One-Click WeChat Article Pipeline from Topic to Draft Box in DSH Local Workbench
Input a trending topic or custom theme → DeepSeek Harness plugin runs a 6-step pipeline → outputs a WeChat Official Account draft ready for review
- Who
- Operators or developers who already have DSH running and want to productize their WeChat Official Account writing workflow
- You get
- Complete topic selection, outline, draft, quality gate, rendering, and image generation in a local Web UI, then push to WeChat Official Account draft box
- You need
- DeepSeek
Constraints
Prerequisites: Node.js ^22.19.0 or >=24.0.0, with npx @deepseek-ai/dsh already working. Requires WeChat Official Account AppID + AppSecret (from the platform admin console "Settings & Development → Basic Configuration"). Image-provider API Key is optional; articles proceed without images if not configured. Credentials and data stay local in ~/.dsh only, never in git, never leaving the machine. This plugin v0.1/v0.2 only supports pushing to the draft box; no mass-publish or freepublish path is available.
What they did
The author developed and open-sourced dsh-wewrite, a DeepSeek Harness plugin that productizes the full WeChat Official Account AI writing pipeline (topic selection, outline, draft, quality gate, rendering, images, draft box) into a local Web workbench. Users install the plugin with one command, configure their Official Account credentials in DSH Web, then start from a trending topic or custom theme. The plugin auto-runs a six-step pipeline to generate the article, supports local editing with AI paragraph rewrite and live WeChat preview, and pushes to the Official Account draft box after human approval.
How it went
The author reports the plugin is released at v0.2.0, compatible with DSH v0.1.x developer preview. The repo CI includes lint, typecheck, 358 unit tests, 52 E2E real-machine cases, P0 visual scanning, and dist-committed consistency checks. The compatibility matrix shows v0.2.0+ passed 358 unit tests + 52 E2E cases green on the 2026-08-20 baseline. (Not tested by this site.)
Worth knowing
1) This plugin v0.1/v0.2 only supports pushing to the draft box (draft/add). There is no code path for mass publish or freepublish; the scheduler target zod literal rejects publish/freepublish/mass by design. 2) WeChat server-side APIs enforce an IP whitelist; if your egress IP changes you must either maintain the whitelist or self-host a relay. 3) DSH v0.1.x is a developer preview with possible breaking changes; the plugin uses feature detection to degrade gracefully. 4) The default image-provider chain is openai (gpt-image-2) → doubao → dashscope → jimeng → minimax → azure_openai → gemini → openrouter → replicate, with automatic fallback; if all fail the article proceeds without images, never blocking publication. 5) AI digest fetching uses an 8-second timeout, 2MB cap, and text/html only; failure degrades to title-based interpretation.
Steps
Install the plugin
Run the install command in your terminal to add the dsh-wewrite plugin to the DSH Web profile. After installation, the "wewrite workbench" entry appears in the DSH Web UI.
Start DSH Web
Launch the DSH Web service and open http://127.0.0.1:3080. Find the "wewrite workbench" tab in the conversation ring, or the "Open Workbench" button at the bottom of the sidebar.
Configure credentials
Open the workbench Settings page, enter your Official Account AppID and AppSecret (from the WeChat platform admin console), then save. Secrets are stored locally only and masked in the UI. Optionally add an image-provider API Key. Click "Connection Test" to verify. If you receive errcode 40164 (IP not in whitelist), add your egress IP to the Official Account whitelist or configure a WeChat API relay address per the FAQ.
Pick a topic and run the writing pipeline
In the "Topics" panel, choose a trending item (built-in Hacker News, with optional custom aggregators). Click to see an AI digest of the linked article. Click "Write this" to start the pipeline, or enter a fixed theme directly in the chat/input box. The pipeline auto-runs six steps: topic → outline → draft → quality gate → render → images, with live progress. A failed gate blocks the default push path; you can revise and retry or explicitly override.
Edit and preview
After the pipeline finishes, enter the editor. Edit in Markdown on the left; a live WeChat preview renders on the right (the HTML matches the final push payload byte-for-byte). Select any paragraph and click "AI Rewrite" in the floating bar, then type a one-sentence instruction (e.g., more casual, cut by half, expand details, more data-driven). AI rewrites only the selected paragraph in-place; press Ctrl+Z to undo. Switch between edit-only, split, and preview-only views.
Push to draft box
When ready, click "Push to Draft Box". If triggered via the agent/dialog, a host-native approval panel pops up first (showing title and gate status). The plugin makes zero WeChat API calls until you click Allow; if approval is unavailable the request is rejected (fail-closed). After success, log in to the WeChat platform admin console under "Content & Interaction → Graphic Materials" to review the draft. Mass publishing must be done manually in the platform admin console; this plugin provides no mass-publish API path.
Prompt you can copy
Reference prompt · Editorial suggestion。Confirm the destination, input variables and output format before use.
This block is the prompt. Ads are never placed here.
No prompt needed — just follow the steps.
After installation, simply tell the DSH agent "use wewrite to write an article about..." to trigger the pipeline. You can also use /wewrite <topic> or @ to reference an existing article.
Source
DeepSeek Harness wewrite 插件:热榜选题→大纲→成稿→配图,审批后推公众号草稿箱 · jerryjiao · GitHub · 2026-08-28T08:32:51Z
Comments & feedback
Tried this method? Share results, problems or improvements. Comments appear after moderation. Please avoid sensitive information.
Loading…
Agent feedback
Reports are submitted by Agents and shared after user confirmation and moderation. They are not platform-verified.