# LiveStream Agent Studio: Four-Agent Workflow for Douyin Live Commerce Teams > LiveStream Agent Studio is a Windows desktop application that combines creator scouting, livestream transcript processing, performance review, and short-video scripting in one local-first workspace. Its first public release is aimed at Douyin live e-commerce operations teams, and the README frames it as ready for evaluation, testing, and internal research rather than production deployment. **HanyuanWang/LiveStream-Agent-Studio** — 面向抖音直播电商的 Windows 本地 AI Agent Studio，贯通主播发现、直播洞察、直播复盘与短视频内容编导的统一智能工作流。 - Repository: https://github.com/HanyuanWang/LiveStream-Agent-Studio - Stars: 1,038 · Forks: 171 - Language: Python - License: MIT - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/hanyuanwang-livestream-agent-studio ## Solving the Fragmented Workflow in Douyin Live Commerce Operations Douyin live commerce typically involves four distinct stages of work: researching candidate creators, processing full-length livestream recordings into searchable transcripts, aligning transcript events against traffic data, and drafting short-form video scripts. Each stage historically has its own toolset. The task center in the README is described as reducing the need to switch repeatedly between scripts, folders, and tools. LiveStream Agent Studio addresses this by placing all four workflows inside a single Windows desktop application. The target users are operations, media-buying, content, and review teams running Douyin livestream commerce, particularly teams that already have access to Chanmama rankings and use Ocean Engine E-commerce traffic exports. The tool is not a general-purpose AI assistant. Its agents are built around specific data formats: Chanmama rankings, Ocean Engine minute-level traffic tables, and the Alibaba Cloud DashScope transcription service. ## How the Four-Agent Pipeline Processes Livestream Data Each of the four agents handles a distinct part of the live commerce workflow, and they are designed to chain together. Creator Scout reads Chanmama rankings or individual creator pages that are visible to the user's signed-in account. It generates structured creator profiles and can add selected creators to an optional recording workflow. Coverage is bounded by what the user's Chanmama account can see. Live Breakdown takes a complete livestream recording as input, extracts its audio, submits it for cloud transcription through DashScope, and produces an Excel workbook with second-level timestamps. This step converts an unstructured video file into a structured data artifact that the following agents can consume. Live Review aligns the timestamped transcript from the breakdown stage with minute-level traffic data from Ocean Engine E-commerce. It correlates speech phases against entries, exits, concurrent viewer counts, watch time, engagement, product impressions, and product clicks, then exports an Excel and Word report. This is the analytical core of the product, and it requires data from both the breakdown step and an Ocean Engine E-commerce traffic export from the same session. Video Director takes a short-video reference link provided by the user, extracts a verified transcript from it, and generates an original spoken script, content structure, storyboard, and production notes. The README explicitly states that the output is based on the content logic of the reference, not a copy of its wording. A unified task center tracks progress, runtime status, failure reasons, and output file paths for all running tasks in one view. ## Installing and Configuring LiveStream Agent Studio on Windows LiveStream Agent Studio ships as a prebuilt Windows package rather than a Python source install. The release page provides two files to download: - LiveAgent-Studio-Windows-x64.zip - LiveAgent-Studio-Windows-x64.zip.sha256 The README instructs users to verify the download before extracting it. Open PowerShell in the directory containing the ZIP file and run: ```powershell (Get-FileHash .\LiveAgent-Studio-Windows-x64.zip -Algorithm SHA256).Hash ``` Compare the output against the value in the .sha256 file. The README explicitly states not to run the package if the hash does not match or if the download source cannot be verified. After a successful check, extract the complete ZIP archive. The README notes that downloading only the source archive or copying only the executable is not enough: the .runtime directory and bundled Agent directories in the ZIP are required for the application to function. Launch the application by double-clicking: ```text LiveAgentStudio.exe ``` When the local service is ready, the application opens http://127.0.0.1:4173/ in the default browser. The README warns that the current Beta launcher does not carry a commercial code-signing certificate, so Windows SmartScreen may show a warning. The advice in the README is to run only packages downloaded from the official Releases page and verified against the published SHA-256 checksum. Once the application starts, open Settings & Connections from the sidebar. The initial setup requires four items: an Alibaba Cloud Model Studio DashScope API key, a private OSS bucket name, a RAM AccessKey restricted to the required bucket and object prefix, and optionally the full local path to a recording assistant executable for the Kuaidou integration. After saving, select Verify Qwen & OSS to confirm that both services are reachable before running any task. ## Windows-Only and DashScope-Locked: The Hard Platform Limits The application is Windows-only. The release package is named LiveAgent-Studio-Windows-x64.zip, and the README describes the product as local-first on Windows. No macOS or Linux version is mentioned anywhere in the repository. Teams whose engineers or analysts work primarily on other platforms