Rodrigo Rodriguez
1a36f4aed2
- botserver/src/llm/local.rs: * Auto-configure embedding-url when empty (http://localhost:8082/v1/embeddings) * Auto-configure llm-url when empty (http://localhost:8081/v1/chat/completions) * Persist defaults to bot_configuration table via ConfigManager * Fix llama-server command path and add LD_LIBRARY_PATH * Fix working_dir for Linux (not just Windows) * Fix embedding server args: --embeddings --pooling mean --ctx-size 512 - botserver/src/core/bootstrap/bootstrap_manager.rs: * Temp disable alm-ci startup (causes 20s hang in bootstrap) * Remove unused alm_ci_health_check import - Database changes: * bot_configuration.embedding-model: 'bge-small-en-v1.5-f32.gguf' (full filename) * bot_configuration.embedding-url: 'http://localhost:8082/v1/embeddings' (auto-generated) - Testing: * Embedding server responds on port 8082 * Generates 384-dimension embeddings * KB file exists: minio://default.gbai/default.gbkb/manual/test-kb-manual.txt * Next: Verify KB indexing and search functionality Refs: #498
73 lines
2 KiB
Python
73 lines
2 KiB
Python
from pathlib import Path
|
|
from typing import Optional
|
|
|
|
from pydantic_settings import BaseSettings, SettingsConfigDict
|
|
|
|
|
|
class Settings(BaseSettings):
|
|
model_config = SettingsConfigDict(
|
|
env_file=".env",
|
|
env_file_encoding="utf-8",
|
|
case_sensitive=False,
|
|
extra="ignore",
|
|
)
|
|
|
|
env: str = "development"
|
|
host: str = "0.0.0.0"
|
|
port: int = 8085
|
|
log_level: str = "INFO"
|
|
api_v1_prefix: str = "/api"
|
|
project_name: str = "BotModels API"
|
|
version: str = "2.0.0"
|
|
api_key: str = "change-me"
|
|
commit: str = "unknown"
|
|
|
|
# External Providers for Speech (Optional)
|
|
groq_api_key: Optional[str] = None
|
|
openai_api_key: Optional[str] = None
|
|
|
|
# Image generation model
|
|
image_model_path: str = "./models/stable-diffusion-v1-5"
|
|
image_steps: int = 4
|
|
image_width: int = 512
|
|
image_height: int = 512
|
|
image_gpu_layers: int = 20
|
|
image_batch_size: int = 1
|
|
|
|
# Video generation model
|
|
video_model_path: str = "./models/zeroscope-v2"
|
|
video_frames: int = 24
|
|
video_fps: int = 8
|
|
video_width: int = 320
|
|
video_height: int = 576
|
|
video_gpu_layers: int = 15
|
|
video_batch_size: int = 1
|
|
|
|
# Speech/TTS model
|
|
speech_model_path: str = "./models/tts"
|
|
|
|
# Vision model (BLIP2 for captioning)
|
|
vision_model_path: str = "./models/blip2"
|
|
|
|
# Whisper model for speech-to-text
|
|
whisper_model_path: str = "./models/whisper"
|
|
|
|
# Real-time Audio model for speech-to-speech
|
|
realtime_audio_model_path: str = "./models/realtime_audio"
|
|
|
|
# Device configuration
|
|
device: str = "cuda"
|
|
|
|
# Output directory for generated files
|
|
output_dir: Path = Path("./outputs")
|
|
|
|
@property
|
|
def is_production(self) -> bool:
|
|
return self.env == "production"
|
|
|
|
|
|
settings = Settings()
|
|
settings.output_dir.mkdir(parents=True, exist_ok=True)
|
|
(settings.output_dir / "images").mkdir(exist_ok=True)
|
|
(settings.output_dir / "videos").mkdir(exist_ok=True)
|
|
(settings.output_dir / "audio").mkdir(exist_ok=True)
|