Brain demo community-brain

checking…

As it is asked in a survey. The brain first distils it into a search query with one LLM call; the result shows the query it used.

Each label counts 0 (Laag), 1 (Midden) or 2 (Hoog). TV, radio and the paper count once; online news counts once, as the higher of paid and free. The score sets the tier; by default 0–1 is negligible, 2–3 low, 4–5 medium and 6–8 high.

Settings

How many of the most relevant prepared articles the search fetches. A tier keeps the first few of them.

An article whose semantic similarity is below this is not recalled, unless it contains a name from the question. 0 switches the floor off, which shows every candidate's score.

On, the two rankings are fused and an article whose detailed text contains a name from the question is recalled even below the floor.

The instructions of the call that turns the question into a search query. Sent with this request only.

Per tier: the lowest score that falls in it, how many memories a community keeps, and the text it reads. The values shown are the defaults; they are sent with this request only. A score below the low tier's is negligible and keeps nothing.