docs: update 2 files and create 1 file

Author: MichaelCurrin

README.md

@@ -8,7 +8,8 @@ This extensions adds a button to the Git extension sidebar - clicking it sends a
 
 ## Note
 
-This extension is not available in the extensions library. You can build and install it locally though.
+- This extension is not available in the extensions library. You can build and install it locally though.
+- This was developed with Cursor quickly, the code and quality has not been verified.
 
 ## Benefits
 

docs/development/about.md

@@ -0,0 +1,35 @@
+# About
+
+## Overview
+
+This extension adds a command and SCM title-bar action to generate a Git commit message using a local LLM (OpenAI-compatible API such as Ollama). It gathers a unified Git diff, prompts the model to produce a conventional commit-style message, and places the result in the Source Control input box.
+
+## Libraries and APIs
+
+- **VS Code**: Uses `vscode` to register the command, show progress and notifications, access the Git extension (`vscode.git`), and write to the SCM input box or clipboard.
+- **Node.js stdlib**: Uses `child_process.exec` (promisified) to run `git` as a fallback for diffs.
+- **HTTP**: Uses global `fetch` in the VS Code runtime to call an OpenAI-compatible chat completions endpoint.
+
+## Getting the Git diff
+
+- **Primary path (Git API)**: Retrieves the first repository from `vscode.git` and tries `repo.diffWithHEAD()`; if unavailable, tries `repo.diffIndexWithHEAD()`.
+- **Fallback (CLI)**: If the API path fails or doesn’t yield a string, runs `git diff --staged` in the repository root using `exec`.
+- **Validation**: If the resulting diff is empty, the command errors and prompts you to stage changes.
+
+## Calling the LLM
+
+- **Endpoint and model**: Configured via Settings → `llmCommitMsg.endpoint` (default `http://localhost:11434/v1`) and `llmCommitMsg.model` (default `gemma3`).
+- **Prompts**: Sends a system prompt to guide style and a user prompt embedding the unified diff in a fenced `diff` block; truncates very large diffs to keep requests reasonable.
+- **Request**: POSTs to `<endpoint>/chat/completions` with `stream: false`, then reads `choices[0].message.content` from the response.
+- **Cleanup**: Strips outer triple-backtick fences if present to return plain text.
+
+## Applying the result in VS Code
+
+- If available, writes the generated message to `repo.inputBox.value` so it appears in the SCM view.
+- If not, copies to the clipboard and shows a notification.
+- The command runs under a progress notification and surfaces success/error messages.
+
+## Configuration
+
+- **llmCommitMsg.endpoint**: Base URL of the OpenAI-compatible API (e.g., Ollama). Default: `http://localhost:11434/v1`.
+- **llmCommitMsg.model**: Model name used for generation. Default: `gemma3`.

docs/usage.md

@@ -4,7 +4,7 @@ Generate a commit message from your staged changes.
 
 ## Steps
 
-1. Stage the changes you want included in the message.
+1. **Stage** the changes you want included in the message.
 1. Open the Source Control view in VS Code.
 1. Click the `Generate Commit Message (LLM)` action in the SCM title bar, or run the command from the Command Palette.
 1. The extension fills the commit message input. If it cannot, it copies the message to the clipboard and shows a notification.