From 6600381bb1af974afed5773379a8d48b9ad0bcc1 Mon Sep 17 00:00:00 2001
From: sanj <67624670+iodrift@users.noreply.github.com>
Date: Thu, 27 Jun 2024 19:27:06 -0700
Subject: [PATCH] Auto-update: Thu Jun 27 19:27:06 PDT 2024
---
sijapi/routers/asr.py | 28 +++++++++++++++++-------
sijapi/routers/llm.py | 12 +++++------
sijapi/routers/note.py | 48 ++++++++++++++++++++---------------------
sijapi/routers/serve.py | 8 +++----
sijapi/routers/tts.py | 12 +++++------
5 files changed, 59 insertions(+), 49 deletions(-)
diff --git a/sijapi/routers/asr.py b/sijapi/routers/asr.py
index 8b160b8..f518164 100644
--- a/sijapi/routers/asr.py
+++ b/sijapi/routers/asr.py
@@ -41,7 +41,7 @@ transcription_results = {}
@asr.post("/transcribe")
@asr.post("/v1/audio/transcription")
async def transcribe_endpoint(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
file: UploadFile = File(...),
params: str = Form(...)
):
@@ -58,7 +58,7 @@ async def transcribe_endpoint(
temp_file.write(await file.read())
temp_file_path = temp_file.name
- transcription_job = await transcribe_audio(file_path=temp_file_path, params=parameters, background_tasks=background_tasks)
+ transcription_job = await transcribe_audio(file_path=temp_file_path, params=parameters, bg_tasks=bg_tasks)
job_id = transcription_job["job_id"]
# Poll for completion
@@ -80,12 +80,13 @@ async def transcribe_endpoint(
# If we've reached this point, the transcription has taken too long
return JSONResponse(content={"status": "timeout", "message": "Transcription is taking longer than expected. Please check back later."}, status_code=202)
-async def transcribe_audio(file_path, params: TranscribeParams, background_tasks: BackgroundTasks):
+async def transcribe_audio(file_path, params: TranscribeParams, bg_tasks: BackgroundTasks):
+ L.DEBUG(f"Transcribing audio file from {file_path}...")
file_path = await convert_to_wav(file_path)
- model = params.model if params.model in WHISPER_CPP_MODELS else 'small'
+ model = params.model if params.model in WHISPER_CPP_MODELS else 'small'
model_path = WHISPER_CPP_DIR / 'models' / f'ggml-{model}.bin'
command = [str(WHISPER_CPP_DIR / 'build' / 'bin' / 'main')]
- command.extend(['-m', str(model_path)])
+ command.extend(['-m', str(model_path)])
command.extend(['-t', str(max(1, min(params.threads or MAX_CPU_CORES, MAX_CPU_CORES)))])
command.extend(['-np']) # Always enable no-prints
@@ -117,7 +118,6 @@ async def transcribe_audio(file_path, params: TranscribeParams, background_tasks
command.extend(['--dtw', params.dtw])
command.extend(['-f', file_path])
-
L.DEBUG(f"Command: {command}")
# Create a unique ID for this transcription job
@@ -127,9 +127,21 @@ async def transcribe_audio(file_path, params: TranscribeParams, background_tasks
transcription_results[job_id] = {"status": "processing", "result": None}
# Run the transcription in a background task
- background_tasks.add_task(process_transcription, command, file_path, job_id)
+ bg_tasks.add_task(process_transcription, command, file_path, job_id)
- return {"job_id": job_id}
+ max_wait_time = 300 # 5 minutes
+ poll_interval = 1 # 1 second
+ start_time = asyncio.get_event_loop().time()
+
+ while asyncio.get_event_loop().time() - start_time < max_wait_time:
+ job_status = transcription_results.get(job_id, {})
+ if job_status["status"] == "completed":
+ return job_status["result"]
+ elif job_status["status"] == "failed":
+ raise Exception(f"Transcription failed: {job_status.get('error', 'Unknown error')}")
+ await asyncio.sleep(poll_interval)
+
+ raise TimeoutError("Transcription timed out")
async def process_transcription(command, file_path, job_id):
try:
diff --git a/sijapi/routers/llm.py b/sijapi/routers/llm.py
index 3433140..ce80ecf 100644
--- a/sijapi/routers/llm.py
+++ b/sijapi/routers/llm.py
@@ -522,7 +522,7 @@ async def summarize_post(file: Optional[UploadFile] = File(None), text: Optional
return summarized_text
@llm.post("/speaksummary")
-async def summarize_tts_endpoint(background_tasks: BackgroundTasks, instruction: str = Form(SUMMARY_INSTRUCT), file: Optional[UploadFile] = File(None), text: Optional[str] = Form(None), voice: Optional[str] = Form(DEFAULT_VOICE), speed: Optional[float] = Form(1.2), podcast: Union[bool, str] = Form(False)):
