feat: added global llm configurations

surfsense_backend/.gitignore

@@ -11,3 +11,4 @@ celerybeat-schedule*
 celerybeat-schedule.*
 celerybeat-schedule.dir
 celerybeat-schedule.bak
+global_llm_config.yaml

surfsense_backend/alembic/versions/36_remove_fk_constraints_for_global_llm_configs.py

@@ -0,0 +1,73 @@
+"""remove_fk_constraints_for_global_llm_configs
+
+Revision ID: 36
+Revises: 35
+Create Date: 2025-11-13 23:20:12.912741
+
+"""
+
+from collections.abc import Sequence
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "36"
+down_revision: str | None = "35"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    """
+    Remove foreign key constraints on LLM preference columns to allow global configs (negative IDs).
+
+    Global LLM configs use negative IDs and don't exist in the llm_configs table,
+    so we need to remove the foreign key constraints that were preventing their use.
+    """
+    # Drop the foreign key constraints
+    op.drop_constraint(
+        "user_search_space_preferences_long_context_llm_id_fkey",
+        "user_search_space_preferences",
+        type_="foreignkey",
+    )
+    op.drop_constraint(
+        "user_search_space_preferences_fast_llm_id_fkey",
+        "user_search_space_preferences",
+        type_="foreignkey",
+    )
+    op.drop_constraint(
+        "user_search_space_preferences_strategic_llm_id_fkey",
+        "user_search_space_preferences",
+        type_="foreignkey",
+    )
+
+
+def downgrade() -> None:
+    """
+    Re-add foreign key constraints (will fail if any negative IDs exist in the table).
+    """
+    # Re-add the foreign key constraints
+    op.create_foreign_key(
+        "user_search_space_preferences_long_context_llm_id_fkey",
+        "user_search_space_preferences",
+        "llm_configs",
+        ["long_context_llm_id"],
+        ["id"],
+        ondelete="SET NULL",
+    )
+    op.create_foreign_key(
+        "user_search_space_preferences_fast_llm_id_fkey",
+        "user_search_space_preferences",
+        "llm_configs",
+        ["fast_llm_id"],
+        ["id"],
+        ondelete="SET NULL",
+    )
+    op.create_foreign_key(
+        "user_search_space_preferences_strategic_llm_id_fkey",
+        "user_search_space_preferences",
+        "llm_configs",
+        ["strategic_llm_id"],
+        ["id"],
+        ondelete="SET NULL",
+    )

surfsense_backend/app/config/__init__.py

@@ -3,6 +3,7 @@ import shutil
 from pathlib import Path
 from typing import Any
 
+import yaml
 from chonkie import AutoEmbeddings, CodeChunker, RecursiveChunker
 from chonkie.embeddings.azure_openai import AzureOpenAIEmbeddings
 from chonkie.embeddings.registry import EmbeddingsRegistry
@@ -80,6 +81,36 @@ def is_ffmpeg_installed():
     return shutil.which("ffmpeg") is not None
 
 
+def load_global_llm_configs():
+    """
+    Load global LLM configurations from YAML file.
+    Falls back to example file if main file doesn't exist.
+
+    Returns:
+        list: List of global LLM config dictionaries, or empty list if file doesn't exist
+    """
+    # Try main config file first
+    global_config_file = BASE_DIR / "app" / "config" / "global_llm_config.yaml"
+
+    # Fall back to example file for testing
+    # if not global_config_file.exists():
+    #     global_config_file = BASE_DIR / "app" / "config" / "global_llm_config.example.yaml"
+    #     if global_config_file.exists():
+    #         print("Info: Using global_llm_config.example.yaml (copy to global_llm_config.yaml for production)")
+
+    if not global_config_file.exists():
+        # No global configs available
+        return []
+
+    try:
+        with open(global_config_file, encoding="utf-8") as f:
+            data = yaml.safe_load(f)
+            return data.get("global_llm_configs", [])
+    except Exception as e:
+        print(f"Warning: Failed to load global LLM configs: {e}")
+        return []
+
+
 class Config:
     # Check if ffmpeg is installed
     if not is_ffmpeg_installed():
@@ -122,6 +153,11 @@ class Config:
     # LLM instances are now managed per-user through the LLMConfig system
     # Legacy environment variables removed in favor of user-specific configurations
 
