A VS Code chat extension that automatically routes your prompts to the best GitHub Copilot model based on complexity, context size, and conversation stage. It can also intelligently attach your workspace files to the model’s context when you ask about your project.
Features
Three model tiers: fast, balanced, powerful – configure which Copilot model goes where.
Automatic complexity analysis: keyword‑based scoring decides the tier.
Conversation awareness: new conversations start with a capable model, follow‑ups become cheaper.
Workspace context injection: let the model “see” your code when needed.
Optional token saving: long conversations are summarised to reduce token usage.
Workspace Context
By default, the router attaches relevant project files only when your prompt hints at it (e.g., “analyze this codebase”, “explain the architecture”). This gives the model actual source code to work with, without wasting tokens on every small query.
Modes:
smart (default) – only attaches workspace files when the prompt contains words like project, codebase, entire, workspace, repository, etc.
always – attaches a representative snapshot of your workspace on every @smart request.
never – no files are attached (original behavior).
What gets attached:
The first 50 files (content truncated to 4,000 characters each).
For larger projects, a list of all other files (just names).
Automatically skips node_modules, .git, build outputs, and respects .gitignore.
Usage
Open VS Code Chat (Ctrl+Shift+I / Cmd+Shift+I).
Type @smart followed by your prompt.
The extension will:
Display the chosen tier and model name.
Attach workspace files if the mode allows it and your prompt triggers the condition.
Send the prompt to the selected model.
Keep track of the conversation for smarter future routing.
Configuration
Setting
Default
Description
aiRouter.fastModel
GPT-5 mini
Model for simple queries
aiRouter.balancedModel
Raptor mini
Model for normal tasks
aiRouter.powerfulModel
GPT-5.5
Model for complex/initial tasks
aiRouter.complexityThreshold
0.6
Score above which a more powerful model is used
aiRouter.tokenLimitBeforeSummarization
80000
Token estimate before auto‑summarisation
aiRouter.enableAutoSummarization
true
Enable conversation summarisation
aiRouter.workspaceContextMode
smart
When to attach workspace files: smart, always, never
Build & Run
npm install
npm run compile
Then press F5 in VS Code to launch a new Extension Development Host.
