 |
+
st.sidebar.success("✅ API połączone")
|
| 112 |
+
else:
|
| 113 |
+
st.sidebar.warning("⚠️ Wprowadź klucz API")
|
| 114 |
+
|
| 115 |
+
st.sidebar.markdown("---")
|
| 116 |
+
|
| 117 |
+
# Ustawienia transkrypcji
|
| 118 |
+
st.sidebar.subheader("⚙️ Ustawienia")
|
| 119 |
+
|
| 120 |
+
max_file_size = st.sidebar.selectbox(
|
| 121 |
+
"Maksymalny rozmiar części:",
|
| 122 |
+
[15, 20, 25, 30],
|
| 123 |
+
index=1,
|
| 124 |
+
help="MB - większe pliki będą dzielone na części"
|
| 125 |
+
)
|
| 126 |
+
|
| 127 |
+
auto_compress = st.sidebar.checkbox(
|
| 128 |
+
"Auto-kompresja dużych plików",
|
| 129 |
+
value=True,
|
| 130 |
+
help="Automatyczna kompresja plików >50MB"
|
| 131 |
+
)
|
| 132 |
+
|
| 133 |
+
language = st.sidebar.selectbox(
|
| 134 |
+
"Język transkrypcji:",
|
| 135 |
+
["pl", "en", "auto"],
|
| 136 |
+
index=0,
|
| 137 |
+
help="Język audio dla Whisper"
|
| 138 |
+
)
|
| 139 |
+
|
| 140 |
+
st.sidebar.markdown("---")
|
| 141 |
+
|
| 142 |
+
# Status systemu
|
| 143 |
+
st.sidebar.subheader("📊 Status")
|
| 144 |
+
|
| 145 |
+
if st.session_state.uploaded_files:
|
| 146 |
+
st.sidebar.info(f"📁 Plików: {len(st.session_state.uploaded_files)}")
|
| 147 |
+
|
| 148 |
+
if st.session_state.transcriptions:
|
| 149 |
+
st.sidebar.info(f"✅ Transkrypcji: {len(st.session_state.transcriptions)}")
|
| 150 |
+
|
| 151 |
+
if st.session_state.final_report:
|
| 152 |
+
st.sidebar.success("📄 Raport gotowy")
|
| 153 |
+
|
| 154 |
+
# Reset session
|
| 155 |
+
if st.sidebar.button("🔄 Reset sesji", type="secondary"):
|
| 156 |
+
for key in list(st.session_state.keys()):
|
| 157 |
+
del st.session_state[key]
|
| 158 |
+
st.rerun()
|
| 159 |
+
|
| 160 |
+
return {
|
| 161 |
+
'openai_key': openai_key,
|
| 162 |
+
'max_file_size': max_file_size,
|
| 163 |
+
'auto_compress': auto_compress,
|
| 164 |
+
'language': language
|
| 165 |
+
}
|
| 166 |
+
|
| 167 |
+
def render_file_upload(self, settings):
|
| 168 |
+
"""Renderuj sekcję upload plików"""
|
| 169 |
+
st.header("📁 Upload plików audio/video")
|
| 170 |
+
|
| 171 |
+
# Research brief
|
| 172 |
+
st.subheader("📋 Brief badawczy (opcjonalny)")
|
| 173 |
+
research_brief = st.text_area(
|
| 174 |
+
"Opisz cele badania, grupę docelową, kluczowe pytania badawcze:",
|
| 175 |
+
value=st.session_state.research_brief,
|
| 176 |
+
height=100,
|
| 177 |
+
help="Ten opis pomoże AI wygenerować lepszy raport"
|
| 178 |
+
)
|
| 179 |
+
st.session_state.research_brief = research_brief
|
| 180 |
+
|
| 181 |
+
# File uploader
|
| 182 |
+
st.subheader("🎙️ Pliki do transkrypcji")
|
| 183 |
+
uploaded_files = st.file_uploader(
|
| 184 |
+
"Wybierz pliki audio/video:",
|
| 185 |
+
type=['mp3', 'wav', 'mp4', 'm4a', 'aac'],
|
| 186 |
+
accept_multiple_files=True,
|
| 187 |
+
help="Obsługiwane formaty: MP3, WAV, MP4, M4A, AAC"
|
| 188 |
+
)
|
| 189 |
+
|
| 190 |
+
if uploaded_files:
|
| 191 |
+
# Walidacja plików
|
| 192 |
+
valid_files = []
|
| 193 |
+
total_size = 0
|
| 194 |
+
|
| 195 |
+
for file in uploaded_files:
|
| 196 |
+
file_size_mb = file.size / (1024 * 1024)
|
| 197 |
+
total_size += file_size_mb
|
| 198 |
+
|
| 199 |
+
# Sprawdź rozmiar pojedynczego pliku
|
| 200 |
+
if file_size_mb > 200: # 200MB limit dla pojedynczego pliku
|
| 201 |
+
st.error(f"❌ {file.name}: Plik za duży ({file_size_mb:.1f}MB). Maksymalnie 200MB.")
|
| 202 |
+
continue
|
| 203 |
+
|
| 204 |
+
valid_files.append({
|
| 205 |
+
'file': file,
|
| 206 |
+
'name': file.name,
|
| 207 |
+
'size_mb': file_size_mb,
|
| 208 |
+
'needs_splitting': file_size_mb > settings['max_file_size']
|
| 209 |
+
})
|
| 210 |
+
|
| 211 |
+
# Wyświetl informacje o plikach
|
| 212 |
+
if valid_files:
|
| 213 |
+
st.success(f"✅ Załadowano {len(valid_files)} plików ({total_size:.1f}MB)")
|
| 214 |
+
|
| 215 |
+
# Tabela z informacjami o plikach
|
| 216 |
+
for i, file_info in enumerate(valid_files):
|
| 217 |
+
col1, col2, col3 = st.columns([3, 1, 1])
|
| 218 |
+
|
| 219 |
+
with col1:
|
| 220 |
+
st.write(f"📄 {file_info['name']}")
|
| 221 |
+
|
| 222 |
+
with col2:
|
| 223 |
+
st.write(f"{file_info['size_mb']:.1f}MB")
|
| 224 |
+
|
| 225 |
+
with col3:
|
| 226 |
+
if file_info['needs_splitting']:
|
| 227 |
+
st.warning("Będzie podzielony")
|
| 228 |
+
else:
|
| 229 |
+
st.success("OK")
|
| 230 |
+
|
| 231 |
+
st.session_state.uploaded_files = valid_files
|
| 232 |
+
return True
|
| 233 |
+
|
| 234 |
+
return False
|
| 235 |
+
|
| 236 |
+
def render_processing_section(self, settings):
|
| 237 |
+
"""Renderuj sekcję przetwarzania"""
|
| 238 |
+
if not st.session_state.uploaded_files:
|
| 239 |
+
st.info("👆 Najpierw załaduj pliki audio/video")
|
| 240 |
+
return
|
| 241 |
+
|
| 242 |
+
if not settings['openai_key']:
|
| 243 |
+
st.warning("⚠️ Wprowadź klucz OpenAI API w sidebarze")
|
| 244 |
+
return
|
| 245 |
+
|
| 246 |
+
st.header("🚀 Przetwarzanie")
|
| 247 |
+
|
| 248 |
+
# Przycisk start
|
| 249 |
+
if st.session_state.processing_status == 'ready':
|
| 250 |
+
if st.button("🎯 Rozpocznij transkrypcję i analizę", type="primary"):
|
| 251 |
+
st.session_state.processing_status = 'running'
|
| 252 |
+
self.process_files(settings)
|
| 253 |
+
|
| 254 |
+
elif st.session_state.processing_status == 'running':
|
| 255 |
+
st.info("⏳ Przetwarzanie w toku...")
|
| 256 |
+
|
| 257 |
+
if st.button("⏹️ Zatrzymaj", type="secondary"):
|
| 258 |
+
st.session_state.processing_status = 'stopped'
|
| 259 |
+
st.warning("Przetwarzanie zatrzymane")
|
| 260 |
+
|
| 261 |
+
# Progress display
|
| 262 |
+
if st.session_state.processing_status == 'running':
|
| 263 |
+
self.render_progress()
|
| 264 |
+
|
| 265 |
+
def render_progress(self):
|
| 266 |
+
"""Renderuj postęp przetwarzania"""
|
| 267 |
+
progress_container = st.container()
|
| 268 |
+
|
| 269 |
+
with progress_container:
|
| 270 |
+
# Overall progress
|
| 271 |
+
total_files = len(st.session_state.uploaded_files)
|
| 272 |
+
completed_files = len(st.session_state.transcriptions)
|
| 273 |
+
|
| 274 |
+
progress = completed_files / total_files if total_files > 0 else 0
|
| 275 |
+
st.progress(progress)
|
| 276 |
+
st.write(f"📊 Postęp ogólny: {completed_files}/{total_files} plików")
|
| 277 |
+
|
| 278 |
+
# Current file info
|
| 279 |
+
if completed_files < total_files:
|
| 280 |
+
current_file = st.session_state.uploaded_files[completed_files]['name']
|
| 281 |
+
st.write(f"🔄 Aktualnie: {current_file}")
|
| 282 |
+
|
| 283 |
+
def process_files(self, settings):
|
| 284 |
+
"""Główna logika przetwarzania plików"""
|
| 285 |
+
try:
|
| 286 |
+
self.log_message("Rozpoczynam przetwarzanie plików")
|
| 287 |
+
|
| 288 |
+
# Container dla live updates
|
| 289 |
+
status_container = st.empty()
|
| 290 |
+
progress_container = st.empty()
|
| 291 |
+
|
| 292 |
+
# 1. Transkrypcja wszystkich plików
|
| 293 |
+
for i, file_info in enumerate(st.session_state.uploaded_files):
|
| 294 |
+
if st.session_state.processing_status != 'running':
|
| 295 |
+
break
|
| 296 |
+
|
| 297 |
+
status_container.info(f"🎙️ Transkrybuję: {file_info['name']}")
|
| 298 |
+
self.log_message(f"Rozpoczynam transkrypcję: {file_info['name']}")
|
| 299 |
+
|
| 300 |
+
try:
|
| 301 |
+
# Przetwórz plik (podział jeśli potrzeba)
|
| 302 |
+
processed_files = self.file_handler.process_file(
|
| 303 |
+
file_info['file'],
|
| 304 |
+
settings['max_file_size'],
|
| 305 |
+
settings['auto_compress']
|
| 306 |
+
)
|
| 307 |
+
|
| 308 |
+
# Transkrypcja
|
| 309 |
+
transcription = self.transcriber.transcribe_files(
|
| 310 |
+
processed_files,
|
| 311 |
+
language=settings['language']
|
| 312 |
+
)
|
| 313 |
+
|
| 314 |
+
st.session_state.transcriptions[file_info['name']] = transcription
|
| 315 |
+
self.log_message(f"✅ Zakończono transkrypcję: {file_info['name']}")
|
| 316 |
+
|
| 317 |
+
# Update progress
|
| 318 |
+
progress = (i + 1) / len(st.session_state.uploaded_files)
|
| 319 |
+
progress_container.progress(progress)
|
| 320 |
+
|
| 321 |
+
except Exception as e:
|
| 322 |
+
self.log_message(f"❌ Błąd transkrypcji {file_info['name']}: {str(e)}", "ERROR")
|
| 323 |
+
st.error(f"Błąd przy {file_info['name']}: {str(e)}")
|
| 324 |
+
|
| 325 |
+
# 2. Generowanie raportu
|
| 326 |
+
if st.session_state.transcriptions and st.session_state.processing_status == 'running':
|
| 327 |
+
status_container.info("📄 Generuję raport badawczy...")
