nynchezyabka/rozittaparser

1. Product and entry points

Rozitta Parser is a local Python desktop application for authenticating to Telegram, downloading chat history/media into portable local archives, and exporting readable or RAG-ready documents. Runtime starts at main.py, which composes the application shell in ui/main_window.py; global settings and output locations are centralized in config.py.

2. Desktop UI and feature orchestration

The main window coordinates feature-specific screens rather than exposing a web/API service. Each major user journey has a UI module paired with an API/service-facing module:

• Authentication: features/auth/ui.py, features/auth/api.py

• Chat selection: features/chats/ui.py, features/chats/api.py

• Archive parsing: features/parser/ui.py, features/parser/api.py

Shared desktop widgets, styles, and calendar controls reside in core/ui_shared/widgets.py, core/ui_shared/styles.py, and core/ui_shared/calendar.py.

3. Telegram access and archive acquisition

Telegram is the principal external service, accessed through the auth/chat/parser feature APIs. The flow is: establish a Telegram session (classic API credentials or Telegram Desktop tdata import), enumerate eligible dialogs, let the user select chats/options, then fetch messages, discussion threads, participants, and media into a local archive. Networking accommodates proxy usage through socks.py and sockshandler.py; retry/rate-limit behavior is isolated in core/retry.py. README identifies Telethon as the Telegram client runtime behavior source.

4. Local persistence and archive lifecycle

Downloaded content is persisted locally in SQLite plus an output directory containing media and generated documents. core/database.py is the storage boundary for archive records; check_db.py supports database inspection/maintenance. Archive updates and longer-running update work are separated into features/update_archive/worker.py and features/update_archive/report_dialog.py. The design keeps Telegram acquisition distinct from subsequent offline export and inspection.

5. Parsing, media, and speech-to-text pipeline

Parser orchestration turns Telegram entities into normalized local records and downloaded assets via features/parser/api.py. Audio/video-note transcription is an optional local post-processing stage: core/stt/worker.py coordinates background work, core/stt/audio_converter.py prepares supported audio, and core/stt/whisper_manager.py manages local Whisper inference. This boundary avoids a cloud transcription dependency; transcription text can then be included in exports.

6. Export and knowledge-corpus generation

The export subsystem reads local archive data and renders DOCX, Markdown, HTML, or JSON, with filtering and grouping options such as dates, posts, participants, and reply context. features/export/generator.py is the central rendering coordinator, while features/export/filters.py and features/export/participants.py shape the selected message set. features/export/knowledge_base.py implements AI-oriented corpus/chunk preparation; features/export/xml_magic.py is the specialized document/XML formatting boundary.

7. Operational boundaries

The application is intended to run locally on Python 3.10–3.13, with portable binary builds produced by .github/workflows/build_binaries.yml. It depends materially on Telegram, local SQLite storage, local filesystem media/output directories, optional SOCKS/MTProto proxy connectivity, and local Whisper models/runtime. Cross-cutting error, logging, merge, and utility behavior is concentrated in core/exceptions.py, core/logger.py, and core/merger.py.