Skip to content

Getting Started

BitVanes is a guided tour through code you didn't write — a pending diff, a fresh AI-generated change, or any function you want to finally understand. It plays the tour back inside your editor: the active statement spotlighted, everything else dimmed, data flow explained step by step.

Two audiences, one tool. Seasoned engineers use it to burn through review diffs quickly (Expert style). Newer developers — and anyone reviewing code an AI generated for them — get plain-language tours that explain the syntax and define the jargon (Learner style).

Choosing your style

Run BitVanes: Set Walkthrough Style (also in the panel's ⋯ menu) or set bitvanes.walkthrough.style:

StyleForWhat changes
ExpertSenior engineers fluent in the languageTerse, high-signal steps; no syntax lessons; invariants and risk only
Standard (default)Working developers new to this codeWhat the code does and why it exists in the flow
LearnerNew devs and vibe codersPlain language, syntax explained on first use, jargon defined, why the pattern matters

Install

text
Search "BitVanes" in the VS Code Extensions view.
bash
code --install-extension bitvanes-v<version>.vsix
bash
git clone https://github.com/BitVanes/BitVanes.git
cd BitVanes && npm ci && npm run compile
# open in VS Code, press F5

Your first tour

  1. Make a change in a git repo — brand-new untracked files count too — or just open a function you want to understand.
  2. Run BitVanes: Walkthrough Working Tree Changes from the Command Palette — or BitVanes: Walkthrough Code Path at Cursor with your cursor inside a function. In a hurry? Instant Walkthrough (No AI) builds a local tour in under a second.
  3. The editor pans to the first step: the active statement is spotlighted, everything outside the enclosing function is dimmed, and a dotted border marks the scope.
  4. Step with alt+] / alt+[, or use the status-bar transport. Hover the highlighted code for the full narrative, variable transitions, and any security note.

Choosing a model

GitHub Copilot (default, zero config)

If you're signed in to Copilot, BitVanes uses VS Code's built-in Language Model API. Nothing to configure, no keys.

Local models — Ollama, LM Studio, vLLM

bash
ollama pull qwen2.5-coder:14b
ollama serve
text
Start the local server (default http://localhost:1234/v1)
and load a chat model.

Then set:

jsonc
{
  "bitvanes.llm.provider": "openai-compatible",
  "bitvanes.llm.baseUrl": "http://localhost:11434/v1", // Ollama; LM Studio: http://localhost:1234/v1
  "bitvanes.llm.modelId": "qwen2.5-coder:14b"
}

Your code never leaves the machine.

Hosted — OpenAI, Anthropic, OpenRouter

Run BitVanes: Set API Key for External LLM Provider, pick a preset, paste your key. The key is stored in VS Code SecretStorage (OS keychain) — there is deliberately no setting that accepts a key.

Keyboard shortcuts

KeyAction
alt+] / alt+[Next / previous step (works anywhere while a walkthrough is active)
] / [Next / previous step when focus is outside the editor and terminal (panel, status bar)
EscExit walkthrough (from the editor) — or the ✕ in the status bar, any time

Released under a proprietary license.