aljaz/enclava
1
0
Fork 0
mirror of https://github.com/aljazceru/enclava.git synced 2025-12-17 07:24:34 +01:00
Code Issues Packages Projects Releases Wiki Activity

rag improvements

Browse Source
This commit is contained in:
Aljaz Ceru
2025-09-23 15:26:54 +02:00
parent 354b43494d
commit f8d127ff42
30 changed files with 817 additions and 2428 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
backend/app/services/llm/config.py
View File

@@ -16,6 +16,7 @@ from .models import ResilienceConfig
class ProviderConfig(BaseModel):
"""Configuration for an LLM provider"""
name: str = Field(..., description="Provider name")
provider_type: str = Field(..., description="Provider type (e.g., 'openai', 'privatemode')")
enabled: bool = Field(True, description="Whether provider is enabled")
base_url: str = Field(..., description="Provider base URL")
api_key_env_var: str = Field(..., description="Environment variable for API key")
@@ -53,9 +54,6 @@ class LLMServiceConfig(BaseModel):
enable_security_checks: bool = Field(True, description="Enable security validation")
enable_metrics_collection: bool = Field(True, description="Enable metrics collection")
# Security settings
security_risk_threshold: float = Field(0.8, ge=0.0, le=1.0, description="Risk threshold for blocking")
security_warning_threshold: float = Field(0.6, ge=0.0, le=1.0, description="Risk threshold for warnings")
max_prompt_length: int = Field(50000, ge=1000, description="Maximum prompt length")
max_response_length: int = Field(32000, ge=1000, description="Maximum response length")
@@ -78,12 +76,6 @@ class LLMServiceConfig(BaseModel):
# Model routing (model_name -> provider_name)
model_routing: Dict[str, str] = Field(default_factory=dict, description="Model to provider routing")
@validator('security_risk_threshold')
def validate_risk_threshold(cls, v, values):
warning_threshold = values.get('security_warning_threshold', 0.6)
if v <= warning_threshold:
raise ValueError("Risk threshold must be greater than warning threshold")
return v
def create_default_config() -> LLMServiceConfig:
@@ -93,6 +85,7 @@ def create_default_config() -> LLMServiceConfig:
# Models will be fetched dynamically from proxy /models endpoint
privatemode_config = ProviderConfig(
name="privatemode",
provider_type="privatemode",
enabled=True,
base_url=settings.PRIVATEMODE_PROXY_URL,
api_key_env_var="PRIVATEMODE_API_KEY",
@@ -119,9 +112,6 @@ def create_default_config() -> LLMServiceConfig:
config = LLMServiceConfig(
default_provider="privatemode",
enable_detailed_logging=settings.LOG_LLM_PROMPTS,
enable_security_checks=settings.API_SECURITY_ENABLED,
security_risk_threshold=settings.API_SECURITY_RISK_THRESHOLD,
security_warning_threshold=settings.API_SECURITY_WARNING_THRESHOLD,
providers={
"privatemode": privatemode_config
},