|
| 328 |
+
self.log_message("Rozpoczynam generowanie raportu")
|
| 329 |
+
|
| 330 |
+
try:
|
| 331 |
+
final_report = self.report_generator.generate_comprehensive_report(
|
| 332 |
+
st.session_state.transcriptions,
|
| 333 |
+
st.session_state.research_brief
|
| 334 |
+
)
|
| 335 |
+
|
| 336 |
+
st.session_state.final_report = final_report
|
| 337 |
+
st.session_state.processing_status = 'completed'
|
| 338 |
+
|
| 339 |
+
status_container.success("✅ Przetwarzanie zakończone!")
|
| 340 |
+
self.log_message("✅ Raport wygenerowany pomyślnie")
|
| 341 |
+
|
| 342 |
+
except Exception as e:
|
| 343 |
+
self.log_message(f"❌ Błąd generowania raportu: {str(e)}", "ERROR")
|
| 344 |
+
st.error(f"Błąd generowania raportu: {str(e)}")
|
| 345 |
+
st.session_state.processing_status = 'error'
|
| 346 |
+
|
| 347 |
+
except Exception as e:
|
| 348 |
+
self.log_message(f"💥 Błąd krytyczny: {str(e)}", "ERROR")
|
| 349 |
+
st.error(f"Błąd krytyczny: {str(e)}")
|
| 350 |
+
st.session_state.processing_status = 'error'
|
| 351 |
+
|
| 352 |
+
def render_results(self):
|
| 353 |
+
"""Renderuj wyniki"""
|
| 354 |
+
if not st.session_state.transcriptions and not st.session_state.final_report:
|
| 355 |
+
return
|
| 356 |
+
|
| 357 |
+
st.header("📊 Wyniki")
|
| 358 |
+
|
| 359 |
+
# Tabs dla różnych widoków
|
| 360 |
+
tab1, tab2, tab3 = st.tabs(["📄 Raport", "🎙️ Transkrypcje", "📋 Logi"])
|
| 361 |
+
|
| 362 |
+
with tab1:
|
| 363 |
+
if st.session_state.final_report:
|
| 364 |
+
st.subheader("📄 Raport z badania")
|
| 365 |
+
|
| 366 |
+
# Download button
|
| 367 |
+
if st.download_button(
|
| 368 |
+
label="📥 Pobierz raport (TXT)",
|
| 369 |
+
data=st.session_state.final_report,
|
| 370 |
+
file_name=f"raport_badawczy_{datetime.now().strftime('%Y%m%d_%H%M')}.txt",
|
| 371 |
+
mime="text/plain"
|
| 372 |
+
):
|
| 373 |
+
st.success("✅ Raport pobierany!")
|
| 374 |
+
|
| 375 |
+
# Display report
|
| 376 |
+
st.markdown("---")
|
| 377 |
+
st.markdown(st.session_state.final_report)
|
| 378 |
+
else:
|
| 379 |
+
st.info("Raport będzie dostępny po zakończeniu przetwarzania")
|
| 380 |
+
|
| 381 |
+
with tab2:
|
| 382 |
+
if st.session_state.transcriptions:
|
| 383 |
+
st.subheader("🎙️ Transkrypcje")
|
| 384 |
+
|
| 385 |
+
for filename, transcription in st.session_state.transcriptions.items():
|
| 386 |
+
with st.expander(f"📄 {filename}"):
|
| 387 |
+
st.write(transcription)
|
| 388 |
+
|
| 389 |
+
# Download individual transcription
|
| 390 |
+
st.download_button(
|
| 391 |
+
label=f"📥 Pobierz {filename}",
|
| 392 |
+
data=transcription,
|
| 393 |
+
file_name=f"transkrypcja_{filename}_{datetime.now().strftime('%Y%m%d_%H%M')}.txt",
|
| 394 |
+
mime="text/plain",
|
| 395 |
+
key=f"download_{filename}"
|
| 396 |
+
)
|
| 397 |
+
else:
|
| 398 |
+
st.info("Transkrypcje będą dostępne po przetworzeniu plików")
|
| 399 |
+
|
| 400 |
+
with tab3:
|
| 401 |
+
st.subheader("📋 Logi procesu")
|
| 402 |
+
|
| 403 |
+
if st.session_state.logs:
|
| 404 |
+
# Scroll to bottom option
|
| 405 |
+
if st.button("⬇️ Przewiń na dół"):
|
| 406 |
+
pass # Auto-scroll jest w CSS
|
| 407 |
+
|
| 408 |
+
# Display logs
|
| 409 |
+
logs_text = "\n".join(st.session_state.logs)
|
| 410 |
+
st.text_area(
|
| 411 |
+
"Logi:",
|
| 412 |
+
value=logs_text,
|
| 413 |
+
height=400,
|
| 414 |
+
disabled=True
|
| 415 |
+
)
|
| 416 |
+
else:
|
| 417 |
+
st.info("Logi będą wyświetlane podczas przetwarzania")
|
| 418 |
+
|
| 419 |
+
def run(self):
|
| 420 |
+
"""Główna funkcja aplikacji"""
|
| 421 |
+
# Sidebar
|
| 422 |
+
settings = self.render_sidebar()
|
| 423 |
+
|
| 424 |
+
# Main content
|
| 425 |
+
st.title("🎙️ FGI/IDI Research Analyzer")
|
| 426 |
+
st.markdown("*Automatyczna transkrypcja i analiza wywiadów fokusowych oraz indywidualnych*")
|
| 427 |
+
st.markdown("---")
|
| 428 |
+
|
| 429 |
+
# File upload section
|
| 430 |
+
files_uploaded = self.render_file_upload(settings)
|
| 431 |
+
|
| 432 |
+
st.markdown("---")
|
| 433 |
+
|
| 434 |
+
# Processing section
|
| 435 |
+
self.render_processing_section(settings)
|
| 436 |
+
|
| 437 |
+
st.markdown("---")
|
| 438 |
+
|
| 439 |
+
# Results section
|
| 440 |
+
self.render_results()
|
| 441 |
+
|
| 442 |
+
# Główna aplikacja
|
| 443 |
+
if __name__ == "__main__":
|
| 444 |
+
try:
|
| 445 |
+
app = FGIIDIAnalyzer()
|
| 446 |
+
app.run()
|
| 447 |
+
except Exception as e:
|
| 448 |
+
st.error(f"💥 Błąd aplikacji: {str(e)}")
|
| 449 |
+
st.code(traceback.format_exc())
|
| 450 |
+
|
| 451 |
+
# Log error for debugging
|
| 452 |
+
with open('error_log.txt', 'w', encoding='utf-8') as f:
|
| 453 |
+
f.write(f"Error: {str(e)}\n\nTraceback:\n{traceback.format_exc()}")
|
| 454 |
+
|
| 455 |
+
st.info("Szczegóły błędu zapisane w error_log.txt")
|
config.py
ADDED
|
@@ -0,0 +1,296 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# config.py - Konfiguracja aplikacji FGI/IDI Analyzer
|
| 2 |
+
|
| 3 |
+
# Kolorystyka NVIDIA - Gaming/Tech Style
|
| 4 |
+
NVIDIA_THEME = {
|
| 5 |
+
'primary': '#00FF88', # Bright neon green (akcenty)
|
| 6 |
+
'secondary': '#1B1B1B', # Very dark gray (tło sekcji)
|
| 7 |
+
'background': '#0A0A0A', # Near black (główne tło)
|
| 8 |
+
'text': '#E0E0E0', # Light gray text
|
| 9 |
+
'text_secondary': '#A0A0A0', # Darker gray for secondary text
|
| 10 |
+
'accent': '#00CC66', # Darker green for hover states
|
| 11 |
+
'error': '#FF4444', # Red
|
| 12 |
+
'warning': '#FFAA00', # Orange
|
| 13 |
+
'success': '#00FF88', # Same as primary
|
| 14 |
+
'border': '#333333', # Dark border
|
| 15 |
+
'card_bg': '#151515', # Card backgrounds
|
| 16 |
+
}
|
| 17 |
+
|
| 18 |
+
# Ustawienia domyślne
|
| 19 |
+
DEFAULT_SETTINGS = {
|
| 20 |
+
'max_file_size_mb': 20,
|
| 21 |
+
'max_total_size_mb': 500,
|
| 22 |
+
'supported_formats': ['mp3', 'wav', 'mp4', 'm4a', 'aac'],
|
| 23 |
+
'whisper_model': 'whisper-1',
|
| 24 |
+
'gpt_model': 'gpt-4o-mini',
|
| 25 |
+
'default_language': 'pl',
|
| 26 |
+
'chunk_overlap_seconds': 30,
|
| 27 |
+
'max_retries': 3,
|
| 28 |
+
'timeout_seconds': 300,
|
| 29 |
+
}
|
| 30 |
+
|
| 31 |
+
# Prompty dla różnych etapów raportowania
|
| 32 |
+
REPORT_PROMPTS = {
|
| 33 |
+
'outline_generator': """
|
| 34 |
+
Jesteś ekspertem analizy badań jakościowych. Na podstawie dostarczonych transkrypcji z wywiadów {interview_type} oraz briefu badawczego, stwórz szczegółowy plan raportu badawczego.
|
| 35 |
+
|
| 36 |
+
TRANSKRYPCJE:
|
| 37 |
+
{transcriptions}
|
| 38 |
+
|
| 39 |
+
BRIEF BADAWCZY:
|
| 40 |
+
{brief}
|
| 41 |
+
|
| 42 |
+
ZADANIE:
|
| 43 |
+
Przeanalizuj materiał i stwórz outline raportu, który:
|
| 44 |
+
1. Odpowie na cele badawcze z briefu
|
| 45 |
+
2. Uwzględni specyfikę {interview_type}
|
| 46 |
+
3. Będzie miał logiczną strukturę od ogółu do szczegółu
|
| 47 |
+
4. Pozwoli na głęboką analizę insights
|
| 48 |
+
|
| 49 |
+
WYMAGANIA:
|
| 50 |
+
- Outline powinien mieć 5-8 głównych sekcji
|
| 51 |
+
- Każda sekcja z 3-5 podpunktami
|
| 52 |
+
- Uwzględnij cytaty/przykłady tam gdzie to sensowne
|
| 53 |
+
- Zaplanuj miejsca na insights, wnioski, rekomendacje
|
| 54 |
+
|
| 55 |
+
FORMAT ODPOWIEDZI:
|
| 56 |
+
```
|
| 57 |
+
# OUTLINE RAPORTU
|
| 58 |
+
|
| 59 |
+
## 1. [Nazwa sekcji]
|
| 60 |
+
- [Podpunkt 1]
|
| 61 |
+
- [Podpunkt 2]
|
| 62 |
+
- [Podpunkt 3]
|
| 63 |
+
|
| 64 |
+
## 2. [Nazwa sekcji]
|
| 65 |
+
...
|
| 66 |
+
```
|
| 67 |
+
""",
|
| 68 |
+
|
| 69 |
+
'section_generator': """
|
| 70 |
+
Jesteś ekspertem analizy badań jakościowych. Napisz szczegółową sekcję raportu zgodnie z planem.
