lake

A terminal LLM agent for Windows and Linux.

What it is

lake is a CLI coding agent that runs against your own LLM server — local llama.cpp / ollama / LM Studio, or any OpenAI-compatible endpoint you control. It ships as a single AOT-published binary per RID, has a focused built-in tool family (filesystem, shell, terminal sessions, HTTP, web search, PDF, image), and persists every session locally so resume / replay / checkpointing all work out of the box.

Recent releases

• 0.1.75 save_to on retrieve.web and http.request — stream downloads directly to a workspace path (binaries, archives, anything too big to inline).
• 0.1.74 Windows Ctrl-C orphan-process fix: Environment.Exit safety net + per-turn IO catch + bounded terminal-disposal.
• 0.1.73 Linux Ctrl-C handling: catch the EIO from PrettyPrompt's cursor-probe right after signal delivery.
• 0.1.71 Backend HTTP retry on 502 / 503 / 504 / 429 with exponential backoff; strip nginx HTML walls from error messages.
• 0.1.70 HTML syntax highlighting dispatches embedded <script> / <style> bodies to the JS / CSS tokenizers.
• 0.1.68 Syntax highlighting for fenced code blocks across 10 languages (csharp, json, html, css, js/ts, python, bash, powershell, yaml, sql).
• 0.1.65 Plan-mode steering: per-turn user-prompt prefix on top of the system-prompt addendum.
• 0.1.63 Yolo agent mode + colour-coded mode tag in the status row (plan = green, auto = yellow, yolo = red).
• 0.1.61 Named sessions: /rename, /resume, auto-naming after N turns via purposes.general.

Currently shipping: 0.5.139. Installation, manifest, and per-RID artifacts are under a separate access-gated path on this host.