Files
LocalAI/backend/go/crispasr/cpp/crispasr_shim.cpp
LocalAI [bot] 76fe0bb929 feat(crispasr): add CrispASR backend — multi-architecture ASR + TTS (#10099)
* feat(crispasr): backend source files (Go gRPC server, C-ABI shim, build files)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* polish(crispasr): brand error strings + fix stale shim comment

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* build(crispasr): register backend in root Makefile

Mirror the whisper Go backend registration for the new crispasr
backend: NOTPARALLEL entry, prepare-test-extra/test-extra hooks,
BACKEND_CRISPASR definition, docker-build target generation, and the
docker-build-backends aggregate target.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* ci(crispasr): add backend build matrix entries

Mirror the 11 whisper golang Dockerfile matrix entries (CPU amd64/arm64,
CUDA 12/13, L4T CUDA 13, Intel SYCL f32/f16, Vulkan amd64/arm64, L4T
arm64, ROCm hipblas) with backend and tag-suffix substituted to crispasr.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(gallery): add crispasr backend gallery entries

Add the crispasr meta anchor and its full set of image gallery entries
(cpu, metal, cuda12/13, rocm, intel-sycl f32/f16, vulkan, L4T arm64,
L4T cuda13 arm64, plus -development variants), mirroring the whisper
backend gallery block.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* ci(crispasr): bump CRISPASR_VERSION via bump_deps workflow

Track CrispStrobe/CrispASR main branch and bump CRISPASR_VERSION in
backend/go/crispasr/Makefile.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* build(crispasr): don't wire fixture-gated test into test-extra

Mirror the whisper Go backend: its AudioTranscription test is gated on
model/audio fixtures and skips in CI, so building crispasr (the heaviest
ggml compile in the tree) inside the unit-test lane adds a long compile
for zero coverage. The backend image build in backend-matrix.yml remains
the authoritative compile check.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* ci(crispasr): add darwin metal build entry (mirror whisper)

The metal-crispasr gallery entries and capabilities.metal mapping
reference -metal-darwin-arm64-crispasr, which is only produced by an
includeDarwin entry. Mirror whisper's darwin metal entry so the tag
actually gets built.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* ci(crispasr): place hipblas matrix entry next to whisper twin

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): register crispasr as pref-only ASR backend + test

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* test(crispasr): port whisper behavioral suite (cancellation + streaming)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* test(crispasr): fix skip message env var names to CRISPASR_*

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): switch shim to crispasr_session_* multi-architecture API

The shim used whisper_full(), which in CrispASR is the whisper-only path:
libcrispasr only transcribes Whisper GGUFs through it. Multi-architecture
transcription (Parakeet, Voxtral, Qwen3-ASR, Canary, Granite, FunASR,
Paraformer, SenseVoice, ...) goes through the crispasr_session_* C-ABI,
which auto-detects the architecture from the GGUF and dispatches to the
matching backend.

Rewrite the C shim around crispasr_session_open / _transcribe_lang /
_result_* and add get_backend() so the selected backend is logged.
load_model now takes a threads param (session_open binds n_threads at
open). The session result is segment+word based with no token IDs and no
per-decode callback, so drop n_tokens / get_token_id /
get_segment_speaker_turn_next / set_new_segment_callback. set_abort is
kept for API parity but is best-effort: the session transcribe is blocking
with no abort hook.

Update the purego bindings and gocrispasr.go to match: tokens are left
empty, speaker-turn handling is removed, and AudioTranscriptionStream
emits one delta per non-empty segment after the blocking decode returns
(no progressive streaming via the session API), preserving the
concat(deltas) == final.Text invariant.

crispasr_session_set_translate is exported by libcrispasr but not declared
in crispasr.h, so it is forward-declared in the shim alongside the
open/transcribe/result functions.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* build(crispasr): link full CrispASR backend set for multi-arch support

The shim's crispasr_session_* dispatch calls into the per-architecture
backend libs (parakeet, voxtral, qwen3_asr, canary, funasr, paraformer,
sensevoice, ...), which CrispASR builds as static archives. Linking only
crispasr + ggml dead-stripped every backend object from the final module
(nm backend-symbol count: 0), leaving a whisper-only .so.