|
| 71 |
+
|
| 72 |
+
CONTEXT:
|
| 73 |
+
- Typ wywiadu: {interview_type}
|
| 74 |
+
- Brief badawczy: {brief}
|
| 75 |
+
- Plan całego raportu: {outline}
|
| 76 |
+
|
| 77 |
+
TRANSKRYPCJE:
|
| 78 |
+
{transcriptions}
|
| 79 |
+
|
| 80 |
+
ZADANIE:
|
| 81 |
+
Napisz sekcję: "{section_title}"
|
| 82 |
+
Podpunkty do uwzględnienia: {section_points}
|
| 83 |
+
|
| 84 |
+
WYMAGANIA:
|
| 85 |
+
- Sekcja powinna mieć 800-1500 słów
|
| 86 |
+
- Użyj konkretnych cytatów z transkrypcji
|
| 87 |
+
- Analizuj głęboko, nie tylko opisuj
|
| 88 |
+
- Połącz insights z celami biznesowymi
|
| 89 |
+
- Używaj podtytułów dla czytelności
|
| 90 |
+
- Zachowaj obiektywność ale wyciągnij wnioski
|
| 91 |
+
|
| 92 |
+
STYLE:
|
| 93 |
+
- Profesjonalny ale przystępny język
|
| 94 |
+
- Strukturyzowany, z jasnymi insights
|
| 95 |
+
- Cytaty w cudzysłowach z oznaczeniem respondenta
|
| 96 |
+
- Wnioski poparte danymi z wywiadów
|
| 97 |
+
""",
|
| 98 |
+
|
| 99 |
+
'section_expander': """
|
| 100 |
+
Otrzymałeś sekcję raportu, która jest zbyt krótka i powierzchowna. Twoim zadaniem jest ją znacznie rozszerzyć i pogłębić.
|
| 101 |
+
|
| 102 |
+
OBECNA SEKCJA:
|
| 103 |
+
{current_section}
|
| 104 |
+
|
| 105 |
+
DOSTĘPNE TRANSKRYPCJE:
|
| 106 |
+
{transcriptions}
|
| 107 |
+
|
| 108 |
+
CONTEXT:
|
| 109 |
+
{brief}
|
| 110 |
+
|
| 111 |
+
ZADANIE:
|
| 112 |
+
Rozszerz tę sekcję do 1000-1500 słów poprzez:
|
| 113 |
+
|
| 114 |
+
1. **Pogłębienie analizy** - zadaj sobie pytania:
|
| 115 |
+
- Jakie są głębsze przyczyny tych zachowań/opinii?
|
| 116 |
+
- Jakie wzorce widać w różnych grupach respondentów?
|
| 117 |
+
- Jak to łączy się z celami biznesowymi?
|
| 118 |
+
|
| 119 |
+
2. **Dodanie cytatów** - znajdź w transkrypcjach:
|
| 120 |
+
- Konkretne przykłady wspierające tezy
|
| 121 |
+
- Różnorodne perspektywy respondentów
|
| 122 |
+
- Emocjonalne reakcje i spontaniczne komentarze
|
| 123 |
+
|
| 124 |
+
3. **Strukturyzacja** - podziel na podsekcje:
|
| 125 |
+
- Główne tematy/wątki
|
| 126 |
+
- Segmenty respondentów
|
| 127 |
+
- Konkretne insights
|
| 128 |
+
|
| 129 |
+
4. **Praktyczne wnioski** - dodaj:
|
| 130 |
+
- Implikacje dla biznesu
|
| 131 |
+
- Możliwe działania
|
| 132 |
+
- Ryzyka i szanse
|
| 133 |
+
|
| 134 |
+
WYMAGANIA:
|
| 135 |
+
- Zachowaj oryginalną strukturę ale ją rozbuduj
|
| 136 |
+
- Dodaj minimum 5 cytatów z transkrypcji
|
| 137 |
+
- Każdy wniosek uzasadnij danymi
|
| 138 |
+
- Użyj podtytułów dla czytelności
|
| 139 |
+
""",
|
| 140 |
+
|
| 141 |
+
'final_assembly': """
|
| 142 |
+
Jesteś ekspertem analizy badań jakościowych. Twoim zadaniem jest sfinalizowanie raportu - dodanie wprowadzenia, executive summary i spójne połączenie wszystkich sekcji.
|
| 143 |
+
|
| 144 |
+
SEKCJE RAPORTU:
|
| 145 |
+
{sections}
|
| 146 |
+
|
| 147 |
+
BRIEF BADAWCZY:
|
| 148 |
+
{brief}
|
| 149 |
+
|
| 150 |
+
METADANE:
|
| 151 |
+
- Typ badania: {interview_type}
|
| 152 |
+
- Liczba wywiadów: {interviews_count}
|
| 153 |
+
- Data analizy: {date}
|
| 154 |
+
|
| 155 |
+
ZADANIE:
|
| 156 |
+
Stwórz kompletny raport dodając:
|
| 157 |
+
|
| 158 |
+
1. **EXECUTIVE SUMMARY** (300-500 słów):
|
| 159 |
+
- Główne insights z każdej sekcji
|
| 160 |
+
- Key takeaways dla biznesu
|
| 161 |
+
- Top 3 rekomendacje
|
| 162 |
+
|
| 163 |
+
2. **WPROWADZENIE** (200-300 słów):
|
| 164 |
+
- Cele badania
|
| 165 |
+
- Metodologia
|
| 166 |
+
- Struktura raportu
|
| 167 |
+
|
| 168 |
+
3. **ZAKOŃCZENIE** (300-400 słów):
|
| 169 |
+
- Podsumowanie głównych wniosków
|
| 170 |
+
- Rekomendacje działań
|
| 171 |
+
- Sugerowane dalsze kroki
|
| 172 |
+
|
| 173 |
+
4. **SPÓJNOŚĆ**:
|
| 174 |
+
- Dodaj przejścia między sekcjami
|
| 175 |
+
- Ujednolic style i terminologię
|
| 176 |
+
- Sprawdź logiczny przepływ
|
| 177 |
+
|
| 178 |
+
FORMAT:
|
| 179 |
+
```
|
| 180 |
+
# RAPORT Z BADANIA [TYP]
|
| 181 |
+
|
| 182 |
+
## EXECUTIVE SUMMARY
|
| 183 |
+
[treść]
|
| 184 |
+
|
| 185 |
+
## 1. WPROWADZENIE
|
| 186 |
+
[treść]
|
| 187 |
+
|
| 188 |
+
## 2. METODOLOGIA
|
| 189 |
+
[treść]
|
| 190 |
+
|
| 191 |
+
[SEKCJE GŁÓWNE]
|
| 192 |
+
|
| 193 |
+
## ZAKOŃCZENIE I REKOMENDACJE
|
| 194 |
+
[treść]
|
| 195 |
+
|
| 196 |
+
## APPENDIX
|
| 197 |
+
- Informacje o respondentach
|
| 198 |
+
- Dodatkowe cytaty
|
| 199 |
+
```
|
| 200 |
+
""",
|
| 201 |
+
|
| 202 |
+
'quality_checker': """
|
| 203 |
+
Otrzymałeś sekcję raportu do oceny jakości. Sprawdź czy spełnia standardy profesjonalnego raportu z badań jakościowych.
|
| 204 |
+
|
| 205 |
+
SEKCJA DO OCENY:
|
| 206 |
+
{section}
|
| 207 |
+
|
| 208 |
+
KRYTERIA OCENY:
|
| 209 |
+
1. **Długość**: Czy ma 800+ słów?
|
| 210 |
+
2. **Głębokość**: Czy analizuje przyczyny, nie tylko opisuje?
|
| 211 |
+
3. **Cytaty**: Czy ma konkretne przykłady z wywiadów?
|
| 212 |
+
4. **Struktura**: Czy ma logiczny podział i podtytuły?
|
| 213 |
+
5. **Insights**: Czy wyciąga praktyczne wnioski?
|
| 214 |
+
6. **Biznesowość**: Czy łączy z celami biznesowymi?
|
| 215 |
+
|
| 216 |
+
ZADANIE:
|
| 217 |
+
Oceń sekcję w skali 1-10 za każde kryterium i podaj:
|
| 218 |
+
- Ogólną ocenę
|
| 219 |
+
- Konkretne problemy do poprawy
|
| 220 |
+
- Sugestie rozszerzeń
|
| 221 |
+
|
| 222 |
+
FORMAT:
|
| 223 |
+
```
|
| 224 |
+
OCENA JAKOŚCI:
|
| 225 |
+
- Długość: X/10
|
| 226 |
+
- Głębokość: X/10
|
| 227 |
+
- Cytaty: X/10
|
| 228 |
+
- Struktura: X/10
|
| 229 |
+
- Insights: X/10
|
| 230 |
+
- Biznesowość: X/10
|
| 231 |
+
|
| 232 |
+
ŚREDNIA: X/10
|
| 233 |
+
|
| 234 |
+
PROBLEMY:
|
| 235 |
+
- [konkretny problem 1]
|
| 236 |
+
- [konkretny problem 2]
|
| 237 |
+
|
| 238 |
+
SUGESTIE:
|
| 239 |
+
- [sugestia poprawy 1]
|
| 240 |
+
- [sugestia poprawy 2]
|
| 241 |
+
|
| 242 |
+
CZY WYMAGA POPRAWY: TAK/NIE
|
| 243 |
+
```
|
| 244 |
+
"""
|
| 245 |
+
}
|
| 246 |
+
|
| 247 |
+
# Ustawienia modeli
|
| 248 |
+
MODEL_SETTINGS = {
|
| 249 |
+
'whisper': {
|
| 250 |
+
'model': 'whisper-1',
|
| 251 |
+
'language': 'pl',
|
| 252 |
+
'temperature': 0,
|
| 253 |
+
'max_retries': 3,
|
| 254 |
+
},
|
| 255 |
+
|
| 256 |
+
'gpt': {
|
| 257 |
+
'model': 'gpt-4o-mini',
|
| 258 |
+
'temperature': 0.3,
|
| 259 |
+
'max_tokens': 4000,
|
| 260 |
+
'max_retries': 3,
|
| 261 |
+
'timeout': 300,
|
| 262 |
+
}
|
| 263 |
+
}
|
| 264 |
+
|
| 265 |
+
# Mapowanie typów wywiadów
|
| 266 |
+
INTERVIEW_TYPES = {
|
| 267 |
+
'fgi': 'Focus Group Interview (wywiad fokusowy)',
|
| 268 |
+
'idi': 'In-Depth Interview (wywiad indywidualny)',
|
| 269 |
+
'auto': 'automatyczne rozpoznanie na podstawie treści'
|
| 270 |
+
}
|
| 271 |
+
|
| 272 |
+
# Ustawienia przetwarzania plików
|
| 273 |
+
FILE_PROCESSING = {
|
| 274 |
+
'max_single_file_mb': 200,
|
| 275 |
+
'chunk_size_mb': 20,
|
| 276 |
+
'supported_audio_formats': ['mp3', 'wav', 'm4a', 'aac'],
|
| 277 |
+
'supported_video_formats': ['mp4', 'mov', 'avi'],
|
| 278 |
+
'compression_quality': 64, # kbps dla audio
|
| 279 |
+
'sample_rate': 16000, # Hz
|
| 280 |
+
}
|
| 281 |
+
|
| 282 |
+
# Komunikaty dla użytkownika
|
| 283 |
+
USER_MESSAGES = {
|
| 284 |
+
'file_too_large': "Plik {filename} jest za duży ({size}MB). Maksymalnie {max_size}MB. Czy chcesz go skompresować automatycznie?",
|
| 285 |
+
'compression_success': "✅ Plik {filename} skompresowany z {old_size}MB do {new_size}MB",
|
| 286 |
+
'transcription_start': "🎙️ Rozpoczynam transkrypcję: {filename}",
|
| 287 |
+
'transcription_success': "✅ Transkrypcja zakończona: {filename}",