+async def summarize_tts_endpoint(bg_tasks: BackgroundTasks, instruction: str = Form(SUMMARY_INSTRUCT), file: Optional[UploadFile] = File(None), text: Optional[str] = Form(None), voice: Optional[str] = Form(DEFAULT_VOICE), speed: Optional[float] = Form(1.2), podcast: Union[bool, str] = Form(False)):
podcast = str_to_bool(str(podcast)) # Proper boolean conversion
text_content = text if text else extract_text(file)
@@ -546,8 +546,8 @@ async def summarize_tts(
timestamp = dt_datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"{timestamp}{filename}.wav"
- background_tasks = BackgroundTasks()
- final_output_path = await generate_speech(background_tasks, summarized_text, voice, "xtts", speed=speed, podcast=podcast, title=filename)
+ bg_tasks = BackgroundTasks()
+ final_output_path = await generate_speech(bg_tasks, summarized_text, voice, "xtts", speed=speed, podcast=podcast, title=filename)
L.DEBUG(f"summary_tts completed with final_output_path: {final_output_path}")
return final_output_path
@@ -578,7 +578,7 @@ def calculate_max_tokens(text: str) -> int:
return min(tokens_count // 4, SUMMARY_CHUNK_SIZE)
-async def extract_text(file: Union[UploadFile, bytes, bytearray, str, Path], background_tasks: BackgroundTasks = None) -> str:
+async def extract_text(file: Union[UploadFile, bytes, bytearray, str, Path], bg_tasks: BackgroundTasks = None) -> str:
if isinstance(file, UploadFile):
file_extension = get_extension(file)
temp_file_path = tempfile.mktemp(suffix=file_extension)
@@ -614,8 +614,8 @@ async def extract_text(file: Union[UploadFile, bytes, bytearray, str, Path], bac
elif file_ext == '.docx':
text_content = await extract_text_from_docx(file_path)
- if background_tasks and 'temp_file_path' in locals():
- background_tasks.add_task(os.remove, temp_file_path)
+ if bg_tasks and 'temp_file_path' in locals():
+ bg_tasks.add_task(os.remove, temp_file_path)
elif 'temp_file_path' in locals():
os.remove(temp_file_path)
diff --git a/sijapi/routers/note.py b/sijapi/routers/note.py
index b079db5..005ad64 100644
--- a/sijapi/routers/note.py
+++ b/sijapi/routers/note.py
@@ -133,7 +133,7 @@ async def build_daily_timeslips(date):
### CLIPPER ###
@note.post("/clip")
async def clip_post(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
file: UploadFile = None,
url: Optional[str] = Form(None),
source: Optional[str] = Form(None),
@@ -142,65 +142,64 @@ async def clip_post(
voice: str = Form(DEFAULT_VOICE),
encoding: str = Form('utf-8')
):
- markdown_filename = await process_article(background_tasks, url, title, encoding, source, tts, voice)
+ markdown_filename = await process_article(bg_tasks, url, title, encoding, source, tts, voice)
return {"message": "Clip saved successfully", "markdown_filename": markdown_filename}
@note.post("/archive")
async def archive_post(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
file: UploadFile = None,
url: Optional[str] = Form(None),
source: Optional[str] = Form(None),
title: Optional[str] = Form(None),
encoding: str = Form('utf-8')
):
- markdown_filename = await process_archive(background_tasks, url, title, encoding, source)
+ markdown_filename = await process_archive(bg_tasks, url, title, encoding, source)
return {"message": "Clip saved successfully", "markdown_filename": markdown_filename}
@note.get("/clip")
async def clip_get(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
url: str,
title: Optional[str] = Query(None),
encoding: str = Query('utf-8'),
tts: str = Query('summary'),
voice: str = Query(DEFAULT_VOICE)
):
- markdown_filename = await process_article(background_tasks, url, title, encoding, tts=tts, voice=voice)
+ markdown_filename = await process_article(bg_tasks, url, title, encoding, tts=tts, voice=voice)
return {"message": "Clip saved successfully", "markdown_filename": markdown_filename}
@note.post("/note/add")
-async def note_add_endpoint(file: Optional[UploadFile] = File(None), text: Optional[str] = Form(None), source: Optional[str] = Form(None)):
+async def note_add_endpoint(file: Optional[UploadFile] = File(None), text: Optional[str] = Form(None), source: Optional[str] = Form(None), bg_tasks: BackgroundTasks = None):
+ L.DEBUG(f"Received request on /note/add...")