Link the same backend set as crispasr-cli so the static archives are
pulled in. After this the module carries the backend symbols (nm count
407, .so grows from ~2.1MB to ~6.7MB) and the session API can dispatch to
every compiled-in architecture.

Also rewrite ${CMAKE_SOURCE_DIR}/examples/talk-llama to
${PROJECT_SOURCE_DIR}/... in the vendored src/CMakeLists.txt: CrispASR
locates its vendored llama.cpp via ${CMAKE_SOURCE_DIR}, which is wrong when
CrispASR is add_subdirectory'd (CMAKE_SOURCE_DIR points at this backend
dir, not the CrispASR root). PROJECT_SOURCE_DIR is correct both standalone
and as a subproject; the sed is idempotent.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* test(crispasr): adapt suite to session API (blocking, no decode callback)

Register the new symbol set (drop the removed token/speaker/callback funcs,
add get_backend; load_model now takes 2 args). The session transcribe is
blocking with no abort hook, so a mid-decode cancel can't interrupt it:
change the cancellation spec to cancel the context before the call and
assert codes.Canceled from the pre-call ctx.Err() check, dropping the
<5s mid-decode timing assertion. The streaming spec still holds with
per-segment post-decode emission (>=2 deltas, concat(deltas) == final.Text).

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(gallery): add CrispASR ASR model entries (-crispasr)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* fix(gallery): keep only session-auto-detectable CrispASR ASR models

The crispasr backend loads models via crispasr_session_open, which
auto-detects the backend from the GGUF general.architecture using
crispasr_detect_backend_from_gguf. Architectures not in that detect
map cannot be opened, so those gallery entries fail to load.

Removed entries whose architecture is not wired into CrispASR
v0.6.11's session auto-detect router (they can be re-added when
upstream maps them):

- Not in the detect map: data2vec, firered-asr, funasr,
  fun-asr-mlt-nano, glm-asr, hubert, kyutai-stt, mega-asr, mimo-asr,
  moonshine{,-de,-streaming,-tiny-de}, omniasr{,-llm,-llm-1b},
  paraformer, sensevoice.
- Pending verification (filename-heuristic routed, not arch-detected):
  parakeet-ctc-0.6b, parakeet-ctc-1.1b. Their GGUFs are routed to the
  fastconformer-ctc backend by a filename heuristic in the model
  registry, which implies general.architecture is not a mapped string.

Kept the parakeet rnnt/tdt_ctc variants: convert-parakeet-to-gguf.py
writes general.architecture="parakeet" unconditionally and encodes the
rnnt/ctc distinction in metadata fields, so they session-auto-detect.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): TTS synthesis via crispasr_session_synthesize (24kHz)

Add tts_synthesize/tts_free/tts_set_voice to the C-ABI shim. They reuse
the already-open g_session (crispasr_session_open auto-detects a TTS
model) and dispatch to the upstream synthesis call, which returns
malloc'd 24 kHz mono float PCM. Orpheus needs a SNAC codec path that we
do not set, so it returns NULL here and surfaces as an error Go-side.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): implement TTS/TTSStream gRPC methods

Bind the new shim functions via purego and implement TTS, TTSStream and
a writeWAV24k helper. synthesize copies the C-owned PCM out before
freeing it; TTS writes a 24 kHz mono 16-bit WAV to req.Dst via
go-audio/wav. CrispASR has no progressive synth, so TTSStream
synthesizes fully, encodes to WAV, and emits the bytes as a single
chunk; it owns the results-channel close (the gRPC server wrapper ranges
until close), mirroring vibevoice-cpp's TTSStream.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): log when a TTS voice override is not honored

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(gallery): add CrispASR vibevoice-tts model entry