|
| 288 |
+
'transcription_error': "❌ Błąd transkrypcji {filename}: {error}",
|
| 289 |
+
'report_generation_start': "📄 Generuję raport badawczy...",
|
| 290 |
+
'report_section_done': "✅ Sekcja '{section}' wygenerowana",
|
| 291 |
+
'report_complete': "🎉 Raport badawczy gotowy!",
|
| 292 |
+
'api_key_missing': "⚠️ Wprowadź klucz OpenAI API",
|
| 293 |
+
'processing_stopped': "⏹️ Przetwarzanie zatrzymane przez użytkownika",
|
| 294 |
+
'no_files_uploaded': "📁 Nie załadowano żadnych plików",
|
| 295 |
+
'session_reset': "🔄 Sesja została zresetowana"
|
| 296 |
+
}
|
file_handler.py
ADDED
|
@@ -0,0 +1,277 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# file_handler.py - Obsługa plików audio/video dla HuggingFace
|
| 2 |
+
|
| 3 |
+
import os
|
| 4 |
+
import tempfile
|
| 5 |
+
import math
|
| 6 |
+
from io import BytesIO
|
| 7 |
+
from typing import List, Dict, Tuple, Union
|
| 8 |
+
import streamlit as st
|
| 9 |
+
|
| 10 |
+
try:
|
| 11 |
+
from pydub import AudioSegment
|
| 12 |
+
PYDUB_AVAILABLE = True
|
| 13 |
+
except ImportError:
|
| 14 |
+
PYDUB_AVAILABLE = False
|
| 15 |
+
st.warning("⚠️ Pydub nie jest dostępny. Zainstaluj: pip install pydub")
|
| 16 |
+
|
| 17 |
+
try:
|
| 18 |
+
import librosa
|
| 19 |
+
import soundfile as sf
|
| 20 |
+
LIBROSA_AVAILABLE = True
|
| 21 |
+
except ImportError:
|
| 22 |
+
LIBROSA_AVAILABLE = False
|
| 23 |
+
|
| 24 |
+
from config import FILE_PROCESSING, USER_MESSAGES
|
| 25 |
+
|
| 26 |
+
class FileHandler:
|
| 27 |
+
"""Klasa do obsługi plików audio/video - optymalizowana dla HuggingFace"""
|
| 28 |
+
|
| 29 |
+
def __init__(self):
|
| 30 |
+
self.temp_files = [] # Lista plików tymczasowych do wyczyszczenia
|
| 31 |
+
self.processing_stats = {}
|
| 32 |
+
|
| 33 |
+
def process_file(self, uploaded_file, max_chunk_size_mb: int = 20, auto_compress: bool = True) -> List[str]:
|
| 34 |
+
"""
|
| 35 |
+
Główna funkcja przetwarzania pliku
|
| 36 |
+
Returns: Lista ścieżek do plików gotowych do transkrypcji
|
| 37 |
+
"""
|
| 38 |
+
try:
|
| 39 |
+
file_size_mb = uploaded_file.size / (1024 * 1024)
|
| 40 |
+
|
| 41 |
+
# Loguj rozpoczęcie przetwarzania
|
| 42 |
+
st.info(f"🔄 Przetwarzam {uploaded_file.name} ({file_size_mb:.1f}MB)")
|
| 43 |
+
|
| 44 |
+
# Sprawdź czy plik wymaga kompresji
|
| 45 |
+
if file_size_mb > 50 and auto_compress:
|
| 46 |
+
compressed_file = self._compress_audio(uploaded_file)
|
| 47 |
+
if compressed_file:
|
| 48 |
+
uploaded_file = compressed_file
|
| 49 |
+
file_size_mb = compressed_file.size / (1024 * 1024)
|
| 50 |
+
st.success(f"✅ Skompresowano do {file_size_mb:.1f}MB")
|
| 51 |
+
|
| 52 |
+
# Sprawdź czy plik wymaga dzielenia
|
| 53 |
+
if file_size_mb > max_chunk_size_mb:
|
| 54 |
+
return self._split_audio_file(uploaded_file, max_chunk_size_mb)
|
| 55 |
+
else:
|
| 56 |
+
# Plik nie wymaga dzielenia - zapisz bezpośrednio
|
| 57 |
+
temp_path = self._save_temp_file(uploaded_file)
|
| 58 |
+
return [temp_path]
|
| 59 |
+
|
| 60 |
+
except Exception as e:
|
| 61 |
+
st.error(f"❌ Błąd przetwarzania {uploaded_file.name}: {str(e)}")
|
| 62 |
+
return []
|
| 63 |
+
|
| 64 |
+
def _compress_audio(self, uploaded_file) -> Union[BytesIO, None]:
|
| 65 |
+
"""Kompresja pliku audio używając pydub"""
|
| 66 |
+
if not PYDUB_AVAILABLE:
|
| 67 |
+
st.warning("Pydub niedostępny - pomijam kompresję")
|
| 68 |
+
return None
|
| 69 |
+
|
| 70 |
+
try:
|
| 71 |
+
# Załaduj audio
|
| 72 |
+
audio_data = uploaded_file.read()
|
| 73 |
+
audio = AudioSegment.from_file(BytesIO(audio_data))
|
| 74 |
+
|
| 75 |
+
# Kompresja: mono, lower bitrate, lower sample rate
|
| 76 |
+
compressed = audio.set_channels(1) # Mono
|
| 77 |
+
compressed = compressed.set_frame_rate(16000) # 16kHz (wystarczy dla mowy)
|
| 78 |
+
|
| 79 |
+
# Export do BytesIO
|
| 80 |
+
output = BytesIO()
|
| 81 |
+
compressed.export(
|
| 82 |
+
output,
|
| 83 |
+
format="mp3",
|
| 84 |
+
bitrate="64k", # Niska jakość dla kompresji
|
| 85 |
+
parameters=["-ac", "1"] # Force mono
|
| 86 |
+
)
|
| 87 |
+
output.seek(0)
|
| 88 |
+
|
| 89 |
+
# Stwórz nowy "uploaded file" object
|
| 90 |
+
output.name = uploaded_file.name.replace('.', '_compressed.')
|
| 91 |
+
output.size = len(output.getvalue())
|
| 92 |
+
|
| 93 |
+
return output
|
| 94 |
+
|
| 95 |
+
except Exception as e:
|
| 96 |
+
st.warning(f"Kompresja nieudana: {str(e)}")
|
| 97 |
+
return None
|
| 98 |
+
|
| 99 |
+
def _split_audio_file(self, uploaded_file, max_chunk_size_mb: int) -> List[str]:
|
| 100 |
+
"""Dzieli plik audio na mniejsze części"""
|
| 101 |
+
try:
|
| 102 |
+
if not PYDUB_AVAILABLE:
|
| 103 |
+
st.error("❌ Pydub wymagany do dzielenia plików. Zainstaluj: pip install pydub")
|
| 104 |
+
return []
|
| 105 |
+
|
| 106 |
+
# Załaduj cały plik audio
|
| 107 |
+
audio_data = uploaded_file.read()
|
| 108 |
+
audio = AudioSegment.from_file(BytesIO(audio_data))
|
| 109 |
+
|
| 110 |
+
# Oblicz parametry dzielenia
|
| 111 |
+
total_duration_ms = len(audio)
|
| 112 |
+
file_size_mb = uploaded_file.size / (1024 * 1024)
|
| 113 |
+
|
| 114 |
+
# Estymacja liczby części na podstawie rozmiaru
|
| 115 |
+
estimated_parts = math.ceil(file_size_mb / max_chunk_size_mb)
|
| 116 |
+
chunk_duration_ms = total_duration_ms // estimated_parts
|
| 117 |
+
|
| 118 |
+
# Dodaj overlap między częściami (30 sekund)
|
| 119 |
+
overlap_ms = 30 * 1000
|
| 120 |
+
|
| 121 |
+
st.info(f"📂 Dzielę na {estimated_parts} części (~{chunk_duration_ms//60000:.1f} min każda)")
|
| 122 |
+
|
| 123 |
+
parts = []
|
| 124 |
+
base_name = os.path.splitext(uploaded_file.name)[0]
|
| 125 |
+
|
| 126 |
+
for i in range(estimated_parts):
|
| 127 |
+
start_ms = max(0, i * chunk_duration_ms - overlap_ms if i > 0 else 0)
|
| 128 |
+
end_ms = min(total_duration_ms, (i + 1) * chunk_duration_ms + overlap_ms)
|
| 129 |
+
|
| 130 |
+
# Wytnij część
|
| 131 |
+
chunk = audio[start_ms:end_ms]
|
| 132 |
+
|
| 133 |
+
# Zapisz do pliku tymczasowego
|
| 134 |
+
temp_fd, temp_path = tempfile.mkstemp(suffix=f"_part{i+1:02d}.mp3", prefix=f"{base_name}_")
|
| 135 |
+
os.close(temp_fd)
|
| 136 |
+
|
| 137 |
+
chunk.export(temp_path, format="mp3", bitrate="128k")
|
| 138 |
+
parts.append(temp_path)
|
| 139 |
+
self.temp_files.append(temp_path)
|
| 140 |
+
|
| 141 |
+
st.success(f"✅ Część {i+1}/{estimated_parts}: {(end_ms-start_ms)//60000:.1f} min")
|
| 142 |
+
|
| 143 |
+
return parts
|
| 144 |
+
|
| 145 |
+
except Exception as e:
|
| 146 |
+
st.error(f"❌ Błąd dzielenia pliku: {str(e)}")
|
| 147 |
+
return []
|
| 148 |
+
|
| 149 |
+
def _save_temp_file(self, uploaded_file) -> str:
|
| 150 |
+
"""Zapisuje uploaded file do pliku tymczasowego"""
|
| 151 |
+
try:
|
| 152 |
+
# Stwórz plik tymczasowy
|
| 153 |
+
suffix = f".{uploaded_file.name.split('.')[-1]}"
|
| 154 |
+
temp_fd, temp_path = tempfile.mkstemp(suffix=suffix)
|
| 155 |
+
|
| 156 |
+
# Zapisz dane
|
| 157 |
+
with os.fdopen(temp_fd, 'wb') as tmp_file:
|
| 158 |
+
tmp_file.write(uploaded_file.read())
|
| 159 |
+
|
| 160 |
+
self.temp_files.append(temp_path)
|
| 161 |
+
return temp_path
|
| 162 |
+
|
| 163 |
+
except Exception as e:
|
| 164 |
+
st.error(f"❌ Błąd zapisu tymczasowego: {str(e)}")
|
| 165 |
+
return ""
|
| 166 |
+
|
| 167 |
+
def get_audio_duration(self, file_path: str) -> float:
|
| 168 |
+
"""Pobierz długość pliku audio w sekundach"""
|
| 169 |
+
try:
|
| 170 |
+
if LIBROSA_AVAILABLE:
|
| 171 |
+
duration = librosa.get_duration(filename=file_path)
