BetaStreetOmnis/xhs_ai_publisher

AI-powered Xiaohongshu/Rednote content creation and publishing tool with PyQt desktop UI, FastAPI service, login-state reuse, preview publish, and automated browser workflows.

What it is

A desktop/web tool that automates content creation and publishing on Xiaohongshu (Little Red Book), a Chinese social media platform similar to Instagram. It uses AI language models to generate post titles and text, then automates the browser to log in and publish — saving creators the repetitive work of writing and posting.

How it works

  1. AI content generation — You type a topic (or pick a trending item from built-in hot lists from Weibo/Baidu/Bilibili). The tool calls an LLM API (OpenAI-compatible, Claude, or Ollama — you configure your own key) to write a title and body. If no API is configured, it falls back to built-in templates.
  2. Image handling — It matches images to the post, lets you upload your own, and can generate cover images and multi-page content images from built-in templates (marketing posters, banners, product shots).
  3. Publishing — Using Playwright browser automation, it logs into Xiaohongshu (phone + verification code, or importing an existing login session from Chrome), previews the post, and publishes it. It supports scheduled publishing for unattended posting.
  4. Multi-account & safety — Multiple user profiles with isolated login states and data; API keys are encrypted locally; all data lives in ~/.xhs_system/.

Tech stack

  • Python 3.8+, PyQt5 desktop GUI, FastAPI + Playwright for a web/API mode (Docker deployable)
  • Local SQLite database, JSON config, encrypted key storage
  • Extensible prompt templates (templates/prompts/*.json) and image templates

Notable extras

  • Web link import: paste a URL (e.g., a WeChat article) and it extracts title/body/images into a draft
  • Hotspot data center: pulls trending topics from multiple Chinese platforms, one click to use as a post topic
  • A documented workflow for controlling the tool via OpenAI's Codex agent from a phone (using a third-party bridge app)

In short: a practical, feature-complete automation tool for Xiaohongshu creators that leans on LLM APIs for the writing part and browser automation for the posting part. It's not building AI models itself — it's an AI-powered application.

Related

  • Project
    injaneity/pi-computer-use

    A Pi extension that enables AI agents to interact with desktop applications by observing UI elements and performing actions like clicking and typing across macOS, Windows, and Linux.

  • Project
    A9T9/RPA

    Ui.Vision is an open‑source browser extension (with optional desktop modules) for recording and running automation macros. It supports classic Selenium‑style commands, OCR/image‑based visual actions, and a Model Context Protocol (MCP) bridge that lets LLM assistants (Claude Code, Cursor, etc.) generate, edit, and execute macros via the `uiv.*` JavaScript API.

  • Project
    ShenSeanChen/waku-agent

    A local-first personal AI assistant with a transparent, lightweight codebase that implements core agent pillars like memory, loops, and evaluation.

  • Project
    blackholll/loonflow

    An open-source process automation platform that uses visual designers to simplify enterprise workflows and provides an MCP server for AI-assisted ticket management.