electron-rare/meet
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

🔧(summary) fix featureflag summary and adding variables in settings

Browse Source 
fixing the checking of the summary-enabled featureflag
cleaning the prints
adding queues name in the settings
renaming the transcribe worker to original name
removing useless variable declarations
This commit is contained in:
Martin Guitteny
2025-09-08 11:16:28 +02:00
parent 7b0c5ee300
commit 486c737156
6 changed files with 28 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/summary/compose.yaml
+1 -1
View File
@@ -16,7 +16,7 @@ services:
".env"
depends_on:
- redis
celery_transcribe_worker:
celery_worker:
container_name: celery_worker_transcribe
build: .
command: celery -A summary.core.celery_worker worker --pool=solo --loglevel=debug -Q transcribe_queue
src/summary/summary/api/route/tasks.py
+3 -2
View File
@@ -6,12 +6,13 @@ from typing import Optional
from celery.result import AsyncResult
from fastapi import APIRouter
from pydantic import BaseModel
from summary.core.config import get_settings
from summary.core.celery_worker import (
process_audio_transcribe_summarize,
process_audio_transcribe_summarize_v2,
)
settings = get_settings()
class TaskCreation(BaseModel):
"""Task data."""
@@ -46,7 +47,7 @@ async def create_task(request: TaskCreation):
request.recording_date,
request.recording_time,
],
queue="transcribe_queue",
queue=settings.transcribe_queue,
)
return {"id": task.id, "message": "Task created"}
src/summary/summary/core/analytics.py
+6 -16
View File
@@ -48,32 +48,22 @@ class Analytics:
except Exception as e:
raise AnalyticsException("Failed to capture analytics event") from e
def feature_enabled(
self, feature_name: str, distinct_id: str = None, email: str = None
def is_feature_enabled(
self, feature_name: str, distinct_id: str = None
) -> bool:
"""
Vérifie si un feature flag est activé dans PostHog.
distinct_id optionnel pour les flags ciblés.
Check if a feature flag is enabled in PostHog for a dinstinct_id.
"""
if self.is_disabled:
return False
if email:
self.capture(
"user_identified",
distinct_id=distinct_id,
properties={"$set": {"email": email}},
)
logger.info(f"[Analytics] Identify user {distinct_id} with email {email}")
try:
logger.info(
f"[Analytics] Check feature flag {feature_name} for user {distinct_id}"
f"Check feature flag {feature_name} for user {distinct_id}"
)
return self._client.is_feature_enabled(feature_name, distinct_id)
return self._client.feature_enabled(feature_name, distinct_id)
except Exception as e:
logger.warning(
f"[Analytics] Impossible de vérifier le flag {feature_name}: {e}"
)
logger.error(f"Error checking feature flag {feature_name}: {e}")
return False
src/summary/summary/core/celery_summarize_worker.py
+10 -22
View File
@@ -50,7 +50,7 @@ def post_with_retries(url, data):
def LLM_call(client, system_prompt, user_prompt, retry=2):
data = {
"model": "Qwen/Qwen2.5-Coder-32B-Instruct-AWQ",
"model": settings.resume_llm_model,
"messages": [
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
@@ -59,38 +59,29 @@ def LLM_call(client, system_prompt, user_prompt, retry=2):
try:
response = client.chat.completions.create(**data)
print(response)
return response.choices[0].message.content
except Exception as e:
if retry > 0:
print(f"Erreur lors de l'appel à l'API : {e}. Nouvelle tentative...")
