deeptide  by paean-ai

AI coding agent for codebases and development workflows

Created 1 month ago
507 stars

Top 61.0% on SourcePulse

GitHubView on GitHub
Project Summary

<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> DeepTide is an AI coding agent designed to navigate and assist within codebases, offering three flavors: a native macOS app, a cross-platform CLI via Bun/Node.js, and a Rust CLI. It targets developers seeking integrated AI assistance across diverse environments, providing a flexible, agentic coding assistant that enhances productivity through deep codebase understanding and LLM integration.

How It Works

DeepTide functions as an agentic coding assistant using a plan-tool-observe-adapt loop, equipped with 30+ tools and 25+ slash commands. It features a permission system and hooks engine for customizable workflows. A key advantage is its unified tide-spec interface contract, ensuring consistency across its native Swift macOS app, the TypeScript/Bun CLI (forwarding to @paean-ai/zero-cli), and the developing Rust CLI. It also includes a local inference runtime with a DeepSeek V4 Flash Metal engine and dsgo gateway for offline AI coding.

Quick Start & Requirements

  • macOS (Native App): Install via curl -fsSL https://deeptide.sh/install.sh | sh. Requires macOS 15+. Launch with tide.
  • Linux/Windows (Bun CLI): Prerequisite: Bun installed and on PATH. Install with bun add -g deeptide. Run with tide or deeptide.
  • Rust CLI: Install with npm install -g deeptide-rs.
  • Local Runtime: Build native components on macOS using npm run build:native. Requires Node.js, npm/bun, and Rust toolchain. High memory footprint for local models (e.g., 128 GB+).
  • Links: macOS install script: https://deeptide.sh/install.sh.

Highlighted Details

  • Agentic Workflow: Supports multi-turn agentic loops (plan -> tool -> observe -> adapt).
  • Extensibility: Features 30+ built-in tools, 25+ slash commands, hooks engine, and sub-agent definition via markdown.
  • LLM Flexibility: Integrates with DeepSeek API by default and supports Anthropic-protocol-compatible endpoints via BYOK.
  • Local Inference: Includes DeepSeek V4 Flash Metal engine and dsgo gateway for offline use, with significant memory requirements for higher fidelity.

Maintenance & Community

Developed by DeepSeek and authored by a8e. Issue tracking is in this repo; CLI issues often forwarded to @paean-ai/zero-cli. Security vulnerabilities should be reported privately. The Rust port is intended to become the canonical cross-platform CLI.

Licensing & Compatibility

Licensed under MIT, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

Native macOS app requires macOS 15+. Bun CLI needs Bun at runtime. High-fidelity local inference models demand substantial system memory (128 GB+). CLI bugs/features are handled upstream in @paean-ai/zero-cli.

Health Check
Last Commit

5 days ago

Responsiveness

Inactive

Pull Requests (30d)
134
Issues (30d)
0
Star History
426 stars in the last 30 days

Explore Similar Projects

Starred by Yineng Zhang Yineng Zhang(Inference Lead at SGLang; Research Scientist at Together AI) and Yaowei Zheng Yaowei Zheng(Author of LLaMA-Factory).

kimi-code by MoonshotAI

17.9%
2k
AI coding agent for terminal-based development
Created 3 weeks ago
Updated 12 hours ago
Feedback? Help us improve.