Skip to content
Charlie Bayne edited this page Mar 28, 2026 · 1 revision

AI Assistant

The AI tab provides a conversational interface for exploring and interpreting your metaproteomics data using natural language. You can ask questions about your experiment, request plots, run analyses, and get help interpreting results — without writing any code.

Supported Providers

Conduit-Summit supports four AI providers:

Provider Models Available
Anthropic Claude Haiku 4.5, Claude Sonnet 4.5, Claude Opus 4.5
OpenAI GPT-4o-mini, GPT-4.1-mini, GPT-4o, GPT-4.1
Google Gemini Gemini 1.5 Flash, Gemini 2.0 Flash, Gemini 1.5 Pro
Ollama (local) Any locally running model (no API key required)

Setup

  1. Select your preferred provider from the dropdown.
  2. Enter your API key in the text field. Keys for Anthropic, OpenAI, and Gemini are auto-filled if the corresponding environment variables (ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY) are set in your R environment.
  3. Select your preferred model.
  4. Optionally, provide a brief description of your experiment in the Experiment Description box. This context helps the AI give more relevant responses.

API keys are held in memory for the current session only and are never persisted or logged.

Obtaining API Keys

Data Privacy

When using cloud-based providers (Anthropic, OpenAI, Gemini), summaries of your data structure and analysis results are sent to the provider's API. Raw data is not transmitted. If data privacy is a concern, use the local Ollama option to keep all processing on your machine.

What Can You Ask?

The AI assistant has access to tools that let it interact with your loaded data. Example questions:

  • "How many samples are in my dataset and what metadata variables do I have?"
  • "Generate a PCA plot colored by treatment group."
  • "What are the most differentially abundant proteins between conditions A and B?"
  • "Explain what the volcano plot is showing me."
  • "Run a GSEA analysis on my limma results and summarize the top pathways."

Clone this wiki locally