What this case covers
Shi Gengxin starts from WeChat history, phone articles, computer web pages, and old documents, using Codex to connect IMA, Obsidian, and Feishu, building a content system from collection, verification, decomposition, to distribution. The article provides a four-directory minimal structure, and an acceptance method for running the first workflow through from one piece of content.
Questions it answers
- What roles do IMA, Obsidian, Feishu, and Codex each play?
- How does original text from different sources fully enter the knowledge base?
- Why reread for acceptance after a successful write?
- Why does the first setup need only four directories?
- How is candidate corpus turned into WeChat Moments, short video, or official account content?
Our observations
- A content system must first ensure sources are complete and traceable, before discussing automatic decomposition and batch production.
- "API success" does not equal business completion; write location, body completeness, and read-back consistency all need acceptance.
- When chat records and personal data are involved, privacy boundaries and the final publishing decision should always remain in human hands.
From the author
If it is only collected, it is still raw material; if it becomes your viewpoint, work, proposal, or even helps you solve a specific problem, then it truly becomes your knowledge.