Graft: AI Assistant for Grafana

An open-source AI assistant plugin for Grafana, enabling natural language interaction with your observability data.

Features

• Natural Language Queries - Ask questions about your dashboards, metrics, logs, traces, alerts or more
• Dual Model Support - Configure separate models for quick responses (Standard) and complex reasoning tasks (Deep Research/Thinking)
• Multiple LLM Providers - Works with Anthropic, OpenAI-compatible APIs as well as Ollama, and LM Studio for local inference
• Grafana MCP Tools - Leverage Grafana's built-in MCP tools for interacting with your Grafana instance
• Chat History - Browse and continue previous conversations
• Prompt Library - Save and reuse commonly used prompts with customizable templates

Screenshots

Chat Interface

Tool Execution

Rich Content Rendering

Prompt Library

Previous Conversations

View all screenshots

Requirements

• Grafana 10.4.0 or later
• Grafana LLM Plugin 1.0.0 or later
• An LLM provider (OpenAI, Anthropic, LM Studio, or Ollama/LM Studio local inference)

Installation

1. Download the latest release from the Releases page
2. Extract the archive to your Grafana plugins directory (typically /var/lib/grafana/plugins/)
3. Configure Grafana to allow the unsigned plugin (see note below)
4. Restart Grafana

Note: Unsigned Plugin

This plugin is currently unsigned. To install it, you must configure Grafana to allow unsigned plugins by adding the following to your grafana.ini or custom.ini:

[plugins]
allow_loading_unsigned_plugins = vikshana-graft-app

Or set the environment variable:

GF_PLUGINS_ALLOW_LOADING_UNSIGNED_PLUGINS=vikshana-graft-app

Configuration

1. Navigate to Administration > Plugins > Graft AI Assistant
2. Click Enable to activate the plugin

LLM Configuration

Graft uses the Grafana LLM Plugin for model configuration. Configure your LLM providers in the Grafana LLM Plugin settings.

Usage

1. Click Graft AI Assistant in the Grafana sidebar under App section
2. Type your question in the chat input
3. Toggle between Standard and Deep Research modes as needed
4. View tool executions and reasoning in expandable sections

Example Queries

• "What's the current CPU usage across all nodes?"
• "Show me error rates for the payment service over the last hour"
• "Explain the spike in memory usage I'm seeing on the dashboard"
• "Help me create an alert for when response times exceed 500ms"

Development

See Development Guide for build instructions, testing, and local development setup.

License

Graft is distributed under AGPL-3.0. See LICENSE for details.