Kernel Chat is an advanced AI-powered CLI tool forked from Google's Gemini-CLI, specifically enhanced for embedded Linux developers and kernel engineers. It combines the power of Google's Gemini AI with specialized tools for embedded systems development, serial console debugging, intelligent code generation, and comprehensive documentation analysis.
See the kernel chat-CLI in action below!
- Direct hardware connection to embedded devices (BeagleBone Black, Raspberry Pi, custom SoCs)
- Real-time serial communication with automatic logging and buffering
- Interactive command execution with natural language interpretation
- Smart command translation: Ask "show me CPU usage" → automatically runs appropriate commands
- Live log summarization using AI for easy debugging
- Ingest technical documentation: PDFs, datasheets, kernel docs, markdown files
- Intelligent cross-referencing: Connects hardware manuals with kernel documentation
- Expert-level synthesis: Provides comprehensive answers spanning multiple technical domains
- Context-aware responses: Understands embedded Linux ecosystem relationships
get_device_info- Comprehensive device identification (CPU, memory, peripherals, board type)get_driver_info- Kernel module and driver analysis with hardware mappingkernel_hotspots- CPU profiling and performance bottleneck identificationreal_time_analysis- Real-time system analysis (latency, scheduling, determinism)
Ask natural language questions and get expert-level responses with automatic code generation:
- "Why is my GPIO switching slower than expected on BeagleBone Black?"
- "How do I debug device tree issues for my I2C sensor?"
- "What's causing high interrupt latency in my real-time application?"
- "Show me the power management configuration for AM335x"
Built-in Code Generation Capabilities from Gemini CLI:
- Driver code generation: Create complete Linux kernel drivers from specifications
- Device tree overlays: Generate DT overlays for custom hardware configurations
- Kernel module templates: Scaffold new kernel modules with proper structure
- Configuration files: Generate kernel configs, Makefiles, and build scripts
Built-in expertise for Texas Instruments AM335x platform:
- Deep understanding of ARM Cortex-A8 architecture
- Power management and boot sequence knowledge
- GPIO, I2C, SPI, UART configuration expertise
- PRU (Programmable Real-time Unit) integration
- Cape ecosystem and device tree overlay management
- Node.js 18+ (Download)
- Serial hardware
- Google AI API Key or Google account for authentication
# Clone the repository
git clone https://github.com/your-username/kernel-chat.git
cd kernel-chat
# Install dependencies
npm install
#Build
npm run build
# Start Kernel Chat
npm start-
Google Account (recommended for personal use):
- Sign in when prompted for up to 60 requests/minute
-
API Key (for higher limits):
export GEMINI_API_KEY="your_api_key_here"
Get your key from Google AI Studio
# Connect to your embedded device
/serial connect /dev/ttyUSB0 115200- What it does: Converts your plain English questions into appropriate Linux commands
- AI Processing: Automatically generates commands, sends them to your device, and provides intelligent summaries
- Best for: Quick diagnostics, system analysis, and getting insights without knowing exact commands
Examples:
~ Check the dmesg and summarize
~ What processes are consuming the most memory?
~ Show me network interface status- What it does: Sends commands directly to your connected device
- Raw Output: Shows exact command output without AI interpretation
- Best for: Specific commands, debugging, and when you need control
Examples:
>uptime
>free -h
>lsusb
>journalctl -f- What it does: Analyzes recent output captured from your device without sending new commands
- AI Analysis: Reviews logs and data already collected, providing insights and explanations
- Best for: Understanding what happened, pattern analysis, and getting context about previous output
Examples:
>dmesg >>> summarize for errors
>ps aux >>> which processes are using too much memory?
>lsusb >>> are all USB devices working properly?
