Console Chat Gpt

Deprecated / successor: For new installs, use tellm instead. It is a small Rust Telegram assistant for Claude, GPT/Grok, Gemini, compatible endpoints, and local Ollama, with per-room model selection and keychain-backed secrets.

---

Table of Contents

• Features
• Installation and Usage
• Examples
• Contributing

---

DISCLAIMER: The intention and implementation of this code are entirely unconnected and unrelated to OpenAI, MistralAI, Anthropic, xAI, Google AI, nor any other LLM lab or related parties. There is no affiliation or relationship with OpenAI, MistralAI, Anthropic, xAI, Google, any other LLM lab or their subsidiaries in any form.

---

Features

• :new: Support conversations via a Telegram bot, including Bot API 10.1 rich message formatting when available. How to create one: From BotFather to 'Hello World' :new:
• :star: OpenAI image generation via Responses API. :star:
• :star: OpenAI Responses API supported. :star:
• :star: Run any OpenAI SDK compatible - just add the model structure with the relevant model_name and base_url to the config.toml file. :star:
• :star: Run Ollama hosted models locally. Ollama should be installed and the selected models already downloaded :star:
• :star: Anthropic Prompt caching Fully supported :star:
• :star: Model Context Protocol (MCP) supported! If you are already using MCP servers just copy your claude_desktop_config.json to the root directory and rename to mcp_config.json to start using with any model! :star:
• Unified chat completion function separated as independent library to be used in any application for seamless cross-provider API experience. The source code is available in Python and TypeScript
• Streaming with all supported models, disabled by default, may be enabled in settings menu
• OpenAI Assistants Beta supported
• AI Managed mode Based on the complexity of the task, automatically determines which model to use.
• Configuration File: Easily customize the app's settings through the config.toml file for complete control over how the app works. Also supported in-app via the settings command.
• Role selection: Users can define the role of the AI in the conversation, allowing for a more personalized and interactive experience.
• Temperature Control: Adjust the temperature of generated responses to control creativity and randomness in the conversation.
• Command Handling: The app responds to various commands entered by the user for easy and intuitive interaction.
• Image input: with selected models.
• Error Handling: Clear and helpful error messages to easily understand and resolve any issues.
• Conversation History: Review previous interactions and save conversations for future reference, providing context and continuity.
• Graceful Exit: Smoothly handle interruptions, ensuring conversations are saved before exiting to avoid loss of progress.
• A nice team: Actively adding features, open for ideas and fixing bugs.

Overall, this app focuses on providing a user-friendly and customizable experience with features that enhance personalization, control, and convenience.

---

Installation and Usage

The script works fine on Linux and MacOS terminals. For Windows it's recommended to use WSL.

1. Clone the repository:

   git clone https://github.com/amidabuddha/console-chat-gpt.git
   

2. Go inside the folder:

   cd console-chat-gpt
   

3. Install the necessary dependencies:

   python3 -m pip install -r requirements.txt
   

4. Get your API key from OpenAI, MistralAI, Anthropic, xAI, Google AI Studio, or the relevant platform, depending on your selected LLM.

5. The config.toml.sample will be automatically copied into config.toml upon first run, with a prompt to enter your API key/s. Feel free to change any of the other defaults that are not available in the settings in-app menu as per your needs.

6. Run the executable:

   python3 main.py
   

   Pro-tip: Create an alias for the executable to run from anywhere.

7. Use the help command within the chat to check the available options.

8. Enjoy

---

Configurable Options in config.toml

[chat.defaults]	Main properties to generate a chat completion/response.
temperature	What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 more focused. May be set for each new chat session if adjust_temperature in [chat.features] is true.
system_role	A system (or developer) message inserted into the model's context. Should be one of the listed in [chat.roles] section. May be set for each new chat session if role_selector in [chat.features] is true.
model	Model ID used to generate the chat completion/response, like gpt-4o or o3. Should be listed in the [chat.models] section, with relevant parameters. May be set for each new chat session if model_selector in [chat.features] is true.

