kukiman010/tg_gpt

Purpose and execution surface

This is a Python Telegram AI-assistant bot. It accepts user text/media, routes requests to selectable LLM providers, supports speech and localization, and manages subscription/payment state. The likely application bootstrap is main.py, with setup centralized in configure.py and cross-component events in signals.py.

Telegram interaction and user state

The control layer represents bot-facing domain state: user identity, active environment/model settings, subscription data, payment records, and media context. These objects form the boundary between Telegram handlers and persistence/provider calls.

Paths: Control/user.py, Control/environment.py, Control/context_model.py, Control/user_media.py.

Conversation/model orchestration

main.py coordinates incoming commands/messages with user configuration and model invocation. data_models.py defines shared data structures, while signals.py likely decouples state changes or asynchronous workflow notifications. Model-specific modules are isolated behind provider adapters rather than embedded in bot handling.

Paths: main.py, data_models.py, signals.py.

Multi-provider AI gateway

The Gpt_models package is the external-model boundary. Separate adapters cover OpenAI/GPT, Claude, Google, YandexGPT, Sber GigaChat, DeepSeek, xAI, and MetaGPT. The runtime selects an adapter based on user/environment settings, sends prompts and context, and returns provider responses to the Telegram interaction layer.

Paths: Gpt_models/gpt_api.py, Gpt_models/claude_api.py, Gpt_models/yandexgpt.py, Gpt_models/sbergpt.py.

Persistence and database boundary

PostgreSQL is the intended persistent store; README requires UTF-8 setup and the repository includes a schema. Database access is separated into connection/core and API-facing layers. Persistent concerns include user profiles, environment/model choices, subscriptions, and payment information.

Paths: database.py, databaseapi.py, tools/tg_base.sql, Control/subscription_data.py.

Billing and entitlement flow

Payments are abstracted by a base provider contract and coordinated by a payment manager. Provider implementations include YooKassa and Crypto Pay. A successful payment should update payment/subscription domain state through the database boundary, enabling or extending access controls for AI services.

Paths: payments/base_pay_system.py, payments/payment_manager.py, payments/yookassa_api.py, payments/crypto_pay_api.py.

Supporting content services and deployment

Speech processing and file handling sit alongside the primary text flow, likely converting or processing user audio/uploads before response generation. Localization is gettext-style catalog based and translated through a dedicated module. The service is containerized with Docker Compose; deployment requires Python 3.8–3.10, PostgreSQL dependencies, and provider credentials/certificates for Yandex and Sber integrations.

Paths: speech.py, file_worker.py, translator.py, locale/LC_MESSAGES, Dockerfile, docker-compose.yml.