CHAIN 239020.XYZ

AutoWeChat

Enterprise Human-in-the-Loop AI WeChat Publishing Workbench

A compliance-first AI publishing workbench built strictly around human-in-the-loop review. Orchestrates 4 automated editorial workflows, 12 hotspot intelligence sources, and 4x parallel DAG sub-agents, writing exclusively to the WeChat Draft Box after fact-checking and AI Critic gating.

  • Python
  • APScheduler
  • trafilatura
  • Knowledge Graph DAG
  • ThreadPoolExecutor
  • SQLite
  • AI Critic
  • WeChat Draft API
Plate
PLATE 03
Role
Primary Author
Status
Production Architecture · Compliant
Year
2026
Scope
Primary Author: Independently engineered the Human-in-the-Loop compliance boundary, 4 editorial workflows, Knowledge Graph DAG parallel writers, fact-checking Critic gate, and 5-dimension console.
PLATE 03 · DECK 4 SLIDES

Interactive HTML Slide Deck

Navigate with controls, keyboard arrows, slide indicators, or Fullscreen presentation mode.

PLATE 03 COMPLIANCE ARCHITECTURE
SLIDE 01 / 04

Human-in-the-Loop Compliance & Draft-Box Boundary

LEFT · AUTOMATED PIPELINE
12 Sources + APSchedulerScheduled ingest & SQLite dedup
trafilatura + DAG 4xDeep scrape & parallel drafting
Fact-Check + AI CriticLive verification & rubric gate
MANDATORY DRAFT BOX BOUNDARYWeChat Draft API Only0% Unsupervised Broadcast
RIGHT · HUMAN AUDIT WORKBENCH
Review citations & fact-check diff
Inspect AI Critic score & tone
Manual editorial publish · 100% audit
100% Human-in-the-Loop
Restricted strictly to WeChat Draft API writesPreserves source citations and audit checklistsFailure isolation prevents corrupted draft pushes
SLIDE 01 / 04
Open HTML Deck

An explicit architectural firewall stops automation at the WeChat Draft Box, guaranteeing 100% human audit before publication.

  • Restricted strictly to WeChat Draft API writes
  • Preserves source citations and audit checklists
  • Failure isolation prevents corrupted draft pushes
  • 4 Automated Workflows
  • 12 Hotspot Data Sources
  • 4x Parallel Sub-Agents
  • 100% Mandatory Human Audit

Context & Engineering Constraints

Unsupervised autoposting bots introduce factual hallucinations, duplicate coverage, and platform compliance risks, while manual cross-source monitoring and long-form technical drafting are prohibitively slow.

System Architecture

  1. 01

    Human-in-the-Loop Compliance Firewall

    API integration is strictly scoped to the WeChat Draft API so every generated issue passes through a mandatory human editorial sign-off desk before broadcast.

  2. 02

    4 Scheduled Workflows & 4x DAG Parallel Drafting

    APScheduler runs hotspots (12 sources), github Trending, aikepu (Knowledge Graph DAG + ThreadPoolExecutor 4-sub-agent drafting), and daily_knowledge pipelines.

  3. 03

    trafilatura Deep Scrape + Fact-Check + AI Critic Gate

    Extracts full article text, runs live web verification and multi-pass AI Critic scoring with automated rewrite loops, and checks SQLite history to block duplicate topics.

Engineering Highlights

  • 100% draft-box delivery boundary eliminates unsupervised publishing while preserving automated research and drafting velocity.
  • The aikepu workflow decomposes long-form explainers via a Knowledge Graph DAG and drafts chapters concurrently with 4 ThreadPoolExecutor sub-agents.
  • Combines trafilatura full-text extraction, live search fact verification, and an independent AI Critic rubric to reject weak drafts automatically.
  • Provides a 5-dimension Linear-inspired light console tracking workflow queues, fact-check logs, Critic scores, SQLite deduplication, and audit sign-off.