return LLM_call(system_prompt, user_prompt, retry=retry - 1)
else:
print(f"Erreur lors de l'appel à l'API : {e}")
return False
logger.error("LLM call failed: %s", e)
return False
@celery.task(
bind=True, autoretry_for=[Exception], max_retries=3, queue="summarize_queue"
bind=True, autoretry_for=[Exception], max_retries=3, queue=settings.summarize_queue
)
def summarize_transcription(self, transcript: str, email: str, sub: str, title: str):
logger.info("Starting summarization task")
prompt_system_TLDR = PROMPT_SYSTEM_TLDR
prompt_system_plan = PROMPT_SYSTEM_PLAN
prompt_system_part = PROMPT_SYSTEM_PART
logger.info("Initiating summarize client")
client_summary = openai.OpenAI(
base_url=settings.resume_endpoint, api_key=settings.resume_api_key
)
out = LLM_call(client_summary, prompt_system_TLDR, transcript)
tldr = LLM_call(client_summary, PROMPT_SYSTEM_TLDR, transcript)
logger.info("TLDR generated")
parts = LLM_call(client_summary, prompt_system_plan, transcript)
parts = LLM_call(client_summary, PROMPT_SYSTEM_PLAN, transcript)
logger.info("Plan generated")
parts = parts.split("\n")
@@ -103,7 +94,7 @@ def summarize_transcription(self, transcript: str, email: str, sub: str, title:
logger.info("Summarizing part: %s", part)
parts_summarized.append(
LLM_call(
client_summary, prompt_system_part, prompt_user_part.format(part=part)
client_summary, PROMPT_SYSTEM_PART, prompt_user_part.format(part=part)
)
)
@@ -111,14 +102,11 @@ def summarize_transcription(self, transcript: str, email: str, sub: str, title:
raw_summary = "\n\n".join(parts_summarized)
prompt_system_next_steps = PROMPT_SYSTEM_NEXT_STEP
prompt_system_cleaning = PROMPT_SYSTEM_CLEANING
next_steps = LLM_call(client_summary, prompt_system_next_steps, transcript)
next_steps = LLM_call(client_summary, PROMPT_SYSTEM_NEXT_STEP, transcript)
logger.info("Next steps generated")
cleaned_summary = LLM_call(client_summary, prompt_system_cleaning, raw_summary)
cleaned_summary = LLM_call(client_summary, PROMPT_SYSTEM_CLEANING, raw_summary)
logger.info("Summary cleaned")
summary = out + "\n\n" + cleaned_summary + "\n\n" + next_steps
summary = tldr + "\n\n" + cleaned_summary + "\n\n" + next_steps
data = {
"title": title + " - Summary",
src/summary/summary/core/celery_worker.py
+4 -6
View File
@@ -175,7 +175,7 @@ def process_audio_transcribe_summarize(filename: str, email: str, sub: str):
instructions = get_instructions(transcription)
summary_response = openai_client.chat.completions.create(
model=settings.openai_llm_model, messages=instructions
model=settings.resume_llm_model, messages=instructions
)
summary = summary_response.choices[0].message.content
@@ -202,7 +202,7 @@ def process_audio_transcribe_summarize(filename: str, email: str, sub: str):
bind=True,
autoretry_for=[exceptions.HTTPError],
max_retries=settings.celery_max_retries,
queue="transcribe_queue",
queue=settings.transcribe_queue,
)
def process_audio_transcribe_summarize_v2(
self,
@@ -320,12 +320,10 @@ def process_audio_transcribe_summarize_v2(
logger.debug("Response body: %s", response.text)
metadata_manager.capture(task_id, settings.posthog_event_success)
if not analytics.feature_enabled("summary-enabled", distinct_id=sub, email=email):
print("Summary generation skipped (feature flag disabled).")
if not analytics.is_feature_enabled("summary-enabled", distinct_id=sub):
logger.info("Summary generation skipped (feature flag disabled).")
else:
logger.info("Queuing summary generation task.")
summarize_transcription.apply_async(
args=[formatted_transcription, email, sub, title], queue="summarize_queue"
args=[formatted_transcription, email, sub, title], queue=settings.summarize_queue
)
src/summary/summary/core/config.py
+4 -2
View File
@@ -35,7 +35,7 @@ class Settings(BaseSettings):
openai_api_key: str
openai_base_url: str = "https://api.openai.com/v1"
openai_asr_model: str = "whisper-1"
openai_llm_model: str = "gpt-4o"
resume_llm_model: str = "Qwen/Qwen2.5-Coder-32B-Instruct-AWQ"
openai_max_retries: int = 0
resume_api_key: str
resume_endpoint: str
@@ -68,7 +68,9 @@ class Settings(BaseSettings):
task_tracker_redis_url: str = "redis://redis/0"
task_tracker_prefix: str = "task_metadata:"
# Queue redis
summarize_queue: str = "summarize_queue"
transcribe_queue: str = "transcribe_queue"
@lru_cache
def get_settings():
"""Load and cache application settings."""