project-nomad/admin/constants
jakeaturner 27be71b46e feat(AI): give chat a sampler instead of the backend's defaults
Chat sent num_ctx, num_predict and nothing else, so every turn decoded at
whatever the backend picked: temperature 0.8, top_p 0.9, no min_p. On a
heavily quantized 1-3B model that is the wandering, repetitive output people
read as "the local model is bad".

Adds `ai.responseStyle` with four profiles. min_p is the lever, not
temperature: it floors relative to the top token, so it truncates hard on a
confident distribution and barely at all on a flat one, which is where small
models start to wander. top_p stays at 1.0 rather than stacked on top of it.
repeat_penalty moves with the style, because a grounded answer legitimately
repeats a part number inside the 64-token window: `focused` relaxes it to
1.05, `creative` tightens it to 1.15.

`auto` is the only style that defers to the model. Plenty of library models
ship author-tuned samplers (qwen3 sets 0.6/0.95/20), so those win key by key
and NOMAD fills only the gap, which in practice is min_p. A style the user
picked by name overrides them, or it would silently do nothing on half
their models.

The compat transport drops min_p, so a profile carries a separate compatTopP.
Passing top_p 1.0 to a backend with no min_p behind it would sample the raw
distribution, which is worse than the 0.9 it replaced.

Resolved in the controller, so the grammar-constrained task calls keep their
temperature 0. An explicit temperature still outranks a profile's, which is
what keeps the eval harness deterministic.

`eval:generation --style` is new and load-bearing: temperature 0 makes min_p,
top_p and top_k inert, so a default run cannot see this change at all.
`--style=off` sends no samplers and no temperature, and is the control.

Verified against Ollama 0.32.9, min_p accepted. 22 unit tests.
2026-09-29 15:54:41 -07:00
..
broadcast.ts feat(benchmark): live telemetry during benchmark runs (#1082) (#1084) 2026-08-04 03:37:39 +00:00
kb_collections.ts feat(eval): standardized RAG quality evaluation harness (#1233) 2026-09-29 15:54:41 -07:00
kiwix.ts fix(zim): extract Kiwix catalog URL into system-wide constant 2026-09-29 15:54:41 -07:00
kv_store.ts feat(AI): give chat a sampler instead of the backend's defaults 2026-09-29 15:54:41 -07:00
link_tile_colors.ts fix(dashboard): show the color picker swatches solid 2026-09-29 15:54:41 -07:00
link_tile_icons.ts feat(dashboard): user-added link tiles for services NOMAD does not manage 2026-09-29 15:54:41 -07:00
map_regions.ts feat(Maps): regional map downloads via go-pmtiles extract (#780) 2026-05-20 10:16:00 -07:00
misc.ts feat: curated content update checking 2026-02-11 21:49:46 -08:00
ollama.ts feat(AI): give chat a sampler instead of the backend's defaults 2026-09-29 15:54:41 -07:00
service_names.ts feat(translate): offline translation for the Information Library (#1292) 2026-09-29 15:54:41 -07:00
supply_depot_docs.ts feat(translate): offline translation for the Information Library (#1292) 2026-09-29 15:54:41 -07:00
zim_extraction.ts feat: zim content embedding 2026-02-08 13:20:10 -08:00