DaatLocus

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Uyari
  • process.env — Environment variable access in webui/.pnp.loader.mjs
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

An agent runtime that truly has experience.

README.md
Daat Locus Logo

Daat Locus

简体中文
Crates.io
CI
License

An agent runtime that truly has experience.

What Is This?

Daat Locus is a long-running local self-governing Agent Runtime.

It is built for work that becomes better through history: maintaining the same
project over time, repeatedly handling the same class of task, remembering your
preferences and practical experience, and distilling them to improve later
behavior.

Core Ideas

Apps For Agents

When humans use a computer, we rarely choose an action from a global list of
everything the machine can do. We open a terminal, read the current output,
enter a command, and wait for the result; or we open a browser, read the current
page, click, navigate, and continue from the new page.

Daat Locus gives agents a similar interaction model.

Apps provide stateful operating surfaces for the runtime. Each App renders the
current state the agent can see, explains when it should be used, explains how
it should be operated, and exposes a local set of tools when focused.

Compared with a flat tool list, this gives the model three things:

  1. Locality: the agent only sees tools relevant to the current operating
    surface.
  2. State grounding: actions are based on the state currently displayed by
    the App, instead of choosing tools out of context.
  3. Temporal continuity: long-running surfaces such as Terminal and Browser
    can be safely continued.

Apps are how Daat Locus turns "tools" into "software operating surfaces".

Therefore, Daat Locus does not need SKILLS.md to explain how a group of tools
should be used. The App itself is self-describing.

Workflow Self-Improvement

Daat Locus executes tasks with workflows as blueprints, then feeds execution
experience back into workflows during an independent sleep phase.

While awake, Daat Locus executes tasks and records practical experience. During
sleep, it organizes that experience, fixes recurring problems, and improves the
workflows that later tasks depend on.

Sleep optimization also attempts to merge similar workflows to avoid unbounded
growth.

Quick Start

The recommended install path is cargo-binstall, which installs the prebuilt
GitHub Release binary for your platform. Normal installs do not need Python,
uv, or PyInstaller.

cargo install cargo-binstall
cargo binstall daat-locus

You can also download the matching archive directly from
GitHub Releases, extract it, and place daat-locus on your
PATH.

On first launch, Daat Locus opens an interactive setup flow.

Source Builds

cargo install daat-locus is available from crates.io. Source builds require
Node.js with Corepack or Yarn available because build.rs builds and embeds
the WebUI.

git clone https://github.com/shadow3aaa/DaatLocus
cd DaatLocus
cargo run --locked

cargo build and cargo run build the WebUI through build.rs and embed the
generated assets into the daemon by default. For a release-style local binary,
run cargo build -p daat-locus --release --locked directly.

Documentation

License

Daat Locus is licensed under the Apache License 2.0.

Yorumlar (0)

Sonuc bulunamadi