+    # Global LLM Configurations (optional)
+    # Load from global_llm_config.yaml if available
+    # These can be used as default options for users
+    GLOBAL_LLM_CONFIGS = load_global_llm_configs()
+
     # Chonkie Configuration | Edit this to your needs
     EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL")
     # Azure OpenAI credentials from environment variables

surfsense_backend/app/config/global_llm_config.example.yaml

@@ -0,0 +1,80 @@
+# Global LLM Configuration
+# 
+# SETUP INSTRUCTIONS:
+# 1. For production: Copy this file to global_llm_config.yaml and add your real API keys
+# 2. For testing: The system will use this example file automatically if global_llm_config.yaml doesn't exist
+#
+# NOTE: The example API keys below are placeholders and won't work.
+# Replace them with your actual API keys to enable global configurations.
+#
+# These configurations will be available to all users as a convenient option
+# Users can choose to use these global configs or add their own
+
+global_llm_configs:
+  # Example: OpenAI GPT-4 Turbo
+  - id: -1
+    name: "Global GPT-4 Turbo"
+    provider: "OPENAI"
+    model_name: "gpt-4-turbo-preview"
+    api_key: "sk-your-openai-api-key-here"
+    api_base: ""
+    language: "English"
+    litellm_params:
+      temperature: 0.7
+      max_tokens: 4000
+
+  # Example: Anthropic Claude 3 Opus
+  - id: -2
+    name: "Global Claude 3 Opus"
+    provider: "ANTHROPIC"
+    model_name: "claude-3-opus-20240229"
+    api_key: "sk-ant-your-anthropic-api-key-here"
+    api_base: ""
+    language: "English"
+    litellm_params:
+      temperature: 0.7
+      max_tokens: 4000
+
+  # Example: Fast model - GPT-3.5 Turbo
+  - id: -3
+    name: "Global GPT-3.5 Turbo"
+    provider: "OPENAI"
+    model_name: "gpt-3.5-turbo"
+    api_key: "sk-your-openai-api-key-here"
+    api_base: ""
+    language: "English"
+    litellm_params:
+      temperature: 0.5
+      max_tokens: 2000
+
+  # Example: Chinese LLM - DeepSeek
+  - id: -4
+    name: "Global DeepSeek Chat"
+    provider: "DEEPSEEK"
+    model_name: "deepseek-chat"
+    api_key: "your-deepseek-api-key-here"
+    api_base: "https://api.deepseek.com/v1"
+    language: "Chinese"
+    litellm_params:
+      temperature: 0.7
+      max_tokens: 4000
+
+  # Example: Groq - Fast inference
+  - id: -5
+    name: "Global Groq Llama 3"
+    provider: "GROQ"
+    model_name: "llama3-70b-8192"
+    api_key: "your-groq-api-key-here"
+    api_base: ""
+    language: "English"
+    litellm_params:
+      temperature: 0.7
+      max_tokens: 8000
+
+# Notes:
+# - Use negative IDs to distinguish global configs from user configs
+# - IDs should be unique and sequential (e.g., -1, -2, -3, etc.)
+# - The 'api_key' field will not be exposed to users via API
+# - Users can select these configs for their long_context, fast, or strategic LLM roles
+# - All standard LiteLLM providers are supported
+

surfsense_backend/app/db.py

@@ -348,15 +348,11 @@ class UserSearchSpacePreference(BaseModel, TimestampMixin):
     )
 
     # User-specific LLM preferences for this search space
-    long_context_llm_id = Column(
-        Integer, ForeignKey("llm_configs.id", ondelete="SET NULL"), nullable=True
-    )
-    fast_llm_id = Column(
-        Integer, ForeignKey("llm_configs.id", ondelete="SET NULL"), nullable=True
-    )
-    strategic_llm_id = Column(
-        Integer, ForeignKey("llm_configs.id", ondelete="SET NULL"), nullable=True
-    )
+    # Note: These can be negative IDs for global configs (from YAML) or positive IDs for custom configs (from DB)
+    # Foreign keys removed to support global configs with negative IDs
+    long_context_llm_id = Column(Integer, nullable=True)
+    fast_llm_id = Column(Integer, nullable=True)
+    strategic_llm_id = Column(Integer, nullable=True)
 