|
| 172 |
+
return duration
|
| 173 |
+
elif PYDUB_AVAILABLE:
|
| 174 |
+
audio = AudioSegment.from_file(file_path)
|
| 175 |
+
return len(audio) / 1000.0 # Convert ms to seconds
|
| 176 |
+
else:
|
| 177 |
+
# Fallback - estymacja na podstawie rozmiaru
|
| 178 |
+
file_size = os.path.getsize(file_path)
|
| 179 |
+
# Przybliżenie: 1MB ≈ 60 sekund dla typowego audio MP3
|
| 180 |
+
return file_size / (1024 * 1024) * 60
|
| 181 |
+
except:
|
| 182 |
+
# Ostateczny fallback
|
| 183 |
+
file_size = os.path.getsize(file_path)
|
| 184 |
+
return file_size / (1024 * 1024) * 60
|
| 185 |
+
|
| 186 |
+
def validate_file(self, uploaded_file) -> Tuple[bool, str]:
|
| 187 |
+
"""Walidacja pliku audio/video"""
|
| 188 |
+
try:
|
| 189 |
+
# Sprawdź rozmiar
|
| 190 |
+
file_size_mb = uploaded_file.size / (1024 * 1024)
|
| 191 |
+
if file_size_mb > FILE_PROCESSING['max_single_file_mb']:
|
| 192 |
+
return False, f"Plik za duży: {file_size_mb:.1f}MB > {FILE_PROCESSING['max_single_file_mb']}MB"
|
| 193 |
+
|
| 194 |
+
# Sprawdź rozszerzenie
|
| 195 |
+
file_ext = uploaded_file.name.split('.')[-1].lower()
|
| 196 |
+
supported_formats = (
|
| 197 |
+
FILE_PROCESSING['supported_audio_formats'] +
|
| 198 |
+
FILE_PROCESSING['supported_video_formats']
|
| 199 |
+
)
|
| 200 |
+
|
| 201 |
+
if file_ext not in supported_formats:
|
| 202 |
+
return False, f"Nieobsługiwany format: .{file_ext}"
|
| 203 |
+
|
| 204 |
+
# Sprawdź czy plik nie jest pusty
|
| 205 |
+
if uploaded_file.size == 0:
|
| 206 |
+
return False, "Plik jest pusty"
|
| 207 |
+
|
| 208 |
+
return True, "OK"
|
| 209 |
+
|
| 210 |
+
except Exception as e:
|
| 211 |
+
return False, f"Błąd walidacji: {str(e)}"
|
| 212 |
+
|
| 213 |
+
def estimate_processing_time(self, uploaded_files: List) -> Dict:
|
| 214 |
+
"""Estymuj czas przetwarzania"""
|
| 215 |
+
total_size_mb = sum(f.size for f in uploaded_files) / (1024 * 1024)
|
| 216 |
+
total_duration_est = total_size_mb * 60 # 1MB ≈ 60s audio
|
| 217 |
+
|
| 218 |
+
# Estymacja czasu transkrypcji (Whisper ~1:10 ratio)
|
| 219 |
+
transcription_time = total_duration_est * 1.1
|
| 220 |
+
|
| 221 |
+
# Estymacja czasu generowania raportu (zależnie od liczby wywiadów)
|
| 222 |
+
report_time = len(uploaded_files) * 30 # ~30s per interview dla raportu
|
| 223 |
+
|
| 224 |
+
return {
|
| 225 |
+
'total_size_mb': total_size_mb,
|
| 226 |
+
'estimated_audio_duration': total_duration_est,
|
| 227 |
+
'estimated_transcription_time': transcription_time,
|
| 228 |
+
'estimated_report_time': report_time,
|
| 229 |
+
'total_estimated_time': transcription_time + report_time
|
| 230 |
+
}
|
| 231 |
+
|
| 232 |
+
def get_file_info(self, uploaded_file) -> Dict:
|
| 233 |
+
"""Pobierz informacje o pliku"""
|
| 234 |
+
file_size_mb = uploaded_file.size / (1024 * 1024)
|
| 235 |
+
file_ext = uploaded_file.name.split('.')[-1].lower()
|
| 236 |
+
|
| 237 |
+
return {
|
| 238 |
+
'name': uploaded_file.name,
|
| 239 |
+
'size_mb': file_size_mb,
|
| 240 |
+
'format': file_ext,
|
| 241 |
+
'needs_compression': file_size_mb > 50,
|
| 242 |
+
'needs_splitting': file_size_mb > 20,
|
| 243 |
+
'estimated_duration': file_size_mb * 60 # Rough estimate
|
| 244 |
+
}
|
| 245 |
+
|
| 246 |
+
def cleanup_temp_files(self):
|
| 247 |
+
"""Wyczyść pliki tymczasowe"""
|
| 248 |
+
cleaned = 0
|
| 249 |
+
for temp_file in self.temp_files:
|
| 250 |
+
try:
|
| 251 |
+
if os.path.exists(temp_file):
|
| 252 |
+
os.remove(temp_file)
|
| 253 |
+
cleaned += 1
|
| 254 |
+
except Exception as e:
|
| 255 |
+
st.warning(f"Nie można usunąć {temp_file}: {e}")
|
| 256 |
+
|
| 257 |
+
self.temp_files = []
|
| 258 |
+
if cleaned > 0:
|
| 259 |
+
st.success(f"🧹 Wyczyszczono {cleaned} plików tymczasowych")
|
| 260 |
+
|
| 261 |
+
def get_processing_stats(self) -> Dict:
|
| 262 |
+
"""Zwróć statystyki przetwarzania"""
|
| 263 |
+
return {
|
| 264 |
+
'temp_files_count': len(self.temp_files),
|
| 265 |
+
'processing_stats': self.processing_stats,
|
| 266 |
+
'libraries_available': {
|
| 267 |
+
'pydub': PYDUB_AVAILABLE,
|
| 268 |
+
'librosa': LIBROSA_AVAILABLE
|
| 269 |
+
}
|
| 270 |
+
}
|
| 271 |
+
|
| 272 |
+
# Test funkcji
|
| 273 |
+
if __name__ == "__main__":
|
| 274 |
+
print("🧪 Test FileHandler")
|
| 275 |
+
handler = FileHandler()
|
| 276 |
+
print(f"📊 Dostępne biblioteki: {handler.get_processing_stats()['libraries_available']}")
|
| 277 |
+
print("✅ FileHandler gotowy do użycia")
|
report_generator.py
ADDED
|
@@ -0,0 +1,399 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# report_generator.py - Inteligentny generator raportów z self-prompting
|
| 2 |
+
|
| 3 |
+
import time
|
| 4 |
+
import streamlit as st
|
| 5 |
+
from typing import Dict, List, Optional, Tuple
|
| 6 |
+
from datetime import datetime
|
| 7 |
+
|
| 8 |
+
try:
|
| 9 |
+
from openai import OpenAI
|
| 10 |
+
OPENAI_AVAILABLE = True
|
| 11 |
+
except ImportError:
|
| 12 |
+
OPENAI_AVAILABLE = False
|
| 13 |
+
st.error("❌ OpenAI library nie jest dostępna")
|
| 14 |
+
|
| 15 |
+
from config import REPORT_PROMPTS, MODEL_SETTINGS, INTERVIEW_TYPES
|
| 16 |
+
|
| 17 |
+
class ReportGenerator:
|
| 18 |
+
"""Inteligentny generator długich raportów badawczych z self-prompting"""
|
| 19 |
+
|
| 20 |
+
def __init__(self, api_key: str):
|
| 21 |
+
if not OPENAI_AVAILABLE:
|
| 22 |
+
raise Exception("OpenAI library nie jest dostępna")
|
| 23 |
+
|
| 24 |
+
self.client = OpenAI(api_key=api_key)
|
| 25 |
+
self.api_key = api_key
|
| 26 |
+
self.generation_stats = {
|
| 27 |
+
'sections_generated': 0,
|
| 28 |
+
'sections_expanded': 0,
|
| 29 |
+
'total_tokens_used': 0,
|
| 30 |
+
'total_cost_estimate': 0,
|
| 31 |
+
'generation_time': 0
|
| 32 |
+
}
|
| 33 |
+
|
| 34 |
+
def generate_comprehensive_report(self, transcriptions: Dict[str, str], brief: str = "") -> str:
|
| 35 |
+
"""
|
| 36 |
+
Główna funkcja generowania kompletnego raportu
|
| 37 |
+
Używa strategii wieloetapowej z self-prompting
|
| 38 |
+
"""
|
| 39 |
+
start_time = time.time()
|
| 40 |
+
|
| 41 |
+
try:
|
| 42 |
+
st.info("📋 Rozpoczynam generowanie raportu...")
|
| 43 |
+
|
| 44 |
+
# Przygotuj dane
|
| 45 |
+
combined_transcriptions = self._combine_transcriptions(transcriptions)
|
| 46 |
+
interview_type = self._detect_interview_type(combined_transcriptions)
|
| 47 |
+
|
| 48 |
+
st.info(f"🔍 Wykryto typ: {INTERVIEW_TYPES.get(interview_type, 'nieznany')}")
|
| 49 |
+
|
| 50 |
+
# ETAP 1: Generowanie outline'u
|
| 51 |
+
st.info("📝 Etap 1/4: Tworzenie struktury raportu...")
|
| 52 |
+
outline = self._generate_outline(combined_transcriptions, brief, interview_type)
|
| 53 |
+
|
| 54 |
+
if not outline:
|
| 55 |
+
raise Exception("Nie udało się wygenerować struktury raportu")
|
| 56 |
+
|
| 57 |
+
# ETAP 2: Generowanie sekcji po sekcji
|
| 58 |
+
st.info("✍️ Etap 2/4: Generowanie treści sekcji...")
|
| 59 |
+
sections = self._generate_sections_iteratively(
|
| 60 |
+
outline, combined_transcriptions, brief, interview_type
|
| 61 |
+
)
|
| 62 |
+
|
| 63 |
+
# ETAP 3: Rozszerzanie zbyt krótkich sekcji (self-prompting)
|
| 64 |
+
st.info("🔍 Etap 3/4: Pogłębianie analizy...")
|
| 65 |
+
expanded_sections = self._expand_short_sections(
|
| 66 |
+
sections, combined_transcriptions, brief
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
# ETAP 4: Finalne scalenie z wprowadzeniem i podsumowaniem
|
| 70 |
+
st.info("📄 Etap 4/4: Finalne scalenie...")