Only vibevoice-tts works through the current shim: qwen3-tts, chatterbox,
and orpheus require companion codec/s3gen/SNAC paths (set_codec_path /
set_s3gen_path) that the shim doesn't wire yet, and kokoro/indextts/voxcpm2
aren't in the session auto-detect map. Those are follow-ups.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* test(crispasr): gated TTS synthesis spec

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* fix(crispasr): satisfy golangci-lint (errcheck defers + unsafeptr nolint)

The crispasr Go file is entirely new, so new-from-merge-base lints every
line (unlike the grandfathered whisper backend it was forked from):
- handle os.RemoveAll / fh.Close return values in AudioTranscription
- annotate the two intentional C-pointer unsafe.Slice sites with //nolint:govet

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): backend: and codec: model options (explicit arch + companion files)

Add two model-config options to the CrispASR backend via opts.Options:

- backend:<name> selects an explicit CrispASR backend (bypassing
  auto-detect) by routing load_model through
  crispasr_session_open_explicit, unlocking architectures the
  detector won't pick on its own (qwen3, cohere, granite, voxtral,
  moonshine, mimo-asr, orpheus, kokoro, chatterbox, etc.).
- codec:<path> loads a companion file (qwen3-tts codec, orpheus SNAC,
  chatterbox s3gen, or mimo-asr tokenizer) via the universal
  crispasr_session_set_codec_path setter after the session opens. A
  relative path resolves against the model directory. rc==0 means
  success or not-applicable; only a negative rc is fatal.

The C shim load_model gains a backend_name argument and a new
set_codec_path entry point; the Go bridge parses the prefix:value
options and registers the new symbol. The vad_only path is unchanged.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(gallery): expand CrispASR models via backend:/codec: options (explicit arch + companions)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* refactor(gallery): use virtual.yaml base for crispasr models

The crispasr entries are just backend + model + a couple options, fully
expressed inline via overrides:/files: in gallery/index.yaml. Point each
url: at the shared gallery/virtual.yaml (the established 'virtual' model
trick) and drop the 36 redundant per-model gallery/*-crispasr.yaml files.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* fix(gallery): drop voice-requiring TTS entries (keep vibevoice-tts)

Real e2e showed qwen3-tts/orpheus/chatterbox don't synthesize through the
current shim: the codec: companion loads fine, but these engines additionally
need a voice pack / voice prompt / reference clip (qwen3-tts base errors
'no voice'; chatterbox is zero-shot cloning; orpheus uses named voices) that
the backend doesn't wire. (qwen3-tts also can't auto-detect: its GGUF arch is
'qwen3tts', unmapped by the detector — would need backend:qwen3-tts.) Removed
to avoid shipping non-working gallery entries; vibevoice-tts (built-in voice,
e2e-verified) remains the working TTS. Voice-pack wiring is a follow-up.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(crispasr): speaker: and voice: TTS options (baked speakers + voice packs/prompts)

speaker:<name> -> crispasr_session_set_speaker_name (baked speakers: qwen3-tts
CustomVoice, orpheus). voice:<path>(+voice_text:<ref>) -> crispasr_session_set_voice
(voice-pack GGUF, or WAV zero-shot clone with ref text). Applied at Load as the
default voice; req.Voice still overrides the speaker per request.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

* feat(gallery): re-add e2e-verified TTS engines (chatterbox, qwen3-tts-customvoice, orpheus)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

---------

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>
2026-05-31 12:11:03 +02:00

