AI Kernel Explorer

Provides a text-based file browser using the Textual framework's DirectoryTree widget to traverse Linux kernel source code hierarchies. Users navigate the file system structure interactively, with the UI rendering directory trees and file listings in real-time. The implementation leverages Textual's reactive event system to handle directory expansion/collapse and file selection without blocking I/O.

Generates intelligent, human-readable summaries of Linux kernel source files by sending file contents to OpenAI's GPT-4o API. The implementation reads selected kernel files, constructs a prompt with the source code, and streams responses back to the TUI. The system handles multi-line code context and generates explanations of kernel subsystem functionality, data structures, and algorithms without requiring local code parsing.

Implements a local cache mechanism that stores AI-generated summaries in ~/.cache/ai-kernel-explorer using file-based storage keyed by source file path. When a user requests a summary for a file that has been previously summarized, the cached response is retrieved and displayed instantly without making a new API call. The cache is transparent to the user and automatically reduces API costs and latency on repeated exploration of the same kernel files.

Allows users to specify which OpenAI model to use for code summarization through the --model command-line flag, defaulting to gpt-4o but supporting alternative models like gpt-4-turbo, gpt-4, or gpt-3.5-turbo. The model selection is passed directly to the OpenAI API client and affects both the quality of summaries and the token cost per request. This design enables users to trade off between summary quality, latency, and API costs based on their specific needs.

Accepts a [root] positional CLI argument allowing users to specify any accessible directory as the starting point for kernel source exploration, defaulting to /usr/src if not provided. This design enables exploration of kernel source from custom locations (e.g., ~/linux-kernel, /opt/kernel-src) without requiring the tool to be reconfigured or reinstalled. The path is validated at startup and used as the root for the DirectoryTree widget.

Streams OpenAI API responses token-by-token into the Textual TUI, rendering summaries in real-time as they are generated rather than waiting for the complete response. The implementation uses OpenAI's streaming API and integrates with Textual's reactive update system to display partial responses without blocking the UI. This approach provides immediate visual feedback to users and makes long summaries feel more responsive.

Reads kernel source files from the filesystem, validates that they are readable and contain text content, and prepares them for AI summarization. The implementation handles file I/O errors gracefully, supports multiple file types (C source, headers, assembly, makefiles), and enforces reasonable file size limits to prevent excessive API token usage. File content is read synchronously but integrated into the async TUI event loop to prevent blocking.

Initializes the OpenAI Python client using the OPENAI_API_KEY environment variable, configuring it with the user-selected model and handling authentication errors at startup. The implementation validates that a valid API key is present before attempting any API calls and provides clear error messages if authentication fails. The client is created once at application startup and reused for all subsequent API requests.