Skip to content

[feature] The compilation pipeline would benefit from an automatic sel… #214

Description

@LanceHaverkamp

The compilation pipeline would benefit from an automatic self-review pass before finalizing pages, since the same model reliably catches errors during linting that it introduced during generation.

Real examples from a single-source wiki: the model misspelled Scofield as 'scoford' during compilation (source had it correct), and added a false claim that Origen was a preterist (source never said this). Both were caught by the same model running openkb lint — but only because a human remembered to run it.

The other day I had it hallucinate part of my bio saying: "He lives in Verona, Wisconsin, is married with two kids, and practices a leadership style of asking 'why' five times to diagnose root problems."

The only correct thing is that I'm married. The rest was completely fabricated—twice.


Diagnostics (auto-collected by openkb feedback)
  • openkb: 0.4.4
  • python: 3.13.5
  • platform: Linux 6.12.100+deb13-amd64
  • kb_initialised: yes
  • AI model: Deepseek-4-flash via opencode zen

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions