chen-yun-demo

Export chat logs and social posts, then organize extracted traits into four Markdown files: procedure, interaction, memory, and personality. Tag each claim with an evidence level such as verbatim, artifact, or impression so reliability stays visible.

This demo uses WeChat private chats, Twitter posts from 2024-2025, and manually written blog articles. The manifest.json records which sources were summarized and which platforms contributed data.

Contradictions go into conflicts.md with both claims, their sources, and a provisional ruling. For example, a data-driven method versus an anti-growth value is ruled as methodology versus values, not a true conflict.

Coverage depends on source material: workplace documents may be underrepresented, memory entries rely on self-narration and can be biased, and some personality items come from friends' impressions rather than direct evidence.

This demo is explicitly a self-distillation for personal backup, not impersonation of others. Third-party content in chat records is anonymized, and the metadata carries an ethics note marking it as self-backup use only.