SuperAPI AI Directory

wiseflow

An agile information mining tool that can extract concise information from various sources such as websites, WeChat official accounts, and social platforms, automatically classify it, and upload it to a database.

Open website

About this tool

Xiaobei is a self-media customer acquisition agent tailored for OPCs/small, medium, and micro enterprises. Its underlying architecture is based on openclaw, and it was developed by bigbrother666sh, the author of Wiseflow (formerly AI Chief Intelligence Officer). Its official website is openclaw-for-business.com.

[Main Capabilities] Content production and distribution: WeChat official account article writing, formatting, and publishing; video generation and multi-platform distribution (WeChat Channels, Douyin, Xiaohongshu, Kuaishou, TikTok); posting on platforms such as Twitter/X, Weibo, and Zhihu; publishing content to WeChat Moments (via the WeCom interface). Creation and replication: viral video trend analysis, imitation writing, and re-creation (supports Douyin, Bilibili, and Xiaohongshu video links); DNA-level replication of top accounts in a niche (currently supports four platforms: Xiaohongshu, WeChat Channels, Douyin, and WeChat official accounts). Operations and customer service: data monitoring of published works and daily scheduled reviews; 24/7 intelligent customer service (pre-sales reception); finding potential customers or conducting market research through social media. Intelligence and assistance: information gathering and intelligence support for 18 categories of sources, including Xiaohongshu, Douyin, Weibo, Zhihu, Kuaishou, Bilibili, Twitter, YouTube, WeChat Channels, LinkedIn, Reddit, news, government affairs, finance, academia, shopping, GitHub, etc., with no key configuration required and completely free; product decks, PPT creation, investment/IR material preparation; website design and production, ICP filing, and other material assistance. Interaction method: users only need to communicate with it via WeChat on their mobile phones to use all functions, and it also supports Feishu and WeCom.

[Video and Card Capabilities] Added talking-head video production capability with B-roll, reducing the cost per video from dozens of yuan in the pure AIGC model to a few yuan. B-roll supports slide generation, paper-cut character animation, and operation screen recording. The first two are based on hyperframes, while the latter is based on Xiaobei's browser automation capability. None require calling AIGC models, and users can also directly provide existing materials. Voice-over audio supports user recording and TTS synthesis. TTS supports custom voices and can mix in accents. Supports adding a digital human synchronized picture in a small screen window, requiring only one photo. The above process supports Xiaobei proposing a plan and autonomously delegating to content-producer to complete the entire production process, with users only needing final acceptance. Added card content production capability: built-in "group chat accidental send single image" creative and "Q&A continuous discussion stream three-image" creative, both generated using HTML code, no need AIGC model.

[Usage and Installation] One-click script installation is recommended (the pre-built tarball route), with no need to pre-install Node/pnpm/git throughout. On macOS/Linux, use a one-line bash command; on Windows, use PowerShell, and Windows users need to run PowerShell as administrator. If GitHub can be accessed normally, use the GitHub route (install.sh/install.ps1); for networks in China, use the atomgit route (install-atomgit.sh/install-atomgit.ps1). After the script finishes, the only manual input is filling in AWK_API_KEY. XIAOBEI_TAG can be used to specify the version. Users who have already installed can rerun the install script to upgrade; users who manually git clone can use scripts/update.sh. For debug mode, use ~/xiaobei/bin/openclaw gateway run. To rebind or add a WeChat binding, use channels login to generate a new QR code.

[Models and Costs] It is recommended to subscribe to Alibaba Cloud Bailian's "Token Plan" package. One package covers thinking and dialogue, image generation, TTS speech synthesis, ASR speech recognition, and video generation; for early familiarization, the Lite version at 39 yuan/month is optional, while for regular use the Standard version at 139 yuan/month is recommended. Video generation uses the quota within the Bailian Token Plan by default; after configuring MODELSTUDIO_API_KEY and WORKSPACE_ID, Wan3.0 can be used; Volcano Seedance and MiniMax H3 are also supported. Image generation awk-img-gen calls qwen-image-3.0-pro by default. Music generation can call Bailian business space fun-music-v1. For memory enhancement, an embedding service in OpenAI API format can optionally be integrated. VIP Club is priced at 168 yuan/year, and remote installation and remote technical support value-added services are also provided.

[System Requirements and Limitations] Ubuntu 22.04 is recommended; WSL2/macOS (arm64 + x64) are supported; Windows 10 1803+ (x64, requires Git Bash or WSL). It is recommended to use acomputer that runs 24/7 for deployment, and do not place personal privacy/confidential files on it. For the network, it is recommended to use a normal residential IP; some publishing capabilities require a fixed IP, and the official relay service is provided. Windows symlinks require developer mode or administrator PowerShell, otherwise it falls back to copying. When awk-img-gen explicitly selects a mode and the corresponding credentials are missing, it will report an error and will not switch automatically.