|
| 71 |
+
final_report = self._assemble_final_report(
|
| 72 |
+
expanded_sections, brief, interview_type, len(transcriptions)
|
| 73 |
+
)
|
| 74 |
+
|
| 75 |
+
# Statystyki
|
| 76 |
+
self.generation_stats['generation_time'] = time.time() - start_time
|
| 77 |
+
|
| 78 |
+
st.success(f"🎉 Raport wygenerowany! ({self.generation_stats['generation_time']:.1f}s)")
|
| 79 |
+
self._log_generation_stats()
|
| 80 |
+
|
| 81 |
+
return final_report
|
| 82 |
+
|
| 83 |
+
except Exception as e:
|
| 84 |
+
st.error(f"❌ Błąd generowania raportu: {str(e)}")
|
| 85 |
+
raise e
|
| 86 |
+
|
| 87 |
+
def _combine_transcriptions(self, transcriptions: Dict[str, str]) -> str:
|
| 88 |
+
"""Połącz wszystkie transkrypcje w jeden tekst z oznaczeniami"""
|
| 89 |
+
combined = []
|
| 90 |
+
|
| 91 |
+
for i, (filename, transcription) in enumerate(transcriptions.items(), 1):
|
| 92 |
+
header = f"\n\n=== WYWIAD {i}: {filename} ===\n\n"
|
| 93 |
+
combined.append(header + transcription)
|
| 94 |
+
|
| 95 |
+
return "\n".join(combined)
|
| 96 |
+
|
| 97 |
+
def _detect_interview_type(self, transcriptions: str) -> str:
|
| 98 |
+
"""Automatyczne rozpoznanie typu wywiadu"""
|
| 99 |
+
text_lower = transcriptions.lower()
|
| 100 |
+
|
| 101 |
+
# Wskaźniki FGI
|
| 102 |
+
fgi_indicators = [
|
| 103 |
+
'moderator', 'grupa', 'wszyscy', 'uczestnicy', 'dyskusja',
|
| 104 |
+
'czy zgadzacie się', 'co myślicie', 'focus group'
|
| 105 |
+
]
|
| 106 |
+
|
| 107 |
+
# Wskaźniki IDI
|
| 108 |
+
idi_indicators = [
|
| 109 |
+
'wywiad indywidualny', 'jeden na jeden', 'interviewer',
|
| 110 |
+
'opowiedz mi', 'jak się czujesz', 'twoje doświadczenie'
|
| 111 |
+
]
|
| 112 |
+
|
| 113 |
+
fgi_score = sum(1 for indicator in fgi_indicators if indicator in text_lower)
|
| 114 |
+
idi_score = sum(1 for indicator in idi_indicators if indicator in text_lower)
|
| 115 |
+
|
| 116 |
+
if fgi_score > idi_score:
|
| 117 |
+
return 'fgi'
|
| 118 |
+
elif idi_score > fgi_score:
|
| 119 |
+
return 'idi'
|
| 120 |
+
else:
|
| 121 |
+
return 'auto'
|
| 122 |
+
|
| 123 |
+
def _generate_outline(self, transcriptions: str, brief: str, interview_type: str) -> Dict:
|
| 124 |
+
"""Generuj strukturę raportu"""
|
| 125 |
+
try:
|
| 126 |
+
prompt = REPORT_PROMPTS['outline_generator'].format(
|
| 127 |
+
transcriptions=transcriptions[:8000], # Limit dla API
|
| 128 |
+
brief=brief or "Brak szczegółowego briefu",
|
| 129 |
+
interview_type=INTERVIEW_TYPES.get(interview_type, 'wywiad')
|
| 130 |
+
)
|
| 131 |
+
|
| 132 |
+
response = self._call_gpt(prompt)
|
| 133 |
+
outline = self._parse_outline(response)
|
| 134 |
+
|
| 135 |
+
st.success(f"✅ Outline: {len(outline)} sekcji zaplanowanych")
|
| 136 |
+
return outline
|
| 137 |
+
|
| 138 |
+
except Exception as e:
|
| 139 |
+
st.error(f"❌ Błąd generowania outline: {e}")
|
| 140 |
+
return {}
|
| 141 |
+
|
| 142 |
+
def _generate_sections_iteratively(self, outline: Dict, transcriptions: str, brief: str, interview_type: str) -> Dict:
|
| 143 |
+
"""Generuj sekcje raportu jedna po drugiej"""
|
| 144 |
+
sections = {}
|
| 145 |
+
|
| 146 |
+
for section_title, section_points in outline.items():
|
| 147 |
+
if not section_title or section_title.startswith('#'):
|
| 148 |
+
continue
|
| 149 |
+
|
| 150 |
+
st.info(f"📝 Generuję: {section_title}")
|
| 151 |
+
|
| 152 |
+
try:
|
| 153 |
+
prompt = REPORT_PROMPTS['section_generator'].format(
|
| 154 |
+
transcriptions=transcriptions,
|
| 155 |
+
brief=brief or "Brak szczegółowego briefu",
|
| 156 |
+
interview_type=INTERVIEW_TYPES.get(interview_type, 'wywiad'),
|
| 157 |
+
outline=str(outline),
|
| 158 |
+
section_title=section_title,
|
| 159 |
+
section_points=section_points
|
| 160 |
+
)
|
| 161 |
+
|
| 162 |
+
section_content = self._call_gpt(prompt)
|
| 163 |
+
sections[section_title] = section_content
|
| 164 |
+
|
| 165 |
+
self.generation_stats['sections_generated'] += 1
|
| 166 |
+
st.success(f"✅ {section_title} ({len(section_content.split())} słów)")
|
| 167 |
+
|
| 168 |
+
# Krótka przerwa żeby nie przekroczyć rate limits
|
| 169 |
+
time.sleep(2)
|
| 170 |
+
|
| 171 |
+
except Exception as e:
|
| 172 |
+
st.warning(f"⚠️ Błąd sekcji '{section_title}': {e}")
|
| 173 |
+
sections[section_title] = f"[BŁĄD GENEROWANIA SEKCJI: {e}]"
|
| 174 |
+
|
| 175 |
+
return sections
|
| 176 |
+
|
| 177 |
+
def _expand_short_sections(self, sections: Dict, transcriptions: str, brief: str) -> Dict:
|
| 178 |
+
"""Self-prompting: rozszerz zbyt krótkie sekcje"""
|
| 179 |
+
expanded_sections = {}
|
| 180 |
+
|
| 181 |
+
for section_title, section_content in sections.items():
|
| 182 |
+
word_count = len(section_content.split())
|
| 183 |
+
|
| 184 |
+
# Sprawdź czy sekcja wymaga rozszerzenia
|
| 185 |
+
if word_count < 500: # Za krótka sekcja
|
| 186 |
+
st.info(f"🔍 Rozszerzam: {section_title} ({word_count} słów)")
|
| 187 |
+
|
| 188 |
+
try:
|
| 189 |
+
prompt = REPORT_PROMPTS['section_expander'].format(
|
| 190 |
+
current_section=section_content,
|
| 191 |
+
transcriptions=transcriptions,
|
| 192 |
+
brief=brief or "Brak szczegółowego briefu"
|
| 193 |
+
)
|
| 194 |
+
|
| 195 |
+
expanded_content = self._call_gpt(prompt)
|
| 196 |
+
expanded_sections[section_title] = expanded_content
|
| 197 |
+
|
| 198 |
+
new_word_count = len(expanded_content.split())
|
| 199 |
+
self.generation_stats['sections_expanded'] += 1
|
| 200 |
+
|
| 201 |
+
st.success(f"✅ Rozszerzone: {section_title} ({word_count} → {new_word_count} słów)")
|
| 202 |
+
|
| 203 |
+
time.sleep(2) # Rate limit protection
|
| 204 |
+
|
| 205 |
+
except Exception as e:
|
| 206 |
+
st.warning(f"⚠️ Nie udało się rozszerzyć '{section_title}': {e}")
|
| 207 |
+
expanded_sections[section_title] = section_content
|
| 208 |
+
else:
|
| 209 |
+
# Sekcja już wystarczająco długa
|
| 210 |
+
expanded_sections[section_title] = section_content
|
| 211 |
+
st.success(f"✅ {section_title} OK ({word_count} słów)")
|
| 212 |
+
|
| 213 |
+
return expanded_sections
|
| 214 |
+
|
| 215 |
+
def _assemble_final_report(self, sections: Dict, brief: str, interview_type: str, interviews_count: int) -> str:
|
| 216 |
+
"""Scal wszystko w finalny raport"""
|
| 217 |
+
try:
|
| 218 |
+
sections_text = "\n\n".join([
|
| 219 |
+
f"## {title}\n\n{content}"
|
| 220 |
+
for title, content in sections.items()
|
| 221 |
+
])
|
| 222 |
+
|
| 223 |
+
prompt = REPORT_PROMPTS['final_assembly'].format(
|
| 224 |
+
sections=sections_text,
|
| 225 |
+
brief=brief or "Brak szczegółowego briefu",
|
| 226 |
+
interview_type=INTERVIEW_TYPES.get(interview_type, 'wywiad'),
|
| 227 |
+
interviews_count=interviews_count,
|
| 228 |
+
date=datetime.now().strftime("%Y-%m-%d")
|
| 229 |
+
)
|
| 230 |
+
|
| 231 |
+
final_report = self._call_gpt(prompt, max_tokens=4000)
|
| 232 |
+
|
| 233 |
+
# Dodaj metadane na koniec
|
| 234 |
+
metadata = f"""
|
| 235 |
+
|
| 236 |
+
---
|
| 237 |
+
|
| 238 |
+
## METADATA RAPORTU
|
| 239 |
+
- **Wygenerowano**: {datetime.now().strftime("%Y-%m-%d %H:%M")}
|
| 240 |
+
- **Typ badania**: {INTERVIEW_TYPES.get(interview_type, 'nieznany')}
|
| 241 |
+
- **Liczba wywiadów**: {interviews_count}
|
| 242 |
+
- **Sekcji wygenerowanych**: {self.generation_stats['sections_generated']}
|
| 243 |
+
- **Sekcji rozszerzonych**: {self.generation_stats['sections_expanded']}
|
| 244 |
+
- **Czas generowania**: {self.generation_stats['generation_time']:.1f}s
|
| 245 |
+
- **Generator**: FGI/IDI Research Analyzer v1.0
|
| 246 |
+
"""
|
| 247 |
+
|
| 248 |
+
return final_report + metadata
|
| 249 |
+
|
| 250 |
+
except Exception as e:
|
| 251 |
+
st.error(f"❌ Błąd finalnego scalenia: {e}")
|
| 252 |
+
# Fallback - zwróć przynajmniej sekcje
|
| 253 |
+
return self._create_fallback_report(sections, brief, interview_type)
|
| 254 |
+
|
| 255 |
+
def _call_gpt(self, prompt: str, max_tokens: int = 3000) -> str:
|
| 256 |
+
"""Wywołanie GPT API z error handling"""
|
| 257 |
+
try:
|
| 258 |
+
response = self.client.chat.completions.create(
|
| 259 |
+
model=MODEL_SETTINGS['gpt']['model'],
|
| 260 |
+
messages=[
|
| 261 |
+
{"role": "system", "content": "Jesteś ekspertem analizy badań jakościowych. Tworzysz profesjonalne, szczegółowe raporty badawcze."},
|
| 262 |
+
{"role": "user", "content": prompt}
|
| 263 |
+
],
|
| 264 |
+
temperature=MODEL_SETTINGS['gpt']['temperature'],
|
| 265 |
+
max_tokens=max_tokens
|
| 266 |
+
)
|
| 267 |
+
|
| 268 |
+
# Statystyki
|
| 269 |
+
if hasattr(response, 'usage'):
|
| 270 |
+
self.generation_stats['total_tokens_used'] += response.usage.total_tokens
|
| 271 |
+
# Estymacja kosztu GPT-4o-mini: ~$0.00015 per 1K tokens
|
| 272 |
+
self.generation_stats['total_cost_estimate'] += (response.usage.total_tokens / 1000) * 0.00015
|
| 273 |
+
|
| 274 |
+
return response.choices[0].message.content
|
| 275 |
+
|
| 276 |
+
except Exception as e:
|
| 277 |
+
if "rate limit" in str(e).lower():
|
| 278 |
+
st.warning("⏳ Rate limit - czekam 60s...")
