chat not using stocks query

[Clare72]

[Clare72]

I think count discrepancy (7 v 9) should be fixed since VirtualFlyBrain/VFBquery#92

[Robbie1977]

VFB does hold the stocks, one hop further than the planner looked: each SplitsTargeting row names its two hemidriver constructs (the FBtp ids in the row id), FindStocks accepts an FBtp, and the stock present in both constructs' results is the split-GAL4 stock — BDSC 68318 for R21B06-DBD ∩ R13F02-AD. A new vfb_find_split_stocks macro runs that chain (one batched run_query) and the answer lists each split with its stock centre and number; 7 of the 9 KCg-d splits have one. The 7 vs 9 is two different queries on the same term (TransgeneExpressionHere vs SplitsTargeting) — "split drivers" wasn't recognised as split-GAL4 wording, so the first turn ran the generic driver query; fixed too. On fix/answer-quality-and-display (PR #51).

[Robbie1977]

Shipped in v4.2.11 (PR #51).

Cause. The stocks guidance pointed the planner at vfb_find_stocks, which takes a FlyBase feature id — nothing to call on an anatomy term — so it ran the expression query instead and said VFB does not list stock numbers. "split drivers" was also not recognised as split-GAL4 wording, which is the 7 vs 9: the first turn ran TransgeneExpressionHere (7 reports), the follow-up SplitsTargeting (9 splits) — two different queries on the same term, not a count bug.

New behaviour. A new vfb_find_split_stocks macro runs SplitsTargeting → FindStocks on each split's two hemidriver constructs (batched into one call) → the stock carrying both. The repro now lists each split with its stock centre and number (BDSC 68318 for R21B06-DBD ∩ R13F02-AD; 7 of the 9 KCg-d splits have one) as prose and as a table linking the FlyBase stock reports; a split with no combined stock is reported as such. "split drivers" counts as split-GAL4 wording. Covered by W8 turns 1–2 in tests/task-battery/workshop.json.