254 lines
8.2 KiB
C++

#include "crispasr_shim.h"
#include "ggml-backend.h"
#include "crispasr.h"
#include <atomic>
#include <vector>
// Opaque session types. crispasr.h declares `struct crispasr_session;` but not
// the result type nor the open/transcribe/result accessors — those are
// CA_EXPORT extern "C" symbols in src/crispasr_c_api.cpp, so we forward-declare
// exactly the ones we use. Signatures verified against
// sources/CrispASR/src/crispasr_c_api.cpp.
struct crispasr_session_result;
extern "C" {
crispasr_session *crispasr_session_open(const char *model_path, int n_threads);
crispasr_session *crispasr_session_open_explicit(const char *model_path,
const char *backend_name,
int n_threads);
int crispasr_session_set_codec_path(crispasr_session *s, const char *path);
void crispasr_session_close(crispasr_session *s);
const char *crispasr_session_backend(crispasr_session *s);
int crispasr_session_set_translate(crispasr_session *s, int enable);
crispasr_session_result *crispasr_session_transcribe_lang(
crispasr_session *s, const float *pcm, int n_samples, const char *language);
int crispasr_session_result_n_segments(crispasr_session_result *r);
const char *crispasr_session_result_segment_text(crispasr_session_result *r,
int i);
int64_t crispasr_session_result_segment_t0(crispasr_session_result *r, int i);
int64_t crispasr_session_result_segment_t1(crispasr_session_result *r, int i);
void crispasr_session_result_free(crispasr_session_result *r);
float *crispasr_session_synthesize(crispasr_session *s, const char *text,
int *out_n_samples);
void crispasr_pcm_free(float *pcm);
int crispasr_session_set_speaker_name(crispasr_session *s, const char *name);
int crispasr_session_set_voice(crispasr_session *s, const char *path,
const char *ref_text_or_null);
}
static crispasr_session *g_session = nullptr;
static crispasr_session_result *g_result = nullptr;
static struct whisper_vad_context *vctx;
static std::vector<float> flat_segs;
static std::atomic<int> g_abort{0};
extern "C" void set_abort(int v) {
g_abort.store(v, std::memory_order_relaxed);
}
static void ggml_log_cb(enum ggml_log_level level, const char *log,
void *data) {
const char *level_str;
if (!log) {
return;
}
switch (level) {
case GGML_LOG_LEVEL_DEBUG:
level_str = "DEBUG";
break;
case GGML_LOG_LEVEL_INFO:
level_str = "INFO";
break;
case GGML_LOG_LEVEL_WARN:
level_str = "WARN";
break;
case GGML_LOG_LEVEL_ERROR:
level_str = "ERROR";
break;
default: /* Potential future-proofing */
level_str = "?????";
break;
}
fprintf(stderr, "[%-5s] ", level_str);
fputs(log, stderr);
fflush(stderr);
}
int load_model(const char *const model_path, int threads,
const char *backend_name) {
whisper_log_set(ggml_log_cb, nullptr);
ggml_backend_load_all();
if (backend_name && *backend_name) {
g_session =
crispasr_session_open_explicit(model_path, backend_name, threads);
} else {
g_session = crispasr_session_open(model_path, threads);
}
if (g_session == nullptr) {
fprintf(stderr, "error: failed to open CrispASR session for model\n");
return 1;
}
fprintf(stderr, "info: CrispASR backend selected: %s\n",
crispasr_session_backend(g_session));
return 0;
}
// set_codec_path forwards a companion file (qwen3-tts codec, orpheus SNAC,
// chatterbox s3gen, or mimo-asr tokenizer) to the active session. Returns 0 on
// success or when the active backend needs no companion, negative on failure,
// and -1 when no session is open.
int set_codec_path(const char *path) {
return g_session ? crispasr_session_set_codec_path(g_session, path) : -1;
}
int load_model_vad(const char *const model_path) {
whisper_log_set(ggml_log_cb, nullptr);
ggml_backend_load_all();
struct whisper_vad_context_params vcparams =
whisper_vad_default_context_params();
// XXX: Overridden to false in upstream due to performance?
// vcparams.use_gpu = true;
vctx = whisper_vad_init_from_file_with_params(model_path, vcparams);