     # Future RBAC fields can be added here
     # role = Column(String(50), nullable=True)  # e.g., 'owner', 'editor', 'viewer'
@@ -365,13 +361,12 @@ class UserSearchSpacePreference(BaseModel, TimestampMixin):
     user = relationship("User", back_populates="search_space_preferences")
     search_space = relationship("SearchSpace", back_populates="user_preferences")
 
-    long_context_llm = relationship(
-        "LLMConfig", foreign_keys=[long_context_llm_id], post_update=True
-    )
-    fast_llm = relationship("LLMConfig", foreign_keys=[fast_llm_id], post_update=True)
-    strategic_llm = relationship(
-        "LLMConfig", foreign_keys=[strategic_llm_id], post_update=True
-    )
+    # Note: Relationships removed because foreign keys no longer exist
+    # Global configs (negative IDs) don't exist in llm_configs table
+    # Application code manually fetches configs when needed
+    # long_context_llm = relationship("LLMConfig", foreign_keys=[long_context_llm_id], post_update=True)
+    # fast_llm = relationship("LLMConfig", foreign_keys=[fast_llm_id], post_update=True)
+    # strategic_llm = relationship("LLMConfig", foreign_keys=[strategic_llm_id], post_update=True)
 
 
 class Log(BaseModel, TimestampMixin):

surfsense_backend/app/routes/chats_routes.py

@@ -68,9 +68,9 @@ async def handle_chat_data(
                 selectinload(UserSearchSpacePreference.search_space).selectinload(
                     SearchSpace.llm_configs
                 ),
-                selectinload(UserSearchSpacePreference.long_context_llm),
-                selectinload(UserSearchSpacePreference.fast_llm),
-                selectinload(UserSearchSpacePreference.strategic_llm),
+                # Note: Removed selectinload for LLM relationships as they no longer exist
+                # Global configs (negative IDs) don't have foreign keys
+                # LLM configs are now fetched manually when needed
             )
             .filter(
                 UserSearchSpacePreference.search_space_id == search_space_id,
@@ -81,6 +81,8 @@ async def handle_chat_data(
         # print("UserSearchSpacePreference:", user_preference)
 
         language = None
+        llm_configs = []  # Initialize to empty list
+
         if (
             user_preference
             and user_preference.search_space
@@ -88,16 +90,36 @@ async def handle_chat_data(
         ):
             llm_configs = user_preference.search_space.llm_configs
 
-            for preferred_llm in [
-                user_preference.fast_llm,
-                user_preference.long_context_llm,
-                user_preference.strategic_llm,
-            ]:
-                if preferred_llm and getattr(preferred_llm, "language", None):
-                    language = preferred_llm.language
-                    break
+            # Manually fetch LLM configs since relationships no longer exist
+            # Check fast_llm, long_context_llm, and strategic_llm IDs
+            from app.config import config as app_config
 
-        if not language:
+            for llm_id in [
+                user_preference.fast_llm_id,
+                user_preference.long_context_llm_id,
+                user_preference.strategic_llm_id,
+            ]:
+                if llm_id is not None:
+                    # Check if it's a global config (negative ID)
+                    if llm_id < 0:
+                        # Look in global configs
+                        for global_cfg in app_config.GLOBAL_LLM_CONFIGS:
+                            if global_cfg.get("id") == llm_id:
+                                language = global_cfg.get("language")
+                                if language:
+                                    break
+                    else:
+                        # Look in custom configs
+                        for llm_config in llm_configs:
+                            if llm_config.id == llm_id and getattr(
+                                llm_config, "language", None
+                            ):
+                                language = llm_config.language
+                                break
+                    if language:
+                        break
+
+        if not language and llm_configs:
             first_llm_config = llm_configs[0]
             language = getattr(first_llm_config, "language", None)
 

surfsense_backend/app/routes/llm_config_routes.py

@@ -1,9 +1,11 @@
+import logging
+
 from fastapi import APIRouter, Depends, HTTPException
 from pydantic import BaseModel
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
-from sqlalchemy.orm import selectinload
 
+from app.config import config
 from app.db import (
     LLMConfig,
     SearchSpace,
@@ -16,6 +18,7 @@ from app.services.llm_service import validate_llm_config
 from app.users import current_active_user
 
 router = APIRouter()
+logger = logging.getLogger(__name__)
 
