Fix merge conflicts and resolve circular import issues
- Resolve conflict markers in logging_config.py, main.py, app.py - Fix circular imports in story_writer services (image/audio/video generation) by using lazy imports for get_story_media_write_dir - Restore clean versions of: - sif_agents.py - tenant_provider_config.py - personalization_service.py - huggingface_provider.py - main_text_generation.py - logger_utils.py - Use setup_clean_logging() consistently across app.py and main.py - Restore verbose_mode handling in start_alwrity_backend.py
This commit is contained in:
ajaysi
@@ -2,8 +2,6 @@
|
||||
|
||||
This service provides the main LLM text generation functionality,
|
||||
migrated from the legacy lib/gpt_providers/text_generation/main_text_generation.py
|
||||
|
||||
This is a clean version that imports from modular components to avoid merge conflicts.
|
||||
"""
|
||||
|
||||
import os
|
||||
@@ -13,47 +11,9 @@ from datetime import datetime
|
||||
from loguru import logger
|
||||
from fastapi import HTTPException
|
||||
|
||||
# Import all functionality from our modular textgen_utils package
|
||||
from .textgen_utils import (
|
||||
llm_text_gen,
|
||||
check_gpt_provider,
|
||||
get_api_key,
|
||||
_normalize_provider,
|
||||
_parse_csv_env,
|
||||
_resolve_provider_sequence,
|
||||
_map_logical_model_to_provider_model,
|
||||
_resolve_model_sequence,
|
||||
)
|
||||
|
||||
# Re-export all the main functions for backward compatibility
|
||||
__all__ = [
|
||||
"llm_text_gen",
|
||||
"check_gpt_provider",
|
||||
"get_api_key",
|
||||
"_normalize_provider",
|
||||
"_parse_csv_env",
|
||||
"_resolve_provider_sequence",
|
||||
"_map_logical_model_to_provider_model",
|
||||
"_resolve_model_sequence",
|
||||
]
|
||||
|
||||
# Maintain any additional constants or configurations that might be needed
|
||||
PREMIUM_HF_MINIMAL_FALLBACK_MODELS = [
|
||||
"openai/gpt-oss-120b:groq",
|
||||
]
|
||||
|
||||
# Legacy compatibility - any imports that other modules might expect
|
||||
from .gemini_provider import gemini_text_response, gemini_structured_json_response
|
||||
from .huggingface_provider import huggingface_text_response, huggingface_structured_json_response
|
||||
<<<<<<< HEAD
|
||||
from .tenant_provider_config import tenant_provider_config_resolver
|
||||
from .routing_policy import (
|
||||
PREMIUM_DEFAULT_MODEL,
|
||||
SIF_LOW_COST_MODEL_DEFAULTS,
|
||||
resolve_text_provider_alias,
|
||||
)
|
||||
=======
|
||||
from ...utils.logger_utils import emit_routing_event
|
||||
|
||||
|
||||
def llm_text_gen(
|
||||
@@ -93,14 +53,17 @@ def llm_text_gen(
|
||||
frequency_penalty = 0.0
|
||||
presence_penalty = 0.0
|
||||
|
||||
# Check for GPT_PROVIDER environment variable
|
||||
env_provider = os.getenv('GPT_PROVIDER', '').lower()
|
||||
if env_provider in ['gemini', 'google']:
|
||||
provider_cfg = tenant_provider_config_resolver.resolve(
|
||||
modality="text",
|
||||
user_id=user_id,
|
||||
)
|
||||
selected_provider = (provider_cfg.selected_providers or [None])[0]
|
||||
if selected_provider in ["gemini", "google"]:
|
||||
gpt_provider = "google"
|
||||
model = "gemini-2.0-flash-001"
|
||||
elif env_provider in ['hf_response_api', 'huggingface', 'hf']:
|
||||
model = provider_cfg.model_policy.get("default_model") or "gemini-2.0-flash-001"
|
||||
elif selected_provider == "huggingface":
|
||||
gpt_provider = "huggingface"
|
||||
model = "mistralai/Mistral-7B-Instruct-v0.3:groq"
|
||||
model = provider_cfg.model_policy.get("default_model") or "mistralai/Mistral-7B-Instruct-v0.3:groq"
|
||||
|
||||
# Default blog characteristics
|
||||
blog_tone = "Professional"
|
||||
@@ -110,64 +73,32 @@ def llm_text_gen(