|
| 279 |
+
time.sleep(60)
|
| 280 |
+
return self._call_gpt(prompt, max_tokens)
|
| 281 |
+
else:
|
| 282 |
+
raise e
|
| 283 |
+
|
| 284 |
+
def _parse_outline(self, outline_text: str) -> Dict:
|
| 285 |
+
"""Parsuj outline z odpowiedzi GPT"""
|
| 286 |
+
outline = {}
|
| 287 |
+
current_section = None
|
| 288 |
+
|
| 289 |
+
for line in outline_text.split('\n'):
|
| 290 |
+
line = line.strip()
|
| 291 |
+
|
| 292 |
+
if line.startswith('## '):
|
| 293 |
+
# Nowa sekcja
|
| 294 |
+
current_section = line[3:].strip()
|
| 295 |
+
outline[current_section] = []
|
| 296 |
+
elif line.startswith('- ') and current_section:
|
| 297 |
+
# Podpunkt sekcji
|
| 298 |
+
outline[current_section].append(line[2:].strip())
|
| 299 |
+
|
| 300 |
+
return outline
|
| 301 |
+
|
| 302 |
+
def _create_fallback_report(self, sections: Dict, brief: str, interview_type: str) -> str:
|
| 303 |
+
"""Fallback raport jeśli final assembly nie zadziała"""
|
| 304 |
+
report_parts = [
|
| 305 |
+
f"# RAPORT Z BADANIA {INTERVIEW_TYPES.get(interview_type, 'INTERVIEW').upper()}",
|
| 306 |
+
f"\n**Data**: {datetime.now().strftime('%Y-%m-%d')}",
|
| 307 |
+
f"**Brief**: {brief or 'Brak szczegółowego briefu'}",
|
| 308 |
+
"\n---\n"
|
| 309 |
+
]
|
| 310 |
+
|
| 311 |
+
for title, content in sections.items():
|
| 312 |
+
report_parts.append(f"## {title}\n\n{content}\n\n")
|
| 313 |
+
|
| 314 |
+
return "\n".join(report_parts)
|
| 315 |
+
|
| 316 |
+
def _log_generation_stats(self):
|
| 317 |
+
"""Wyświetl statystyki generowania"""
|
| 318 |
+
stats = self.generation_stats
|
| 319 |
+
|
| 320 |
+
st.info(f"""
|
| 321 |
+
📊 **Statystyki generowania:**
|
| 322 |
+
- Sekcji: {stats['sections_generated']} wygenerowanych, {stats['sections_expanded']} rozszerzonych
|
| 323 |
+
- Tokeny: ~{stats['total_tokens_used']:,}
|
| 324 |
+
- Koszt: ~${stats['total_cost_estimate']:.4f}
|
| 325 |
+
- Czas: {stats['generation_time']:.1f}s
|
| 326 |
+
""")
|
| 327 |
+
|
| 328 |
+
def evaluate_section_quality(self, section_content: str) -> Dict:
|
| 329 |
+
"""Oceń jakość sekcji (dla debugowania)"""
|
| 330 |
+
try:
|
| 331 |
+
prompt = REPORT_PROMPTS['quality_checker'].format(section=section_content)
|
| 332 |
+
evaluation = self._call_gpt(prompt, max_tokens=500)
|
| 333 |
+
|
| 334 |
+
# Parsuj ocenę (uproszczone)
|
| 335 |
+
lines = evaluation.split('\n')
|
| 336 |
+
scores = {}
|
| 337 |
+
|
| 338 |
+
for line in lines:
|
| 339 |
+
if ':' in line and '/10' in line:
|
| 340 |
+
criterion = line.split(':')[0].strip()
|
| 341 |
+
score = line.split(':')[1].strip().split('/')[0]
|
| 342 |
+
try:
|
| 343 |
+
scores[criterion] = int(score)
|
| 344 |
+
except:
|
| 345 |
+
pass
|
| 346 |
+
|
| 347 |
+
needs_improvement = 'TAK' in evaluation.upper()
|
| 348 |
+
|
| 349 |
+
return {
|
| 350 |
+
'scores': scores,
|
| 351 |
+
'needs_improvement': needs_improvement,
|
| 352 |
+
'evaluation_text': evaluation
|
| 353 |
+
}
|
| 354 |
+
|
| 355 |
+
except Exception as e:
|
| 356 |
+
return {'error': str(e)}
|
| 357 |
+
|
| 358 |
+
def get_generation_stats(self) -> Dict:
|
| 359 |
+
"""Zwróć statystyki generowania"""
|
| 360 |
+
return self.generation_stats.copy()
|
| 361 |
+
|
| 362 |
+
# Funkcje pomocnicze
|
| 363 |
+
def estimate_report_length(transcriptions: Dict[str, str]) -> Dict:
|
| 364 |
+
"""Estymuj długość finalnego raportu"""
|
| 365 |
+
total_words = sum(len(text.split()) for text in transcriptions.values())
|
| 366 |
+
|
| 367 |
+
# Raporty są zwykle 15-25% długości transkrypcji
|
| 368 |
+
estimated_report_words = int(total_words * 0.2)
|
| 369 |
+
estimated_pages = estimated_report_words / 250 # ~250 słów na stronę
|
| 370 |
+
|
| 371 |
+
return {
|
| 372 |
+
'transcription_words': total_words,
|
| 373 |
+
'estimated_report_words': estimated_report_words,
|
| 374 |
+
'estimated_pages': estimated_pages,
|
| 375 |
+
'estimated_generation_time': len(transcriptions) * 120 # ~2 min per interview
|
| 376 |
+
}
|
| 377 |
+
|
| 378 |
+
# Test modułu
|
| 379 |
+
if __name__ == "__main__":
|
| 380 |
+
print("🧪 Test ReportGenerator")
|
| 381 |
+
|
| 382 |
+
# Test bez prawdziwego API
|
| 383 |
+
try:
|
| 384 |
+
generator = ReportGenerator("test-key")
|
| 385 |
+
print("✅ ReportGenerator zainicjalizowany")
|
| 386 |
+
|
| 387 |
+
# Test estymacji
|
| 388 |
+
test_transcriptions = {
|
| 389 |
+
"test1.mp3": "To jest przykładowa transkrypcja wywiadu. " * 100,
|
| 390 |
+
"test2.mp3": "To jest druga transkrypcja z badania. " * 150
|
| 391 |
+
}
|
| 392 |
+
|
| 393 |
+
estimates = estimate_report_length(test_transcriptions)
|
| 394 |
+
print(f"📊 Estymacja: {estimates['estimated_report_words']} słów, {estimates['estimated_pages']:.1f} stron")
|
| 395 |
+
|
| 396 |
+
except Exception as e:
|
| 397 |
+
print(f"❌ Błąd testu: {e}")
|
| 398 |
+
|
| 399 |
+
print("✅ Test zakończony")
|
transcription.py
ADDED
|
@@ -0,0 +1,287 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# transcription.py - Moduł transkrypcji audio używając OpenAI Whisper
|
| 2 |
+
|
| 3 |
+
import os
|
| 4 |
+
import time
|
| 5 |
+
import streamlit as st
|
| 6 |
+
from typing import List, Dict, Optional
|
| 7 |
+
from pathlib import Path
|
| 8 |
+
|
| 9 |
+
try:
|
| 10 |
+
from openai import OpenAI
|
| 11 |
+
OPENAI_AVAILABLE = True
|
| 12 |
+
except ImportError:
|
| 13 |
+
OPENAI_AVAILABLE = False
|
| 14 |
+
st.error("❌ OpenAI library nie jest dostępna. Zainstaluj: pip install openai")
|
| 15 |
+
|
| 16 |
+
from config import MODEL_SETTINGS, USER_MESSAGES
|
| 17 |
+
|
| 18 |
+
class AudioTranscriber:
|
| 19 |
+
"""Klasa do transkrypcji audio używając OpenAI Whisper API"""
|
| 20 |
+
|
| 21 |
+
def __init__(self, api_key: str):
|
| 22 |
+
if not OPENAI_AVAILABLE:
|
| 23 |
+
raise Exception("OpenAI library nie jest dostępna")
|
| 24 |
+
|
| 25 |
+
self.client = OpenAI(api_key=api_key)
|
| 26 |
+
self.api_key = api_key
|
| 27 |
+
self.transcription_stats = {
|
| 28 |
+
'total_files': 0,
|
| 29 |
+
'successful': 0,
|
| 30 |
+
'failed': 0,
|
| 31 |
+
'total_duration': 0,
|
| 32 |
+
'total_cost_estimate': 0
|
| 33 |
+
}
|
| 34 |
+
|
| 35 |
+
def transcribe_files(self, file_paths: List[str], language: str = "pl") -> str:
|
| 36 |
+
"""
|
| 37 |
+
Transkrypcja listy plików audio
|
| 38 |
+
Returns: Połączona transkrypcja wszystkich plików
|
| 39 |
+
"""
|
| 40 |
+
transcriptions = []
|
| 41 |
+
|
| 42 |
+
for i, file_path in enumerate(file_paths):
|
| 43 |
+
if not os.path.exists(file_path):
|
| 44 |
+
st.error(f"❌ Plik nie istnieje: {file_path}")
|
| 45 |
+
continue
|
| 46 |
+
|
| 47 |
+
try:
|
| 48 |
+
# Pokaż postęp
|
| 49 |
+
if len(file_paths) > 1:
|
| 50 |
+
st.info(f"🎙️ Transkrybuję część {i+1}/{len(file_paths)}")
|
| 51 |
+
|
| 52 |
+
# Transkrypcja pojedynczego pliku
|
| 53 |
+
transcription = self._transcribe_single_file(file_path, language)
|
| 54 |
+
|
| 55 |
+
if transcription:
|
| 56 |
+
transcriptions.append(transcription)
|
| 57 |
+
self.transcription_stats['successful'] += 1
|
| 58 |
+
st.success(f"✅ Część {i+1} zakończona")
|
| 59 |
+
else:
|
| 60 |
+
self.transcription_stats['failed'] += 1
|
| 61 |
+
st.error(f"❌ Błąd części {i+1}")
|
| 62 |
+
|
| 63 |
+
except Exception as e:
|
| 64 |
+
st.error(f"❌ Błąd transkrypcji części {i+1}: {str(e)}")
|
| 65 |
+
self.transcription_stats['failed'] += 1
|
| 66 |
+
|
| 67 |
+
# Połącz wszystkie transkrypcje
|
| 68 |
+
if transcriptions:
|
| 69 |
+
# Jeśli było więcej niż jeden plik, dodaj separatory
|
| 70 |
+
if len(transcriptions) > 1:
|
| 71 |
+
final_transcription = "\n\n=== CZĘŚĆ 1 ===\n\n".join([
|
| 72 |
+
transcriptions[0]
|
| 73 |
+
] + [
|
| 74 |
+
f"=== CZĘŚĆ {i+1} ===\n\n{text}"
|
| 75 |
+
for i, text in enumerate(transcriptions[1:], 1)
|
| 76 |
+
])
|
| 77 |
+
else:
|
| 78 |
+
final_transcription = transcriptions[0]
|
| 79 |
+
|
| 80 |
+
return final_transcription
|
| 81 |
+
else:
|
| 82 |
+
raise Exception("Wszystkie transkrypcje zakończone błędem")
|
| 83 |
+
|
| 84 |
+
def _transcribe_single_file(self, file_path: str, language: str = "pl") -> Optional[str]:
|
| 85 |
+
"""Transkrypcja pojedynczego pliku"""
|
| 86 |
+
try:
|
| 87 |
+
self.transcription_stats['total_files'] += 1
|
| 88 |
+
|
| 89 |
+
# Sprawdź rozmiar pliku
|
| 90 |
+
file_size = os.path.getsize(file_path)
|
| 91 |
+
file_size_mb = file_size / (1024 * 1024)
|
| 92 |
+
|
| 93 |
+
# OpenAI Whisper ma limit 25MB
|
| 94 |
+
if file_size_mb > 25:
|
| 95 |
+
raise Exception(f"Plik za duży dla Whisper API: {file_size_mb:.1f}MB > 25MB")
|
| 96 |
+
|
| 97 |
+
st.info(f"📤 Wysyłam do Whisper ({file_size_mb:.1f}MB)...")
