diff options
| author | Onur Gündüz <ogunduz@gmail.com> | 2026-07-04 02:37:58 +0200 |
|---|---|---|
| committer | Onur Gündüz <ogunduz@gmail.com> | 2026-07-04 02:37:58 +0200 |
| commit | ed23a1571fe6911eba9d4727bef30c82df7be06d (patch) | |
| tree | dd9efa8053620f3b9ee662375837671c9880adc8 /tools/server/server.h | |
| parent | c2b369f634d1e3087465a65b1a956aea872a156c (diff) | |
| download | ffwd-ed23a1571fe6911eba9d4727bef30c82df7be06d.tar.gz ffwd-ed23a1571fe6911eba9d4727bef30c82df7be06d.zip | |
server: per-model input token cap, auto-truncate, and 429 admission
Diffstat (limited to 'tools/server/server.h')
| -rw-r--r-- | tools/server/server.h | 4 |
1 files changed, 3 insertions, 1 deletions
diff --git a/tools/server/server.h b/tools/server/server.h index 97fb57f..151a793 100644 --- a/tools/server/server.h +++ b/tools/server/server.h @@ -39,7 +39,9 @@ typedef struct { int port; int batch_size; int max_batch_tokens; - int max_request_tokens; /* per-request total-token cap; 0 = default 120000 */ + int max_request_tokens; /* per-request total-token cap; 0 = default 120000 */ + int max_concurrent_requests; /* in-flight request cap before 429; 0 = default 512 */ + int auto_truncate; /* truncate over-long inputs instead of rejecting them */ int batch_wait_us; /* GPU build tuning; ignored by backends that do not use it. */ int gpu_quantize_bits; |