|
||||
blog_output_format = "markdown"
|
||||
blog_length = 2000
|
||||
|
||||
# Check which providers have API keys available using APIKeyManager
|
||||
api_key_manager = APIKeyManager()
|
||||
available_providers = []
|
||||
if api_key_manager.get_api_key("gemini"):
|
||||
available_providers.append("google")
|
||||
if api_key_manager.get_api_key("hf_token"):
|
||||
available_providers.append("huggingface")
|
||||
for provider in ("google", "huggingface"):
|
||||
if get_api_key(provider, user_id=user_id):
|
||||
available_providers.append(provider)
|
||||
|
||||
preferred_provider = env_provider or None
|
||||
flow_type = "text_generation"
|
||||
route_intent = "primary"
|
||||
fallback_count = 0
|
||||
fallback_models_tried = []
|
||||
|
||||
# If no environment variable set, auto-detect based on available keys
|
||||
if not env_provider:
|
||||
# Prefer Google Gemini if available, otherwise use Hugging Face
|
||||
if "google" in available_providers:
|
||||
gpt_provider = "google"
|
||||
model = "gemini-2.0-flash-001"
|
||||
elif "huggingface" in available_providers:
|
||||
gpt_provider = "huggingface"
|
||||
model = "mistralai/Mistral-7B-Instruct-v0.3:groq"
|
||||
if gpt_provider not in available_providers:
|
||||
logger.warning(f"[llm_text_gen] Provider {gpt_provider} unavailable for user {user_id}, falling back.")
|
||||
if available_providers:
|
||||
gpt_provider = available_providers[0]
|
||||
else:
|
||||
logger.error("[llm_text_gen] No API keys found for supported providers.")
|
||||
raise RuntimeError("No LLM API keys configured. Configure GEMINI_API_KEY or HF_TOKEN to enable AI responses.")
|
||||
else:
|
||||
# Environment variable was set, validate it's supported
|
||||
if gpt_provider not in available_providers:
|
||||
logger.warning(f"[llm_text_gen] Provider {gpt_provider} not available, falling back to available providers")
|
||||
if "google" in available_providers:
|
||||
gpt_provider = "google"
|
||||
model = "gemini-2.0-flash-001"
|
||||
elif "huggingface" in available_providers:
|
||||
gpt_provider = "huggingface"
|
||||
model = "mistralai/Mistral-7B-Instruct-v0.3:groq"
|
||||
else:
|
||||
raise RuntimeError("No supported providers available.")
|
||||
raise RuntimeError("No LLM API keys configured for tenant or environment defaults.")
|
||||
|
||||
# Ensure downstream provider clients (currently env-based) receive resolved key
|
||||
resolved_key = get_api_key(gpt_provider, user_id=user_id)
|
||||
if gpt_provider == "google" and resolved_key:
|
||||
os.environ["GEMINI_API_KEY"] = resolved_key
|
||||
os.environ.setdefault("GOOGLE_API_KEY", resolved_key)
|
||||
elif gpt_provider == "huggingface" and resolved_key:
|
||||
os.environ["HF_TOKEN"] = resolved_key
|
||||
|
||||
if gpt_provider == "huggingface" and preferred_hf_models:
|
||||
model = preferred_hf_models[0]
|
||||
logger.info(f"[llm_text_gen] Using preferred low-cost HF model: {model}")
|
||||
|
||||
fallback_models_tried.append(model)
|
||||
|
||||
logger.debug(f"[llm_text_gen] Using provider: {gpt_provider}, model: {model}")
|
||||
emit_routing_event(
|
||||
logger,
|
||||
flow_type=flow_type,
|
||||
route_intent=route_intent,
|
||||
provider_selected=gpt_provider,
|
||||
model_selected=model,
|
||||
preferred_provider=preferred_provider,
|
||||
fallback_count=fallback_count,
|
||||
fallback_models_tried=fallback_models_tried,
|
||||
tenant_user_id=user_id,
|
||||
extra={"available_providers": available_providers},
|
||||
)
|
||||
|
||||
# Map provider name to APIProvider enum (define at function scope for usage tracking)
|
||||
from models.subscription_models import APIProvider
|
||||
@@ -311,8 +242,7 @@ def llm_text_gen(