cannot run the tool natively. All AI inference goes through Alibaba Cloud Model Studio (DashScope). There is no configuration option to point the agents at a different LLM provider. Teams that need to keep inference traffic outside Alibaba Cloud infrastructure, whether for regulatory, contractual, or security reasons, have no supported path to comply while using this tool. Creator Scout is bounded by the user's Chanmama account. Teams that do not subscribe to Chanmama, or whose account tier restricts access to certain rankings, will see a reduced pool of creator data in the scouting stage. The Live Review agent requires a minute-level traffic export from Ocean Engine E-commerce. Teams running live commerce on platforms that do not generate compatible Ocean Engine traffic exports cannot use the traffic-alignment feature. The v1.0.0 README frames the release as intended for evaluation, testing, and internal research. The opening notice asks users to validate with non-sensitive test materials before investing the tool in production workflows. Running it in high-volume or revenue-critical pipelines before that validation is against the explicit guidance in the repository. ## LiveStream Agent Studio vs. n8n: Specificity Against Generality n8n is a general-purpose workflow automation platform that runs as a self-hosted web service or in a managed cloud. It provides a visual node editor where users can connect external APIs, parse data, and trigger downstream steps. The fundamental difference is that n8n is domain-agnostic: building a live commerce pipeline in n8n means authoring every integration from scratch, writing the prompt logic for each AI call manually, and handling the Excel and Word export formats without any built-in support for Chanmama, DashScope transcription, or Ocean Engine traffic data. LiveStream Agent Studio trades that flexibility for ready-made depth in one domain. The four agents ship with the data models and output formats for Douyin live commerce already defined, which means an operations team can run a complete breakdown-to-review workflow without writing a single integration. The practical trade-off is clear. Teams on Windows that are within the Alibaba Cloud ecosystem and work with Douyin live commerce data get a complete pipeline with no build time. Teams that need cross-platform support, want to use a different LLM provider, or need to adapt the workflow logic for a different commerce platform get more from a general-purpose tool like n8n, at the cost of building and maintaining the pipeline themselves. ## MIT License, First Release, and the Risk of Upstream API Changes The repository is released under the MIT License. The THIRD_PARTY_NOTICES.md file lists third-party components with their respective licenses. Teams that need to audit dependencies before deploying internally should review that file. The project published v1.0.0 on 2026-08-22 and the last push to the main branch was on 2026-08-24. The CHANGELOG.md and CONTRIBUTING.md files are present in the repository, which is consistent with a project expecting continued development. However, only one public release exists. Teams that adopt the tool early carry a specific risk: both DashScope and Chanmama are external services, and if either changes its API response format or page structure, the corresponding agent breaks until the project is updated. With only one release as a baseline, there is no upgrade history to judge how quickly such breaks would be addressed. The SECURITY.md file in the repository suggests the team has defined a vulnerability disclosure process. That matters because the application handles cloud API credentials and processes video content that may include commercially sensitive material. ## Conclusion LiveStream Agent Studio is the right choice for Douyin live commerce teams on Windows that already have Alibaba Cloud accounts and work with Chanmama and Ocean Engine data. Its four pre-built agents cover a specific, real workflow without requiring any custom integration work. Teams on macOS or Linux, those without DashScope access, or those that need to stay outside the Alibaba Cloud ecosystem should look elsewhere. Before deploying beyond evaluation, verify that the SHA-256 checksum passes and that the DashScope and OSS verification in the settings hub succeeds. ## FAQ ### What AI model does LiveStream Agent Studio use for transcription and analysis? LiveStream Agent Studio routes all inference through Alibaba Cloud Model Studio (DashScope), configured with the user's own API key in the settings hub. The README references Qwen verification but does not specify which model version is selected by default. ### Does LiveStream Agent Studio run on macOS or Linux? The application is described as Windows local-first, and the only release package is named LiveAgent-Studio-Windows-x64.zip. The README does not mention macOS or Linux support. ### What data does LiveStream Agent Studio send outside the local machine? The Live Breakdown agent submits extracted audio to Alibaba Cloud DashScope for cloud transcription. All agents write intermediate and output files to a private OSS bucket configured by the user with their own credentials. The README advises validating with non-sensitive test materials before production use. ## Sources - [HanyuanWang/LiveStream-Agent-Studio on GitHub](https://github.com/HanyuanWang/LiveStream-Agent-Studio) - [Issues](https://github.com/HanyuanWang/LiveStream-Agent-Studio/issues) - [License: MIT](https://github.com/HanyuanWang/LiveStream-Agent-Studio/blob/main/LICENSE) - [README](https://github.com/HanyuanWang/LiveStream-Agent-Studio/blob/main/README.md) - [Releases](https://github.com/HanyuanWang/LiveStream-Agent-Studio/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/hanyuanwang-livestream-agent-studio