Server enhancements + adopted upstream fixes (Metal, security, KV fidelity)#1

Merged

elkaix merged 29 commits into

mainfrom

adopt-upstream-fixes

Jul 19, 2026

Commits on Jul 2, 2026

Commits on Jul 19, 2026

Reject oversized snapshot token_count before allocating

gigioneggiando
authored and
elkaix
committed
Cap file-declared lengths in agent KV cache reads

gigioneggiando
authored and
elkaix
committed
Unlink bash tool output temp file on job cleanup

gigioneggiando
authored and
elkaix
committed
Bound GGUF metadata/tensor counts by file size before allocating

gigioneggiando
authored and
elkaix
committed
Bound tokenizer merges count to INT32_MAX like the tokens table

gigioneggiando
authored and
elkaix
committed
Bound file-declared length prefixes against remaining file size

gigioneggiando
authored and
elkaix
committed
Cross-check tensor data size against shape in FP8/FP4 dequant

gigioneggiando
authored and
elkaix
committed
Resolve Metal shader sources relative to the executable

authored and
elkaix
committed
metal: keep expert pread threads on the important IO tier

jasontitus
authored and
elkaix
committed
metal: report pread pool dispatch stats in the timing summary

jasontitus
authored and
elkaix
committed
Fix double-free in JSON request parser (null *out on failure)

gigioneggiando
authored and
elkaix
committed
Handle NaN in json_int() (undefined double->int cast)

gigioneggiando
authored and
elkaix
committed
Make tool-output validation O(n) instead of O(n^2)

gigioneggiando
authored and
elkaix
committed
server: surface unclosed thinking as reasoning_content, not content

stefanwichmann
authored and
elkaix
committed
Canonicalize blank reasoning blocks

JordiPosthumus
authored and
elkaix
committed