Настройка Codex AI агента: от установки на Ubuntu до команд в Telegram
The transcript covers the setup of a Codex AI agent with voice message support in Telegram, specifically addressing the limitation of standard ChatGPT subscriptions with Whisper. The speaker walks through obtaining an OpenAI API key to enable Whisper transcription and debugging a file extension error with OGA voice files. By the end, the bot successfully receives and transcribes Telegram voice messages.
Summary
The video begins by highlighting a key limitation: a standard ChatGPT subscription does not support tools like Whisper, which is required for transcribing voice messages sent to a Telegram bot. The speaker explains that without Whisper, the bot simply cannot process voice input, and an additional tool must be configured to enable this functionality.
To resolve this, the speaker walks through the process of obtaining an OpenAI API key. This involves logging into the OpenAI platform at platform.openai.com, navigating to the settings and API keys section, creating a new secret API key tied to a specific project, and then sending that key to the agent so it can authenticate and use Whisper when voice messages are received.
After providing the API key, the speaker tests the bot by sending a voice message. An error occurs, and the speaker investigates by sending the full error text to the agent and asking it to fix the problem. The agent diagnoses the issue: Telegram voice messages arrive with an .OGA file extension, but the OpenAI API rejects this suffix. Additionally, FFmpeg is not installed in the environment. The agent proposes and implements a minimal fix — saving the files with an .OGG extension without altering the binary container — effectively resolving the compatibility issue without requiring FFmpeg.
After the fix is applied, the speaker conducts a second test by sending another voice message asking the bot to confirm it can receive voice input by outputting a specific number sequence. The bot successfully transcribes the voice message and responds with the correct sequence, confirming that voice message support is now fully functional. The speaker briefly mentions that the same approach applies to video messages as well.
Key Insights
- The speaker explains that a standard ChatGPT subscription does not grant access to Whisper, meaning the bot cannot transcribe voice messages without an explicit OpenAI API key being provided to the agent.
- The speaker demonstrates that the OpenAI API key must be generated at platform.openai.com under Settings > API Keys, tied to a specific project, and then manually sent to the agent to unlock Whisper functionality.
- The agent self-diagnoses the voice message error, identifying that Telegram sends voice files with an .OGA extension which the OpenAI API rejects, and that FFmpeg is not installed in the environment.
- The agent proposes and applies a minimal fix — renaming the file extension from .OGA to .OGG without altering the binary container — rather than requiring FFmpeg installation, resolving the transcription error.
- After the fix, the speaker confirms successful voice message transcription by having the bot respond with a specific number sequence dictated in a voice message, validating that Whisper integration is working end-to-end.
Topics
Transcript
[0:00] Unfortunately, with a standard Chat GPT subscription we cannot use tools like Whisper. Whisper allows you to transcribe messages that come as a voice message. That is, if we now record a voice message for our bot, it will most likely not accept it. He needs an extra tool. The first tool is specifically for receiving voice messages. But Wiper, unfortunately, does not work with the standard CHGPT subscription. Therefore, in order for Whisper to work, we need to send [0:30] him the Open Apk key so that he can use Wisper when he receives voice messages. To do this, we need to go to the Open AI personal account. The account is located at platformopai.com. Log in to…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Игорь Зуевич
Сравнение Claude Code, Gemini и Codex — тесты на контент, анализ и изображения
The speaker is comparing AI coding assistants — Claude Code, Gemini, and Codex — by testing their ability to create image-generation agents. Each tool is being evaluated on how well it writes prompts and structures sub-agents for text-to-image and image-to-image tasks.
Экономия токенов и контекстного окна: настройка памяти нейросети – Compact сжатие контекстного окна
The transcript explains the 'Compact' command in AI agent sessions, which compresses the entire conversation history to reduce context window usage without switching sessions. It also covers related commands for full history clearing and automatic triggers for compact execution. The speaker recommends configuring automatic notifications or auto-execution when the context window reaches a certain threshold.
HR-агент на Claude Code: создаёт других сотрудников сам по запросу
The video demonstrates setting up a Telegram HR bot using Claude Code, walking through the process of creating a bot via BotFather, obtaining a Telegram user ID, and letting the code automatically generate the HR agent. The presenter shows how the system initializes itself by installing dependencies and launching the bot with minimal manual configuration.
Живые скиллы для Claude Code, Gemini и Codex | Как настроить super skills для ИИ агентов
The video explains how to create 'living' or 'super' skills for AI agents like Claude Code, Gemini, and Codex that self-update based on user feedback. The presenter demonstrates how these skills collect feedback after each task, store it in log files, and periodically incorporate that feedback into the skill's core instructions. This approach reduces token consumption and ensures skills continuously improve over time.
Планировщик для ИИ агента: 50 идей и готовые файлы для Claude, Codex и Gemini
The speaker describes their use of a scheduling/routing tool called Routins, which they compare to N8N, to automate daily delivery of curated YouTube videos on topics like Gemini and Claude. The system runs autonomously, collecting videos and generating concise summaries of their content without manual intervention.