Runic: Interactive AI Research tool

• 

  Runic: AI beyond just walls of txt

• 

  Java Script for simulations

• 

  Image generation using Nano Banana

• 

  Stays grounded using real data source like Wiki pedia

• 

  Using Gemini interactions API to execute Python code for complex problems

• 

  When installed locally the user can add their own API key to use Gemini

• 

  Right-click images lets you save them to the project.

• 

  Project overviews help you keep track of all you projects.

• 

  Review Note, Fact, and questions you discovered. Highlighting text on any canvas lets you save it to notes.

• 

  Executing python

• 

  Real-time data for Space weather and USGS

• 

  Or just create a spot to discover new recipes and find new favorites

Runic: ​The Research OS — Grounding AI in Reality through Python and Native Context. ​The Problem: The "Wall of Text" Trap ​Most AI research tools suffer from two fatal flaws: Hallucinations and Passivity. ​Users are buried in "Walls of Text" that they have to fact-check manually. ​The research process is fragmented—users find a source in one tab, analyze it in another, and write their report in a third. ​There is no "fluidity" between finding a fact and proving its validity through data. ​The Solution: Runic ​Runic is a high-performance research environment that transforms Gemini from a chatbot into a Digital Lab Assistant. It bridges the gap between raw information and verified insights by allowing users to harvest "Native Context" and trigger real-time data simulations. ​With Runic, research isn't a conversation; it's an extraction. ​The Tech Stack ​Frontend: Svelte 5 (utilizing Runes for fine-grained, high-performance state management). ​State Logic: Reactive Svelte Store architecture for instant snapshotting and context persistence. ​AI Engine: Gemini 3 Flash (optimized for speed and massive context windows).
​Sandbox: Secure IFrame sandboxing for executing dynamic HTML/JS research components. ​Analysis: Native Python Code Execution via the Gemini API.
​How We Used Gemini 3 ​We didn't just use Gemini for text; we leveraged the full Gemini Tool-Kit: ​Google Search Grounding: We integrated native grounding to ensure every research claim is backed by real-world, 2026-current data, eliminating the "knowledge cutoff" issue.
​Code Execution (The Lab): When users need data analyzed, Gemini writes and executes Python in a sandboxed environment. This powers our Gallery Component, where scientific charts are generated on the fly.
​Massive Context Window: Runic utilizes the large context window to hold "Snapshots" of entire research sessions, allowing users to pivot between topics without the AI losing the thread. ​Multimodal Asset Management: Gemini produces high-fidelity data visualizations (Base64) which Runic automatically captures, captions, and prepares for final reporting. ​The Journey: From Chatbot to OS ​The journey began with a simple question: "Why do I have to copy-paste from my AI into a spreadsheet?" The development process involved: ​Architecting the "Active Context": Moving away from standard chat history and building a "Highlight-to-Context" feature where the user decides what the AI remembers. ​Mastering Svelte 5: Implementing the new Runes ($state, $effect) to handle the high-speed data flow between the AI's Python output and the UI's Gallery. ​Refining the "Hybrid" UI: Creating a workspace that balances a Wiki-style reader, a Python console, and a visual asset gallery—all on one screen. ​What's Next for Runic? ​Collaborative Context: Multi-user research sessions where a team can "seed" a shared context. ​Deep PDF Orchestration: More complex document templating to turn research sessions into publication-ready papers instantly. ​Expanded Toolsets: Integrating more specialized API tools for even deeper scientific verification.

Built With

• firebase
• gemini
• javascript
• svelte5
• vite