Super Simple Whisper Server

.gitignore

@@ -31,6 +31,7 @@ build-sanitize-thread/
 /talk-llama
 /bench
 /quantize
+/server
 /lsp
 
 arm_neon.h

Makefile

@@ -1,4 +1,4 @@
-default: main bench quantize
+default: main bench quantize server
 
 ifndef UNAME_S
 UNAME_S := $(shell uname -s)
@@ -338,7 +338,7 @@ libwhisper.so: $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) -shared -o libwhisper.so $(WHISPER_OBJ) $(LDFLAGS)
 
 clean:
-	rm -f *.o main stream command talk talk-llama bench quantize lsp libwhisper.a libwhisper.so
+	rm -f *.o main stream command talk talk-llama bench quantize server lsp libwhisper.a libwhisper.so
 
 #
 # Examples
@@ -359,6 +359,9 @@ bench: examples/bench/bench.cpp $(WHISPER_OBJ)
 quantize: examples/quantize/quantize.cpp $(WHISPER_OBJ) $(SRC_COMMON)
 	$(CXX) $(CXXFLAGS) examples/quantize/quantize.cpp $(SRC_COMMON) $(WHISPER_OBJ) -o quantize $(LDFLAGS)
 
+server: examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ)
+	$(CXX) $(CXXFLAGS) examples/server/server.cpp $(SRC_COMMON) $(WHISPER_OBJ) -o server $(LDFLAGS)
+
 stream: examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ)
 	$(CXX) $(CXXFLAGS) examples/stream/stream.cpp $(SRC_COMMON) $(SRC_COMMON_SDL) $(WHISPER_OBJ) -o stream $(CC_SDL) $(LDFLAGS)
 

examples/CMakeLists.txt

@@ -65,6 +65,7 @@ elseif(CMAKE_JS_VERSION)
 else()
     add_subdirectory(main)
     add_subdirectory(stream)
+    add_subdirectory(server)
     add_subdirectory(command)
     add_subdirectory(bench)
     add_subdirectory(quantize)

examples/main/main.cpp

@@ -165,8 +165,8 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
         else if (arg == "-m"    || arg == "--model")           { params.model           = argv[++i]; }
         else if (arg == "-f"    || arg == "--file")            { params.fname_inp.emplace_back(argv[++i]); }
         else if (arg == "-oved" || arg == "--ov-e-device")     { params.openvino_encode_device = argv[++i]; }
-        else if (arg == "-ls"   || arg == "--log-score")       { params.log_score = true; }
-        else if (arg == "-ng"   || arg == "--no-gpu")          { params.use_gpu = false; }
+        else if (arg == "-ls"   || arg == "--log-score")       { params.log_score       = true; }
+        else if (arg == "-ng"   || arg == "--no-gpu")          { params.use_gpu         = false; }
         else {
             fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
             whisper_print_usage(argc, argv, params);

examples/server/CMakeLists.txt

@@ -0,0 +1,6 @@
+set(TARGET server)
+add_executable(${TARGET} server.cpp httplib.h json.hpp)
+
+include(DefaultTargetOptions)
+
+target_link_libraries(${TARGET} PRIVATE common whisper ${CMAKE_THREAD_LIBS_INIT})

examples/server/README.md

@@ -0,0 +1,59 @@
+# whisper.cpp http server
+
+Simple http server. WAV Files are passed to the inference model via http requests.
+
+```
+./server -h
+
+usage: ./bin/server [options]
+
+options:
+  -h,        --help              [default] show this help message and exit
+  -t N,      --threads N         [4      ] number of threads to use during computation
+  -p N,      --processors N      [1      ] number of processors to use during computation
+  -ot N,     --offset-t N        [0      ] time offset in milliseconds
+  -on N,     --offset-n N        [0      ] segment index offset
+  -d  N,     --duration N        [0      ] duration of audio to process in milliseconds
+  -mc N,     --max-context N     [-1     ] maximum number of text context tokens to store
+  -ml N,     --max-len N         [0      ] maximum segment length in characters
+  -sow,      --split-on-word     [false  ] split on word rather than on token
+  -bo N,     --best-of N         [2      ] number of best candidates to keep
+  -bs N,     --beam-size N       [-1     ] beam size for beam search
+  -wt N,     --word-thold N      [0.01   ] word timestamp probability threshold
+  -et N,     --entropy-thold N   [2.40   ] entropy threshold for decoder fail
+  -lpt N,    --logprob-thold N   [-1.00  ] log probability threshold for decoder fail
+  -debug,    --debug-mode        [false  ] enable debug mode (eg. dump log_mel)
+  -tr,       --translate         [false  ] translate from source language to english
+  -di,       --diarize           [false  ] stereo audio diarization
+  -tdrz,     --tinydiarize       [false  ] enable tinydiarize (requires a tdrz model)
+  -nf,       --no-fallback       [false  ] do not use temperature fallback while decoding
+  -ps,       --print-special     [false  ] print special tokens
+  -pc,       --print-colors      [false  ] print colors
+  -pp,       --print-progress    [false  ] print progress
+  -nt,       --no-timestamps     [false  ] do not print timestamps
+  -l LANG,   --language LANG     [en     ] spoken language ('auto' for auto-detect)
+  -dl,       --detect-language   [false  ] exit after automatically detecting language
+             --prompt PROMPT     [       ] initial prompt
+  -m FNAME,  --model FNAME       [models/ggml-base.en.bin] model path
+  -oved D,   --ov-e-device DNAME [CPU    ] the OpenVINO device used for encode inference
+  --host HOST,                   [127.0.0.1] Hostname/ip-adress for the server
+  --port PORT,                   [8080   ] Port number for the server
+```
+
+## request examples
+
+**/inference**
+```
+curl 127.0.0.1:8080/inference \
+-H "Content-Type: multipart/form-data" \
+-F file="@<file-path>" \
+-F temperature="0.2" \
+-F response-format="json"
+```
+
+**/load**
+```
+curl 127.0.0.1:8080/load \
+-H "Content-Type: multipart/form-data" \
+-F model="<path-to-model-file>"
+```