[chat.managed]	Settings dedicated to the AI Managed mode. Not available to be edited from within a chat session.
assistant	The preferred model that will evaluate your prompt and select the best available model out of the four configured below to handle it. Should be listed in the [chat.models] section, with relevant parameters.
assistant_role	Custom instruction to the evaluation model. Change this only if you know exactly what you are doing!
assistant_generalist	Your preferred general purpose model, typically the one you use the most for any type of queries. Should be listed in the [chat.models] section, with relevant parameters.
assistant_fast	When speed is preferred to accuracy. Should be listed in the [chat.models] section, with relevant parameters.
assistant_thinker	A reasoning model for complex tasks. Should be listed in the [chat.models] section, with relevant parameters.
assistant_coder	Your preferred model to handle Coding and Math questions. Should be listed in the [chat.models] section, with relevant parameters.
prompt	When AI Managed mode is used frequently the Y/N prompt may be disabled by changing this to false.

[chat.features]	Configurable options of the chat application. Some are accessible from within a chat session via the settings command.
model_selector	A selection list of models available in section [chat.models] of config.toml. When true this list may be modified at the beginning of each new chat session.
adjust_temperature	Prompt to change the temperature for each chat session. When true the temperature value may be modified at the beginning of each new chat session.
role_selector	A selection list of roles available in section [chat.roles] of config.toml. When true this list may be modified at the beginning of each new chat session.
save_chat_on_exit	When true automatically save the chat session upon using the exit command in chat.
continue_chat	When true offers a list of previously saved chat sessions to be continued in a new session. The list may be modified from within a chat session via the chats command.
debug	Application logging - not yet implemented.
disable_intro_help_message	All chat commands available in help are printed upon chat initialization. This is targeted at new users and may be disabled by setting to false.
assistant_mode	Enable Open AI Assistants API as an available selection upon chat initialization.
ai_managed	Enable AI Managed mode to allow a model to automatically select the best model according to your prompt. Detailed settings below.
streaming	If set to true, the model response data will be streamed to the terminal client. Telegram mode always sends completed replies.
mcp_client	Setting to false will prevent the default initialization of MCP servers for each chat if not needed.

[chat.telegram]	Enable Telegram UI for cnversations.
enabled	If set to true, starts Telegram bot polling loop instead of the terminal chat UI.
bot_token	Telegram bot token from BotFather, in the format <digits>:<token>.
allowed_chat_ids	List of chat IDs allowed to use the bot. Pairing mode activates when this list is empty and no model room mappings (telegram_chat_id/telegram_chat_ids) are configured: the bot prints a one-time code in its terminal, and the chat that sends /pair <code> is claimed and saved here; all other chats are denied. The code rotates every 10 minutes and repeated wrong attempts lock the offending chat out temporarily.
admin_chat_ids	List of chat IDs allowed to run admin-only commands such as /shutdown.
debug_context	If true, prints Telegram session/memory debug snapshots to terminal logs for troubleshooting context issues.
max_concurrent_updates	Maximum number of Telegram updates processed in parallel. Default is 8 when omitted. Different chats run concurrently, while each chat remains strictly ordered and isolated.

Telegram replies are sent through sendRichMessage when the Bot API supports it, so model Markdown can render as rich headings, lists, tables, quotes, formulas, details blocks, and similar structured output. If rich message delivery is unavailable or Telegram rejects a malformed rich block, the bot falls back to the classic sendMessage path.

Per-model Telegram room mapping is also supported inside each model block via:

• telegram_chat_id = <chat_id>
• telegram_chat_ids = [<chat_id_1>, <chat_id_2>]

When a chat room is mapped to a model, that room is pinned to this model (model switching commands are disabled there), while all other commands (/mode, /role, /reasoning, /websearch, /webfetch, etc.) remain available.

Adding your OpenAI SDK supported model

Add an entry at the end of your config.toml file. Use the following example structure:

[chat.models.{{model name that makes sense to you}}]
api_key = "YOUR_API_KEY"
base_url = {{the custom base URL, for example "https://api.deepseek.com/v1"}}
model_input_pricing_per_1k = 0.015 (currently the price is not calculated but you may adjust this for your reference)
model_max_tokens = 32000 (the suported context window of the selected model)
model_name = {{model name exactly as specified in the relevant API reference, for example "deepseek-reasoner"}}
model_output_pricing_per_1k = 0.075 (currently the price is not calculated but you may adjust this for your reference)
reasoning_effort = (the desired value if supported, such as "medium", otherwise false)

---

Examples

• Prompt example:

  

• Markdown visualization example:

  

• Settings and help:

  

You can find more examples on our Examples page.

---

Contributing

Contributions are welcome! If you find any bugs, have feature requests, or want to contribute improvements, please open an issue or submit a pull request.