Where to go depending on what you want to change, and the three rules that are not negotiable.
| Wrong information about an organisation | Open an issue on scraper
with the link to the page that contradicts it. A correction without a source cannot be
applied. |
| An organisation wants its listing changed or removed | scraper.
The site is regenerated from the packages by site/build_site.py in that
repository, then the generated HTML is pushed to the public repo. |
| Collection or verification | scraper |
| Search or answers | rag |
| The conversation interface | chatbot |
| Startup, deployment | stack |
| This site | The public repository receives generated HTML only. Editing it directly would be overwritten by the next build. |
1. A citation cannot be bypassed. No change may allow a service to be published without a verbatim quote and a link. This is not advice, it is an automatic check: a build that finds a service without one refuses to write anything at all.
2. The category reference is frozen. It is shared with the partners who review the directory. A code added unilaterally breaks their work without their knowing. Proposals go through an issue, never a direct edit.
3. Some fields are never set by a machine. Marking a speciality is a human decision. A program that filled it in would produce a claim nobody validated, wearing the appearance of verified data.
They are not precautions in principle. Each one comes from an observed failure. A directory where nobody can check a single line is no better than a well-formatted rumour, and a rule that lives only in one person's head disappears with that person. That is why the rules above are, as far as possible, enforced by programs rather than written in a document.