Bring Your Own Key - General Interest Webinar

This feature allows GitHub Copilot Chat and Agent modes to route queries to remotely running LLMs (such as those provided by Anthropic, Google, OpenAI and recently SHARCNET) instead of GitHub’s cloud models.

SHARCNET Brine¶

SHARCNET Brine gives Canadian researchers access to hosted AI models on SHARCNET infrastructure, without the usual HPC queue wait. It provides an always-on, OpenAI-compatible API for hosted models that support chat, tool calling, and audio transcription. The current model list includes:

DeepSeek-V4-Flash-0731
Qwen3.8-27B
North-Mini-Code-1.0

🛡️ Data privacy

Brine runs on hardware in the Graham data centre, so data is processed in Canada rather than sent to a commercial provider outside the country. Even so, users should not submit sensitive data. This includes personal or health information, student or financial records, credentials, confidential third-party data, export-controlled data, Indigenous or community-governed data, and anything you are not authorized to send to a centrally hosted AI service.

🛠️ How to set it up...

Install LiteLLM Provider for GitHub Copilot Chat VSCode extension.
Manage Language Models in VS Code:
- Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P or simply F1) and search for Chat: Manage Language Models.
- Select Add Models $\rightarrow$ LiteLLM.
- Enter Brine for Group Name.
- Point the Base URL to https://brine.sharcnet.ca.
- Enter SHARCNET for Label.
- Provide your API Key when asked.
- Press Enter for the rest.
Switch Models: You can now select the models provided by Brine directly inside the Copilot Chat / Agent dropdown.

✨ Instruction for the AI agent

Add support for writing images in TGA file format to stb.hpp using stbi_write_tga() function that's already included in stb_image_write.h

Pros & Cons¶

Pros

✅ Free for API Key holders
✅ Faster than local LLMs
✅ More capable than local LLMs

Cons

❌ No data privacy
❌ Network dependency