 
 # Helper function to check search space access
@@ -43,16 +46,11 @@ async def get_or_create_user_preference(
 ) -> UserSearchSpacePreference:
     """Get or create user preference for a search space"""
     result = await session.execute(
-        select(UserSearchSpacePreference)
-        .filter(
+        select(UserSearchSpacePreference).filter(
             UserSearchSpacePreference.user_id == user_id,
             UserSearchSpacePreference.search_space_id == search_space_id,
         )
-        .options(
-            selectinload(UserSearchSpacePreference.long_context_llm),
-            selectinload(UserSearchSpacePreference.fast_llm),
-            selectinload(UserSearchSpacePreference.strategic_llm),
-        )
+        # Removed selectinload options since relationships no longer exist
     )
     preference = result.scalars().first()
 
@@ -88,6 +86,58 @@ class LLMPreferencesRead(BaseModel):
     strategic_llm: LLMConfigRead | None = None
 
 
+class GlobalLLMConfigRead(BaseModel):
+    """Schema for reading global LLM configs (without API key)"""
+
+    id: int
+    name: str
+    provider: str
+    custom_provider: str | None = None
+    model_name: str
+    api_base: str | None = None
+    language: str | None = None
+    litellm_params: dict | None = None
+    is_global: bool = True
+
+
+# Global LLM Config endpoints
+
+
+@router.get("/global-llm-configs", response_model=list[GlobalLLMConfigRead])
+async def get_global_llm_configs(
+    user: User = Depends(current_active_user),
+):
+    """
+    Get all available global LLM configurations.
+    These are pre-configured by the system administrator and available to all users.
+    API keys are not exposed through this endpoint.
+    """
+    try:
+        global_configs = config.GLOBAL_LLM_CONFIGS
+
+        # Remove API keys from response
+        safe_configs = []
+        for cfg in global_configs:
+            safe_config = {
+                "id": cfg.get("id"),
+                "name": cfg.get("name"),
+                "provider": cfg.get("provider"),
+                "custom_provider": cfg.get("custom_provider"),
+                "model_name": cfg.get("model_name"),
+                "api_base": cfg.get("api_base"),
+                "language": cfg.get("language"),
+                "litellm_params": cfg.get("litellm_params", {}),
+                "is_global": True,
+            }
+            safe_configs.append(safe_config)
+
+        return safe_configs
+    except Exception as e:
+        raise HTTPException(
+            status_code=500, detail=f"Failed to fetch global LLM configs: {e!s}"
+        ) from e
+
+
 @router.post("/llm-configs", response_model=LLMConfigRead)
 async def create_llm_config(
     llm_config: LLMConfigCreate,
@@ -309,13 +359,49 @@ async def get_user_llm_preferences(
             session, user.id, search_space_id
         )
 
+        # Helper function to get config (global or custom)
+        async def get_config_for_id(config_id):
+            if config_id is None:
+                return None
+
+            # Check if it's a global config (negative ID)
+            if config_id < 0:
+                for cfg in config.GLOBAL_LLM_CONFIGS:
+                    if cfg.get("id") == config_id:
+                        # Return as LLMConfigRead-compatible dict
+                        return {
+                            "id": cfg.get("id"),
+                            "name": cfg.get("name"),
+                            "provider": cfg.get("provider"),
+                            "custom_provider": cfg.get("custom_provider"),
+                            "model_name": cfg.get("model_name"),
+                            "api_key": "***GLOBAL***",  # Don't expose the actual key
+                            "api_base": cfg.get("api_base"),
+                            "language": cfg.get("language"),
+                            "litellm_params": cfg.get("litellm_params"),
+                            "created_at": None,
+                            "search_space_id": search_space_id,
+                        }
+                return None
+
+            # It's a custom config, fetch from database
+            result = await session.execute(
+                select(LLMConfig).filter(LLMConfig.id == config_id)
+            )
+            return result.scalars().first()
+
+        # Get the configs (from DB for custom, or constructed for global)
+        long_context_llm = await get_config_for_id(preference.long_context_llm_id)
+        fast_llm = await get_config_for_id(preference.fast_llm_id)
+        strategic_llm = await get_config_for_id(preference.strategic_llm_id)
+
         return {
             "long_context_llm_id": preference.long_context_llm_id,
             "fast_llm_id": preference.fast_llm_id,
             "strategic_llm_id": preference.strategic_llm_id,
-            "long_context_llm": preference.long_context_llm,
-            "fast_llm": preference.fast_llm,
-            "strategic_llm": preference.strategic_llm,
+            "long_context_llm": long_context_llm,
+            "fast_llm": fast_llm,
+            "strategic_llm": strategic_llm,
         }
     except HTTPException:
         raise
@@ -353,29 +439,57 @@ async def update_user_llm_preferences(
 