>cat /var/log/syslog >>> what happened in the last boot?# Get comprehensive device information
> get device infoThis tool will call the get_device_info tool which provides:
- System Identity Overview: High-level "at-a-glance" summary of your Linux system's hardware and software configuration
- Hardware Information: Device model, CPU architecture, and platform details for compatibility assessment
- Software Stack Details: Kernel version, Linux distribution, and user-space environment information
- Initial System Assessment: Quick orientation tool for understanding what you're working with before deep debugging
# Analyze real-time performance
> analyze real-time performance on this systemThis tool will call the real_time_analysis tool which provides:
- RT Kernel Analysis: Checks real-time kernel configuration, preemption settings, and boot parameters
- Interrupt & Scheduling Analysis: Examines IRQ handling, RT task priorities, and CPU affinity settings
- Performance Diagnostics: Identifies latency sources, priority inversions, and timing issues
- System Configuration Review: Analyzes CPU isolation, power management, and memory settings affecting RT performance
- Hardware & Network RT Features: Reviews CPU features, network stack configuration
# Check for kernel performance bottlenecks and interrupt issues
> get me the kernel hotspots connected to serialThis will call the kernel_hotspots tool which provides:
- CPU Utilization Analysis: Shows how busy the processor was during a 5-second monitoring window
- Time Distribution Breakdown: Details how CPU time was split between user space, kernel space, interrupts, and deferred work
- Performance Profiling: Identifies the exact kernel functions consuming the most CPU time (when perf is available)
- Interrupt Analysis: Shows the most frequent hardware interrupts and their rates
- System Health Assessment: Provides a "medical check-up" for your Linux kernel's performance
What This Helps You Debug:
- Performance Issues: Identify if CPU overload is causing system problems
- Driver Problems: Spot kernel functions that might be stuck in loops or consuming excessive resources
- Hardware Malfunctions: Detect interrupt storms or abnormal hardware behavior
- System Bottlenecks: Understand where your kernel is spending its processing time
- Resource Contention: See if specific subsystems (network, I2C, timers) are overactive
> Get driver info from the serial deviceThis tool calls the get_driver_info tool which provides:
Key Information:
- Overview: High-level summary of how many devices were inspected
- Device Count: Total hardware devices detected by the kernel
- Binding Status: How many devices have drivers loaded vs. unbound
- Health Check: Overall assessment of your hardware driver situation
# Ingest documentation
/rag add linux-6.14.11/Documentation/devicetree --tag kernel
/rag add am335x_manual.pdf --tag beaglebone
# Ask complex technical questions
/ask "How do I configure GPIO interrupts on AM335x with proper device tree bindings?" --tag kernel
/ask "What are the power domain considerations for GPIO performance on BeagleBone Black?" --tag <tag you created># Real-world problem solving
> My GPIO switching is only 1MHz instead of 10MHz on BeagleBone Black, what could be wrong?
> I'm getting kernel panics during high CPU load with GPIO operations, help me debug this
> Design a real-time system for motor control with sub-100μs response time on BeagleBone Black# Driver development
> Create a Linux kernel driver for an SPI-connected accelerometer sensor
> Generate a device tree overlay for BeagleBone Black with I2C sensor at address 0x48
> Write a GPIO interrupt handler for BeagleBone Black with proper error handling
# System configuration
> Generate a kernel config for real-time BeagleBone Black system with PRU support
> Create a Makefile for building a custom kernel module with cross-compilation
> Write initialization scripts for embedded Linux system startup/serial connect <port> <baud>- Connect to device/serial send <command>- Send command to device/serial prompt <natural_language>- AI-interpreted command execution/serial summarize [query]- AI analysis of logs/serial disconnect- Close connection
/rag add <path> [--tag TAG]- Ingest documentation/rag list- Show ingested documents/rag status- Show store information/rag clear- Clear documentation store/ask "question"- Query documentation with AI
/memory show- Display AI's current context/memory add <text>- Add information to AI memory/memory refresh- Reload GEMINI.md files
- Development environment only: Not intended for production systems
- Serial access: Tool executes commands directly on connected devices
- Network isolation: Use in controlled, isolated networks
- Audit trail: All commands and responses are logged
- Multi-board support (Raspberry Pi, NVIDIA Jetson, custom SoCs)
- Automated test generation from device tree analysis
- Performance regression detection
- Visual system topology mapping
- Real-time anomaly detection (kernel panics, OOM, IRQ storms)
- Enhanced automated driver generation from datasheet analysis and hardware specs
- Complete system code generation (bootloaders, kernel configs, userspace apps)
- Team collaboration features with shared knowledge base
- Support for other OSs (Zephyr , QNX)
- To support other models (OpenAI, Anthropic, Local LLM's..)
For questions, issues, or contributions related to this Kernel Chat project:
Email: kondaraviteja1@gmail.com
GitHub: @Ravi-Teja-konda
We welcome contributions from the community! Please feel free to submit a pull request.
- AI Model: Google Gemini (1M+ token context)
- Supported Platforms: Linux, macOS, Windows (with WSL)
- Target Hardware: ARM Cortex-A8/A9, ARM64, x86_64 embedded systems
- Serial Protocols: UART, USB-to-Serial adapters
- Documentation Formats: PDF, Markdown, HTML, plain text
- Google Gemini Team for the foundational Gemini-CLI
- BeagleBoard.org for excellent hardware and documentation
- Linux Kernel Community for comprehensive documentation
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
If you find this project useful, consider starring it on GitHub to help others discover it!