if not file and not text:
+ L.WARN(f"... without any file or text!")
raise HTTPException(status_code=400, detail="Either text or a file must be provided")
else:
- result = await process_for_daily_note(file, text, source)
+ result = await process_for_daily_note(file, text, source, bg_tasks)
L.INFO(f"Result on /note/add: {result}")
return JSONResponse(result, status_code=204)
-async def process_for_daily_note(file: Optional[UploadFile] = File(None), text: Optional[str] = None, source: Optional[str] = None):
+async def process_for_daily_note(file: Optional[UploadFile] = File(None), text: Optional[str] = None, source: Optional[str] = None, bg_tasks: BackgroundTasks = None):
now = datetime.now()
-
transcription_entry = ""
file_entry = ""
if file:
+ L.DEBUG("File received...")
file_content = await file.read()
audio_io = BytesIO(file_content)
file_type, _ = mimetypes.guess_type(file.filename)
-
- if 'audio' in file_type:
- subdir = "Audio"
- elif 'image' in file_type:
- subdir = "Images"
- else:
- subdir = "Documents"
-
+ L.DEBUG(f"Processing as {file_type}...")
+ subdir = file_type.title() or "Documents"
absolute_path, relative_path = assemble_journal_path(now, subdir=subdir, filename=file.filename)
+ L.DEBUG(f"Destination path: {absolute_path}")
+
with open(absolute_path, 'wb') as f:
f.write(file_content)
+ L.DEBUG(f"Processing {f.name}...")
if 'audio' in file_type:
- transcription = await asr.transcribe_audio(file_path=absolute_path, params=asr.TranscribeParams(model="small-en", language="en", threads=6))
+ transcription = await asr.transcribe_audio(file_path=absolute_path, params=asr.TranscribeParams(model="small-en", language="en", threads=6), bg_tasks=bg_tasks)
file_entry = f"![[{relative_path}]]"
elif 'image' in file_type:
@@ -209,7 +208,6 @@ async def process_for_daily_note(file: Optional[UploadFile] = File(None), text:
else:
file_entry = f"[Source]({relative_path})"
-
text_entry = text if text else ""
L.DEBUG(f"transcription: {transcription}\nfile_entry: {file_entry}\ntext_entry: {text_entry}")
return await add_to_daily_note(transcription, file_entry, text_entry, now)
@@ -262,7 +260,7 @@ async def handle_text(title:str, summary:str, extracted_text:str, date_time: dat
async def process_document(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
document: File,
title: Optional[str] = None,
tts_mode: str = "summary",
@@ -301,7 +299,7 @@ added: {timestamp}
datetime_str = datetime.now().strftime("%Y%m%d%H%M%S")
audio_filename = f"{datetime_str} {readable_title}"
audio_path = await tts.generate_speech(
- background_tasks=background_tasks,
+ bg_tasks=bg_tasks,
text=tts_text,
voice=voice,
model="eleven_turbo_v2",
@@ -336,7 +334,7 @@ added: {timestamp}
async def process_article(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
url: str,
title: Optional[str] = None,
encoding: str = 'utf-8',
@@ -397,7 +395,7 @@ tags:
datetime_str = datetime.now().strftime("%Y%m%d%H%M%S")
audio_filename = f"{datetime_str} {readable_title}"
try:
- audio_path = await tts.generate_speech(background_tasks=background_tasks, text=tts_text, voice=voice, model="eleven_turbo_v2", podcast=True, title=audio_filename,
+ audio_path = await tts.generate_speech(bg_tasks=bg_tasks, text=tts_text, voice=voice, model="eleven_turbo_v2", podcast=True, title=audio_filename,
output_dir=Path(OBSIDIAN_VAULT_DIR) / OBSIDIAN_RESOURCES_DIR)
audio_ext = Path(audio_path).suffix
obsidian_link = f"![[{OBSIDIAN_RESOURCES_DIR}/{audio_filename}{audio_ext}]]"
@@ -502,7 +500,7 @@ async def html_to_markdown(url: str = None, source: str = None) -> Optional[str]