|
| 98 |
+
|
| 99 |
+
# Otwórz plik i wyślij do API
|
| 100 |
+
with open(file_path, 'rb') as audio_file:
|
| 101 |
+
transcript = self.client.audio.transcriptions.create(
|
| 102 |
+
model=MODEL_SETTINGS['whisper']['model'],
|
| 103 |
+
file=audio_file,
|
| 104 |
+
language=language if language != 'auto' else None,
|
| 105 |
+
temperature=MODEL_SETTINGS['whisper']['temperature']
|
| 106 |
+
)
|
| 107 |
+
|
| 108 |
+
# Estymacja kosztu (Whisper API: $0.006 per minute)
|
| 109 |
+
estimated_duration = file_size_mb * 60 # Rough estimate: 1MB ≈ 1 minute
|
| 110 |
+
estimated_cost = (estimated_duration / 60) * 0.006
|
| 111 |
+
self.transcription_stats['total_duration'] += estimated_duration
|
| 112 |
+
self.transcription_stats['total_cost_estimate'] += estimated_cost
|
| 113 |
+
|
| 114 |
+
st.success(f"✅ Transkrypcja otrzymana (~{estimated_duration:.1f}s audio)")
|
| 115 |
+
|
| 116 |
+
return transcript.text
|
| 117 |
+
|
| 118 |
+
except Exception as e:
|
| 119 |
+
st.error(f"❌ Błąd Whisper API: {str(e)}")
|
| 120 |
+
|
| 121 |
+
# Jeśli błąd rate limit, poczekaj i spróbuj ponownie
|
| 122 |
+
if "rate limit" in str(e).lower():
|
| 123 |
+
st.warning("⏳ Rate limit - czekam 60s i próbuję ponownie...")
|
| 124 |
+
time.sleep(60)
|
| 125 |
+
return self._transcribe_single_file(file_path, language)
|
| 126 |
+
|
| 127 |
+
return None
|
| 128 |
+
|
| 129 |
+
def transcribe_with_retries(self, file_path: str, language: str = "pl", max_retries: int = 3) -> Optional[str]:
|
| 130 |
+
"""Transkrypcja z ponawianiem przy błędach"""
|
| 131 |
+
for attempt in range(max_retries):
|
| 132 |
+
try:
|
| 133 |
+
result = self._transcribe_single_file(file_path, language)
|
| 134 |
+
if result:
|
| 135 |
+
return result
|
| 136 |
+
|
| 137 |
+
except Exception as e:
|
| 138 |
+
st.warning(f"⚠️ Próba {attempt + 1}/{max_retries} nieudana: {str(e)}")
|
| 139 |
+
|
| 140 |
+
if attempt < max_retries - 1:
|
| 141 |
+
wait_time = (attempt + 1) * 30 # Exponential backoff
|
| 142 |
+
st.info(f"⏳ Czekam {wait_time}s przed następną próbą...")
|
| 143 |
+
time.sleep(wait_time)
|
| 144 |
+
else:
|
| 145 |
+
st.error(f"❌ Wszystkie {max_retries} prób nieudane")
|
| 146 |
+
|
| 147 |
+
return None
|
| 148 |
+
|
| 149 |
+
def estimate_transcription_time(self, file_paths: List[str]) -> Dict:
|
| 150 |
+
"""Estymuj czas i koszt transkrypcji"""
|
| 151 |
+
total_size = sum(os.path.getsize(path) for path in file_paths if os.path.exists(path))
|
| 152 |
+
total_size_mb = total_size / (1024 * 1024)
|
| 153 |
+
|
| 154 |
+
# Estymacje
|
| 155 |
+
estimated_duration_minutes = total_size_mb # 1MB ≈ 1 minute
|
| 156 |
+
estimated_api_time = estimated_duration_minutes * 0.1 # Whisper jest ~10x szybszy niż realtime
|
| 157 |
+
estimated_cost = estimated_duration_minutes * 0.006 # $0.006 per minute
|
| 158 |
+
|
| 159 |
+
return {
|
| 160 |
+
'total_size_mb': total_size_mb,
|
| 161 |
+
'estimated_audio_duration': estimated_duration_minutes,
|
| 162 |
+
'estimated_processing_time': estimated_api_time,
|
| 163 |
+
'estimated_cost_usd': estimated_cost,
|
| 164 |
+
'files_count': len(file_paths)
|
| 165 |
+
}
|
| 166 |
+
|
| 167 |
+
def validate_api_key(self) -> bool:
|
| 168 |
+
"""Sprawdź czy klucz API działa"""
|
| 169 |
+
try:
|
| 170 |
+
# Spróbuj pobrać listę modeli
|
| 171 |
+
models = self.client.models.list()
|
| 172 |
+
return True
|
| 173 |
+
except Exception as e:
|
| 174 |
+
st.error(f"❌ Nieprawidłowy klucz API: {str(e)}")
|
| 175 |
+
return False
|
| 176 |
+
|
| 177 |
+
def get_transcription_stats(self) -> Dict:
|
| 178 |
+
"""Zwróć statystyki transkrypcji"""
|
| 179 |
+
return self.transcription_stats.copy()
|
| 180 |
+
|
| 181 |
+
def detect_interview_type(self, transcription: str) -> str:
|
| 182 |
+
"""
|
| 183 |
+
Automatyczne rozpoznanie typu wywiadu na podstawie treści
|
| 184 |
+
Returns: 'fgi', 'idi', lub 'unknown'
|
| 185 |
+
"""
|
| 186 |
+
text_lower = transcription.lower()
|
| 187 |
+
|
| 188 |
+
# Wskaźniki FGI (Focus Group)
|
| 189 |
+
fgi_indicators = [
|
| 190 |
+
'moderator', 'grupa', 'wszyscy', 'kto jeszcze', 'a państwo',
|
| 191 |
+
'czy zgadzacie się', 'co myślicie', 'focus group',
|
| 192 |
+
'uczestnicy', 'grupa fokusowa', 'dyskusja grupowa'
|
| 193 |
+
]
|
| 194 |
+
|
| 195 |
+
# Wskaźniki IDI (Individual)
|
| 196 |
+
idi_indicators = [
|
| 197 |
+
'wywiad indywidualny', 'jeden na jeden', 'prywatnie',
|
| 198 |
+
'osobiście', 'indywidualne', 'w cztery oczy'
|
| 199 |
+
]
|
| 200 |
+
|
| 201 |
+
fgi_score = sum(1 for indicator in fgi_indicators if indicator in text_lower)
|
| 202 |
+
idi_score = sum(1 for indicator in idi_indicators if indicator in text_lower)
|
| 203 |
+
|
| 204 |
+
# Sprawdź także liczbę różnych głosów/osób
|
| 205 |
+
# (FGI zwykle ma więcej przerywników, overlapping speech)
|
| 206 |
+
interruption_patterns = ['...', '[', ']', '(', ')', '--']
|
| 207 |
+
interruption_count = sum(text_lower.count(pattern) for pattern in interruption_patterns)
|
| 208 |
+
|
| 209 |
+
if fgi_score > idi_score and interruption_count > 10:
|
| 210 |
+
return 'fgi'
|
| 211 |
+
elif idi_score > fgi_score:
|
| 212 |
+
return 'idi'
|
| 213 |
+
elif interruption_count > 20: # Dużo przerywników = prawdopodobnie grupa
|
| 214 |
+
return 'fgi'
|
| 215 |
+
else:
|
| 216 |
+
return 'unknown'
|
| 217 |
+
|
| 218 |
+
def clean_transcription(self, transcription: str) -> str:
|
| 219 |
+
"""Oczyszczenie i formatowanie transkrypcji"""
|
| 220 |
+
try:
|
| 221 |
+
# Usuń nadmiarowe spacje
|
| 222 |
+
lines = transcription.split('\n')
|
| 223 |
+
cleaned_lines = []
|
| 224 |
+
|
| 225 |
+
for line in lines:
|
| 226 |
+
line = line.strip()
|
| 227 |
+
if line: # Pomijaj puste linie
|
| 228 |
+
# Usuń nadmiarowe spacje
|
| 229 |
+
line = ' '.join(line.split())
|
| 230 |
+
cleaned_lines.append(line)
|
| 231 |
+
|
| 232 |
+
# Połącz z pojedynczymi przerwami linii
|
| 233 |
+
cleaned = '\n\n'.join(cleaned_lines)
|
| 234 |
+
|
| 235 |
+
# Dodaj informacje metadata na początek
|
| 236 |
+
metadata = f"""TRANSKRYPCJA AUDIO
|
| 237 |
+
Data: {time.strftime('%Y-%m-%d %H:%M')}
|
| 238 |
+
Typ: {self.detect_interview_type(cleaned).upper()}
|
| 239 |
+
Długość: ~{len(cleaned.split())} słów
|
| 240 |
+
|
| 241 |
+
---
|
| 242 |
+
|
| 243 |
+
"""
|
| 244 |
+
|
| 245 |
+
return metadata + cleaned
|
| 246 |
+
|
| 247 |
+
except Exception as e:
|
| 248 |
+
st.warning(f"⚠️ Błąd czyszczenia transkrypcji: {e}")
|
| 249 |
+
return transcription
|
| 250 |
+
|
| 251 |
+
# Funkcje pomocnicze dla kompatybilności
|
| 252 |
+
def validate_audio_file(file_path: str) -> bool:
|
| 253 |
+
"""Sprawdź czy plik audio jest prawidłowy"""
|
| 254 |
+
if not os.path.exists(file_path):
|
| 255 |
+
return False
|
| 256 |
+
|
| 257 |
+
# Sprawdź rozmiar
|
| 258 |
+
file_size = os.path.getsize(file_path)
|
| 259 |
+
if file_size == 0:
|
| 260 |
+
return False
|
| 261 |
+
|
| 262 |
+
# Sprawdź rozszerzenie
|
| 263 |
+
valid_extensions = ['.mp3', '.wav', '.mp4', '.m4a', '.aac']
|
| 264 |
+
file_ext = Path(file_path).suffix.lower()
|
| 265 |
+
|
| 266 |
+
return file_ext in valid_extensions
|
| 267 |
+
|
| 268 |
+
# Test modułu
|
| 269 |
+
if __name__ == "__main__":
|
| 270 |
+
print("🧪 Test AudioTranscriber")
|
| 271 |
+
|
| 272 |
+
# Test bez prawdziwego API key
|
| 273 |
+
try:
|
| 274 |
+
transcriber = AudioTranscriber("test-key")
|
| 275 |
+
print("✅ AudioTranscriber zainicjalizowany")
|
| 276 |
+
|
| 277 |
+
# Test rozpoznania typu wywiadu
|
| 278 |
+
test_fgi = "Moderator: Co wszyscy myślicie o produkcie? Czy zgadzacie się z tym?"
|
| 279 |
+
test_idi = "Interviewer: A teraz opowiedz mi o swoich doświadczeniach..."
|
| 280 |
+
|
| 281 |
+
print(f"Test FGI: {transcriber.detect_interview_type(test_fgi)}")
|
| 282 |
+
print(f"Test IDI: {transcriber.detect_interview_type(test_idi)}")
|
| 283 |
+
|
| 284 |
+
except Exception as e:
|
| 285 |
+
print(f"❌ Błąd testu: {e}")
|
| 286 |
+
|
| 287 |
+
print("✅ Test zakończony")
|