aichat and aider: terminal workflows and configuration
Learn how the speaker uses aichat roles and documentation RAG, then configures aider for code comments, Git changes and focused test runs.
Player not loading? Watch on YouTube
This Russian-language overview, uploaded on July 31, 2025, describes the speaker's terminal workflows with aichat and aider. Part of the recording predates the later outdoor sections, but no recording dates or software versions are supplied. The configuration details reflect that demonstration.
The speaker uses aichat for questions, named sessions and reusable prompt roles. Examples include generating shell commands in Zsh, turning Git diffs into commit messages and correcting selected text in Vim. The documentation RAG example retrieves material to answer questions about configuration, with a command for showing source links. Sessions can follow the current directory, and the speaker describes model-generated summaries for long conversations.
Aider is the coding assistant used for repository work. The walkthrough covers file watching for requests written in code comments, asking questions before applying edits, Git-based undo and an architect mode that separates planning from implementation. The speaker recommends adding only relevant files to context and limiting test output. He warns that automatic lint fixes can get stuck in a loop.
Both workflows use an OpenAI-compatible API, with YandexGPT in the examples. The speaker criticizes its responses to large requests. This is an API-backed demonstration; it does not establish offline operation or show how to run models locally.