|
||||
model=model,
|
||||
temperature=temperature,
|
||||
max_tokens=max_tokens,
|
||||
system_prompt=system_instructions,
|
||||
tenant_user_id=user_id
|
||||
system_prompt=system_instructions
|
||||
)
|
||||
else:
|
||||
response_text = huggingface_text_response(
|
||||
@@ -321,8 +251,7 @@ def llm_text_gen(
|
||||
temperature=temperature,
|
||||
max_tokens=max_tokens,
|
||||
top_p=top_p,
|
||||
system_prompt=system_instructions,
|
||||
tenant_user_id=user_id
|
||||
system_prompt=system_instructions
|
||||
)
|
||||
else:
|
||||
logger.error(f"[llm_text_gen] Unknown provider: {gpt_provider}")
|
||||
@@ -366,34 +295,17 @@ def llm_text_gen(
|
||||
try:
|
||||
logger.info(f"[llm_text_gen] Trying SINGLE fallback provider: {fallback_provider}")
|
||||
actual_provider_used = fallback_provider
|
||||
fallback_count += 1
|
||||
route_intent = "fallback"
|
||||
|
||||
# Update provider enum for fallback
|
||||
if fallback_provider == "google":
|
||||
provider_enum = APIProvider.GEMINI
|
||||
actual_provider_name = "gemini"
|
||||
fallback_model = "gemini-2.0-flash-lite"
|
||||
fallback_models_tried.append(fallback_model)
|
||||
elif fallback_provider == "huggingface":
|
||||
provider_enum = APIProvider.MISTRAL
|
||||
actual_provider_name = "huggingface"
|
||||
fallback_model = "mistralai/Mistral-7B-Instruct-v0.3:groq"
|
||||
fallback_models_tried.append(fallback_model)
|
||||
|
||||
emit_routing_event(
|
||||
logger,
|
||||
flow_type=flow_type,
|
||||
route_intent=route_intent,
|
||||
provider_selected=fallback_provider,
|
||||
model_selected=fallback_model,
|
||||
preferred_provider=preferred_provider,
|
||||
fallback_count=fallback_count,
|
||||
fallback_models_tried=fallback_models_tried,
|
||||
tenant_user_id=user_id,
|
||||
extra={"available_providers": available_providers},
|
||||
)
|
||||
|
||||
if fallback_provider == "google":
|
||||
if json_struct:
|
||||
response_text = gemini_structured_json_response(
|
||||
@@ -422,8 +334,7 @@ def llm_text_gen(
|
||||
model="mistralai/Mistral-7B-Instruct-v0.3:groq",
|
||||
temperature=temperature,
|
||||
max_tokens=max_tokens,
|
||||
system_prompt=system_instructions,
|
||||
tenant_user_id=user_id
|
||||
system_prompt=system_instructions
|
||||
)
|
||||
else:
|
||||
response_text = huggingface_text_response(
|
||||
@@ -432,8 +343,7 @@ def llm_text_gen(
|
||||
temperature=temperature,
|
||||
max_tokens=max_tokens,
|
||||
top_p=top_p,
|
||||
system_prompt=system_instructions,
|
||||
tenant_user_id=user_id
|
||||
system_prompt=system_instructions
|
||||
)
|
||||
|
||||
# TRACK USAGE after successful fallback call
|
||||
@@ -472,18 +382,16 @@ def check_gpt_provider(gpt_provider: str) -> bool:
|
||||
supported_providers = ["google", "huggingface"]
|
||||
return gpt_provider in supported_providers
|
||||
|
||||
def get_api_key(gpt_provider: str) -> Optional[str]:
|
||||
def get_api_key(gpt_provider: str, user_id: Optional[str] = None) -> Optional[str]:
|
||||
"""Get API key for the specified provider."""
|
||||
try:
|
||||
api_key_manager = APIKeyManager()
|
||||
provider_mapping = {
|
||||
"google": "gemini",
|
||||
"huggingface": "hf_token"
|
||||
"huggingface": "huggingface"
|
||||
}
|
||||
|
||||
mapped_provider = provider_mapping.get(gpt_provider, gpt_provider)
|
||||
return api_key_manager.get_api_key(mapped_provider)
|
||||
key, _source = tenant_provider_config_resolver.resolve_provider_key(mapped_provider, user_id=user_id)
|
||||
return key
|
||||
except Exception as e:
|
||||
logger.error(f"[get_api_key] Error getting API key for {gpt_provider}: {str(e)}")
|
||||
return None
|
||||
>>>>>>> pr-421
|
||||
|
||||
Reference in New Issue
Block a user