gptel-slim-tools - Surgical context management for faster LLM workflows

[Jeremias-A-Queiroz]

Hi everyone!

I’d like to share a project I’ve been developing to bridge a gap in my daily interaction with LLMs in Emacs: emacs-gptel-slim-tools.

The goal is simple: Maximum precision with minimal context. Instead of copying/pasting fragments or sending entire files (which can be noisy and token-heavy), these tools allow you to inject specific code blocks or metadata directly into gptel with surgical precision.

The Background:
I work as a SysAdmin/DevOps in a heavily regulated healthcare cooperative where the Development and Infrastructure teams are strictly separated. I often receive deployment tasks for applications I didn't write and that are sometimes poorly documented.

Before these tools, I spent hours inspecting source code or pinging developers to understand how to build the proper manifests (especially Secrets and ConfigMaps). While gptel is a lifesaver, the friction of manual context management was still a bottleneck.

The Workflow:
With gptel-slim-tools, I can now:

• Identify precisely what the LLM needs to see (env vars, logic blocks, or config patterns) without leaving the buffer.
• Eliminate the "Copy-Paste" fatigue by sending exactly what matters in seconds.
• Identify inconsistencies—like hardcoded variables that should be in a manifest—before they hit production.

I’m sharing this because I believe other SysAdmins using Emacs might find it useful for bridging the gap between "code I didn't write" and "infrastructure I must maintain." However, I believe this is equally valuable for anyone who needs to investigate large codebases or get LLM support on the fly without constant context "noise."

The Repo: https://github.com/jeremias-a-queiroz/emacs-gptel-slim-tools

I’d love to hear your thoughts or how you manage context "noise" in your own workflows!