semble-hooks — automatic code context for every prompt via Semble

[benediktkraus]

What I built

semble-hooks — hooks that automatically inject the most relevant code chunks from your project into every Gemini CLI prompt.

Powered by Semble semantic code search (98% token savings vs grep+read, CPU-only, 16MB static model, no API key).

Gemini CLI integration

• SessionStart → warms up the Semble index for your project
• BeforeAgent → runs semble search with your prompt, injects top-5 code chunks as <relevant-code>
• Installed as Gemini extension under ~/.gemini/extensions/semble-hooks/
• Graceful degradation: if semble is not installed, hooks silently pass through
• Zero npm dependencies

Install

pip install semble
git clone https://github.com/benediktkraus/semble-hooks.git
cd semble-hooks
./install.sh gemini

Also supports Claude Code, Codex CLI, and OpenClaw.

Before vs After

Before: Agent spends 5-10 turns doing grep → read → grep → read to find relevant code.
After: Agent gets the 5 most relevant code chunks in the first turn.

Feedback welcome!

[aniruddhaadak80]

Hey there. Thanks for sharing this.

This is a really interesting idea and something we have been discussing internally. We love seeing how the community is pushing the boundaries of the CLI.

We will definitely keep this in mind for future iterations.