if (vctx == nullptr) {
fprintf(stderr, "error: Failed to init model as VAD\n");
return 1;
}
return 0;
}
int vad(float pcmf32[], size_t pcmf32_len, float **segs_out,
size_t *segs_out_len) {
if (!whisper_vad_detect_speech(vctx, pcmf32, pcmf32_len)) {
fprintf(stderr, "error: failed to detect speech\n");
return 1;
}
struct whisper_vad_params params = whisper_vad_default_params();
struct whisper_vad_segments *segs =
whisper_vad_segments_from_probs(vctx, params);
size_t segn = whisper_vad_segments_n_segments(segs);
// fprintf(stderr, "Got segments %zd\n", segn);
flat_segs.clear();
for (int i = 0; i < segn; i++) {
flat_segs.push_back(whisper_vad_segments_get_segment_t0(segs, i));
flat_segs.push_back(whisper_vad_segments_get_segment_t1(segs, i));
}
// fprintf(stderr, "setting out variables: %p=%p -> %p, %p=%zx -> %zx\n",
// segs_out, *segs_out, flat_segs.data(), segs_out_len, *segs_out_len,
// flat_segs.size());
*segs_out = flat_segs.data();
*segs_out_len = flat_segs.size();
// fprintf(stderr, "freeing segs\n");
whisper_vad_free_segments(segs);
// fprintf(stderr, "returning\n");
return 0;
}
// threads, diarize and prompt are accepted for Go-side API parity but unused
// in Phase 1: the thread count is fixed at session open, and diarization and
// the initial prompt are separate CrispASR features not yet wired through the
// session ASR path.
int transcribe(uint32_t threads, char *lang, bool translate, bool diarize,
float pcmf32[], size_t pcmf32_len, size_t *segs_out_len,
char *prompt) {
(void)threads;
(void)diarize;
(void)prompt;
if (!g_session) {
return 1;
}
// Reset stale abort flag from any prior cancelled call. set_abort remains
// best-effort: the session transcribe call is blocking and exposes no abort
// hook, so a mid-decode abort cannot interrupt it.
g_abort.store(0, std::memory_order_relaxed);
crispasr_session_set_translate(g_session, translate ? 1 : 0);
if (g_result) {
crispasr_session_result_free(g_result);
g_result = nullptr;
}
const char *language = (lang && *lang) ? lang : nullptr;
g_result = crispasr_session_transcribe_lang(g_session, pcmf32, (int)pcmf32_len,
language);
if (!g_result) {
fprintf(stderr, "error: transcription failed\n");
return 1;
}
*segs_out_len = crispasr_session_result_n_segments(g_result);
return 0;
}
const char *get_segment_text(int i) {
if (!g_result) {
return "";
}
return crispasr_session_result_segment_text(g_result, i);
}
int64_t get_segment_t0(int i) {
if (!g_result) {
return 0;
}
return crispasr_session_result_segment_t0(g_result, i);
}
int64_t get_segment_t1(int i) {
if (!g_result) {
return 0;
}
return crispasr_session_result_segment_t1(g_result, i);
}
const char *get_backend(void) {
return g_session ? crispasr_session_backend(g_session) : "";
}
// TTS uses the already-open session (crispasr_session_open auto-detects a TTS
// model). Output is 24 kHz mono float PCM (upstream CrispASR convention),
// malloc'd by the C API; the caller must release it via tts_free.
float *tts_synthesize(const char *text, int *out_n_samples) {
if (out_n_samples) *out_n_samples = 0;
if (!g_session || !text) return nullptr;
return crispasr_session_synthesize(g_session, text, out_n_samples);
}
void tts_free(float *pcm) {
if (pcm) crispasr_pcm_free(pcm);
}
int tts_set_voice(const char *name) {
if (!g_session || !name || !*name) return 0;
return crispasr_session_set_speaker_name(g_session, name);
}
// tts_set_voice_file loads a voice from a file: a .gguf path selects a voice
// pack, a .wav path with a non-empty ref_text performs zero-shot voice cloning
// (the C API returns -2 when ref_text is required but missing). Returns -1 when
// no session is open or path is null.
int tts_set_voice_file(const char *path, const char *ref_text) {
if (!g_session || !path) return -1;
const char *ref = (ref_text && *ref_text) ? ref_text : nullptr;
return crispasr_session_set_voice(g_session, path, ref);
}