async def process_archive(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
url: str,
title: Optional[str] = None,
encoding: str = 'utf-8',
diff --git a/sijapi/routers/serve.py b/sijapi/routers/serve.py
index 5452c4a..eea8725 100644
--- a/sijapi/routers/serve.py
+++ b/sijapi/routers/serve.py
@@ -136,7 +136,7 @@ async def hook_changedetection(webhook_data: dict):
@serve.post("/cl/search")
-async def hook_cl_search(request: Request, background_tasks: BackgroundTasks):
+async def hook_cl_search(request: Request, bg_tasks: BackgroundTasks):
client_ip = request.client.host
L.DEBUG(f"Received request from IP: {client_ip}")
data = await request.json()
@@ -150,7 +150,7 @@ async def hook_cl_search(request: Request, background_tasks: BackgroundTasks):
json.dump(payload, file, indent=2)
for result in results:
- background_tasks.add_task(cl_search_process_result, result)
+ bg_tasks.add_task(cl_search_process_result, result)
return JSONResponse(content={"message": "Received"}, status_code=status.HTTP_200_OK)
@serve.post("/cl/docket")
@@ -283,7 +283,7 @@ def shellfish_run_widget_command(args: List[str]):
### COURTLISTENER FUNCTIONS ###
-async def cl_docket(data, client_ip, background_tasks: BackgroundTasks):
+async def cl_docket(data, client_ip, bg_tasks: BackgroundTasks):
payload = data['payload']
results = data['payload']['results']
timestamp = datetime.now().strftime("%Y%m%d-%H%M%S")
@@ -292,7 +292,7 @@ async def cl_docket(data, client_ip, background_tasks: BackgroundTasks):
json.dump(payload, file, indent=2)
for result in results:
- background_tasks.add_task(cl_docket_process, result)
+ bg_tasks.add_task(cl_docket_process, result)
return JSONResponse(content={"message": "Received"}, status_code=status.HTTP_200_OK)
async def cl_docket_process(result):
diff --git a/sijapi/routers/tts.py b/sijapi/routers/tts.py
index 34cfbd0..bb76645 100644
--- a/sijapi/routers/tts.py
+++ b/sijapi/routers/tts.py
@@ -87,7 +87,7 @@ def select_voice(voice_name: str) -> str:
@tts.post("/v1/audio/speech")
async def generate_speech_endpoint(
request: Request,
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
model: str = Form("eleven_turbo_v2"),
text: Optional[str] = Form(None),
file: Optional[UploadFile] = File(None),
@@ -110,7 +110,7 @@ async def generate_speech_endpoint(
else:
return await stream_tts(text_content, speed, voice, voice_file)
else:
- return await generate_speech(background_tasks, text_content, voice, voice_file, model, speed, podcast)
+ return await generate_speech(bg_tasks, text_content, voice, voice_file, model, speed, podcast)
except Exception as e:
L.ERR(f"Error in TTS: {str(e)}")
L.ERR(traceback.format_exc())
@@ -118,7 +118,7 @@ async def generate_speech_endpoint(
async def generate_speech(
- background_tasks: BackgroundTasks,
+ bg_tasks: BackgroundTasks,
text: str,
voice: str = None,
voice_file: UploadFile = None,
@@ -142,8 +142,8 @@ async def generate_speech(
elif model == "xtts":
L.INFO(f"Using XTTS2")
- final_output_dir = await local_tts(text, speed, voice, voice_file, podcast, background_tasks, title, output_dir)
- background_tasks.add_task(os.remove, str(final_output_dir))
+ final_output_dir = await local_tts(text, speed, voice, voice_file, podcast, bg_tasks, title, output_dir)
+ bg_tasks.add_task(os.remove, str(final_output_dir))
return str(final_output_dir)
else:
raise HTTPException(status_code=400, detail="Invalid model specified")
@@ -282,7 +282,7 @@ async def local_tts(
voice: str,
voice_file = None,
podcast: bool = False,
- background_tasks: BackgroundTasks = None,
+ bg_tasks: BackgroundTasks = None,
title: str = None,
output_path: Optional[Path] = None
) -> str: