Skip to content

Server: preserve raw prompts in /v1/completions #1012

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kyuz0 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Server: preserve raw prompts in /v1/completions #1012

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@
/tests/test_q4k_dot
/tests/test_sampling
/tests/test_quality_api
/tests/test_raw_completions
/tests/test_cuda_q8_scratch
/tests/test_cuda_dspark_moe
/tests/test_session_state
Expand Down
8 changes: 8 additions & 0 deletions Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -754,12 +754,20 @@ test-quality-api: tests/test_quality_api.c gguf-tools/quality-testing/score_offi
$(CC) $(QUALITY_CFLAGS) -I. -ffunction-sections -fdata-sections -o tests/test_quality_api tests/test_quality_api.c -Wl,$(if $(filter Darwin,$(UNAME_S)),-dead_strip,--gc-sections) -lm
./tests/test_quality_api

.PHONY: test-raw-completions
tests/test_raw_completions: tests/test_raw_completions.c ds4_server.c ds4.h ds4_kvstore.h ds4_tool_text.h $(CPU_CORE_OBJS) ds4_help.o ds4_kvstore.o rax.o
$(CC) $(CFLAGS) -Wno-unused-function -DDS4_NO_GPU -I. -ffunction-sections -fdata-sections -o $@ $< $(CPU_CORE_OBJS) ds4_help.o ds4_kvstore.o rax.o -Wl,$(if $(filter Darwin,$(UNAME_S)),-dead_strip,--gc-sections) $(LDLIBS)

test-raw-completions: tests/test_raw_completions
./tests/test_raw_completions

ds4.o ds4_cpu.o ds4_agent.o ds4_agent_cpu.o ds4_server.o ds4_server_cpu.o \
ds4_test.o ds4_agent_test.o \
ds4_cpu_test_hooks.o ds4_cuda_test_hooks.o tests/test_session_state.o \
tests/test_session_state_gpu.o: ds4_tool_text.h

clean:
rm -f tests/test_raw_completions
rm -f tests/test_metal_ssd_experts
rm -f tests/test_cuda_q8_scratch
rm -f tests/test_cuda_dspark_moe
Expand Down
2 changes: 2 additions & 0 deletions docs/SERVER.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,8 @@ Max only with sufficient context; otherwise it falls back to normal thinking.
`xhigh` maps to normal thinking, not Think Max. Use `think:false`, a disabled
thinking object, or a non-thinking model alias for direct answers.

`/v1/completions` continues a raw prompt without adding a system message or chat template. `prompt` accepts one string or one non-empty flat array of integer token IDs. Strings receive the model start sequence and recognize explicitly spelled special tokens; token arrays are used exactly as supplied. Batches, nested or mixed arrays, and IDs outside the model vocabulary are rejected. The compatibility fields `thinking`, `think`, and `reasoning_effort` are validated but do not change the prompt or suppress thinking tags in the output. Model stop tokens and explicit client stop sequences still apply.

## Multiple sessions

```sh
Expand Down
Loading