pi-diet-semble

One of my diet context engineering and workflow extensions. Explore the complete collection: https://www.npmjs.com/~probabilityengineer

On-demand Semble semantic code search tools for Pi.

pi-diet-semble gives Pi agents compact, model-visible tools for exploratory code discovery with Semble. It is designed for behavior and intent searches — finding where authentication, persistence, rendering, or other concepts are implemented — without MCP, automatic scans, or large prompt snippets.

Use Semble first for behavior, intent, or conceptual discovery when the relevant symbol or file is unknown. Results are discovery candidates, not exhaustive exact-text matches, syntax matches, or resolved symbol references.

Tools

• semble_search — natural-language, symbol, or identifier code search
• semble_related — find code related to a known file and line

Both tools support includeTextFiles: true to pass Semble's --include-text-files flag for Markdown, YAML, JSON, and other non-code text files.

Requirements

Install the Semble CLI so semble is on your PATH:

pip install semble
# or
uv tool install semble

Install

From npm:

pi install npm:pi-diet-semble

From GitHub:

pi install git:github.com/ProbabilityEngineer/pi-diet-semble

For project-local install, add -l:

pi install -l npm:pi-diet-semble

For local testing:

pi -e ./index.ts

Usage

Ask Pi questions like:

• "Find where authentication is handled."
• "Search for the save model implementation."
• "Find code related to src/auth.ts line 42."

The extension nudges agents to describe the desired behavior in natural language and inspect candidate source locations to verify relevance before drawing conclusions or editing. Its guidance does not assume other extensions are installed.

Prompt overhead

pi-diet-semble registers two compact tools and routing guidance only. It does not inject search results or repository state into prompts.

Development

npm install
npm run typecheck

License

MIT