master
jinja : support ensure_ascii=true, string repetition and int/float self-filtering (#21623)
fit-params : refactor + add option to output estimated memory per device (#22171)
vendor : update cpp-httplib to 0.30.0 (#18660)
llava : expose as a shared library for downstream projects (#3613)
libs : rename libcommon -> libllama-common (#21936)
common/autoparser : allow space after tool call (#22073)
common : replace wrap_for_generation with a prefix convenience function and fix gpt-oss (#20912)
chat : avoid including json in chat.h (#21306)
autoparser: support case of JSON_NATIVE with per-call markers (test case: Reka-Edge) (#21892)
common : add gemma 4 specialized parser (#21418)
server : speculative checkpointing (#19493)
cli: fix stripping of \n in multiline input (#21485)
cli : add command and file auto-completion (#19985)
debug: make common_debug_print_tensor readable (#19331)
chore : correct typos [no ci] (#20041)
common : add download cancellation and temp file cleanup (#21813)
server : refactor "use checkpoint" logic (#22114)
common : fix split model migration (#21019)
server: Parse port numbers from MCP server URLs in CORS proxy (#20208)
common : Generalized XML-style tool-call parsing with streaming support (GLM 4.5/4.6 + MiniMax M2 + SeedOSS + Kimi-K2 + Qwen3-Coder + Apriel-1.5 + Xiaomi-MiMo) (#16932)
cli : fix reasoning responses in CLI (#18961)
common/json-schema: fix: handle non-capturing groups (?:...) in JSON schema pattern converter (#21124)
common : add nemotron 3 parsing (#18077)
sampling : add support for backend sampling (#17004)
spec : add self‑speculative decoding (no draft model required) + refactor (#18471)
fix: correct misspellings in code comments (#21217)
spec : add ngram-mod (#19164)
ngram-mod : fix build [no ci] (#19216)
common/gemma4 : handle parsing edge cases (#21760)
preset: allow named remote preset (#18728)
common: support remote preset (#18520)
common : inhibit lazy grammar sampler while reasoning is active (#20970)
common : fix iterator::end() dereference (#20445)
`common`: add partial regex support (#12808)
common/parser: handle reasoning budget (#20297)