         for _key, llm_config_id in update_data.items():
             if llm_config_id is not None:
-                # Verify the LLM config belongs to the search space
-                result = await session.execute(
-                    select(LLMConfig).filter(
-                        LLMConfig.id == llm_config_id,
-                        LLMConfig.search_space_id == search_space_id,
-                    )
-                )
-                llm_config = result.scalars().first()
-                if not llm_config:
-                    raise HTTPException(
-                        status_code=404,
-                        detail=f"LLM configuration {llm_config_id} not found in this search space",
-                    )
+                # Check if this is a global config (negative ID)
+                if llm_config_id < 0:
+                    # Validate global config exists
+                    global_config = None
+                    for cfg in config.GLOBAL_LLM_CONFIGS:
+                        if cfg.get("id") == llm_config_id:
+                            global_config = cfg
+                            break
 
-                # Collect language for consistency check
-                languages.add(llm_config.language)
+                    if not global_config:
+                        raise HTTPException(
+                            status_code=404,
+                            detail=f"Global LLM configuration {llm_config_id} not found",
+                        )
 
-        # Check if all selected LLM configs have the same language
+                    # Collect language for consistency check (if explicitly set)
+                    lang = global_config.get("language")
+                    if lang and lang.strip():  # Only add non-empty languages
+                        languages.add(lang.strip())
+                else:
+                    # Verify the LLM config belongs to the search space (custom config)
+                    result = await session.execute(
+                        select(LLMConfig).filter(
+                            LLMConfig.id == llm_config_id,
+                            LLMConfig.search_space_id == search_space_id,
+                        )
+                    )
+                    llm_config = result.scalars().first()
+                    if not llm_config:
+                        raise HTTPException(
+                            status_code=404,
+                            detail=f"LLM configuration {llm_config_id} not found in this search space",
+                        )
+
+                    # Collect language for consistency check (if explicitly set)
+                    if llm_config.language and llm_config.language.strip():
+                        languages.add(llm_config.language.strip())
+
+        # Language consistency check - only warn if there are multiple explicit languages
+        # Allow mixing configs with and without language settings
         if len(languages) > 1:
-            raise HTTPException(
-                status_code=400,
-                detail="All selected LLM configurations must have the same language setting",
+            # Log warning but allow the operation
+            logger.warning(
+                f"Multiple languages detected in LLM selection for search_space {search_space_id}: {languages}. "
+                "This may affect response quality."
             )
+            # Don't raise an exception - allow users to proceed
+            # raise HTTPException(
+            #     status_code=400,
+            #     detail="All selected LLM configurations must have the same language setting",
+            # )
 
         # Update user preferences
         for key, value in update_data.items():
@@ -384,19 +498,50 @@ async def update_user_llm_preferences(
         await session.commit()
         await session.refresh(preference)
 
-        # Reload relationships
-        await session.refresh(
-            preference, ["long_context_llm", "fast_llm", "strategic_llm"]
-        )
+        # Helper function to get config (global or custom)
+        async def get_config_for_id(config_id):
+            if config_id is None:
+                return None
+
+            # Check if it's a global config (negative ID)
+            if config_id < 0:
+                for cfg in config.GLOBAL_LLM_CONFIGS:
+                    if cfg.get("id") == config_id:
+                        # Return as LLMConfigRead-compatible dict
+                        return {
+                            "id": cfg.get("id"),
+                            "name": cfg.get("name"),
+                            "provider": cfg.get("provider"),
+                            "custom_provider": cfg.get("custom_provider"),
+                            "model_name": cfg.get("model_name"),
+                            "api_key": "***GLOBAL***",  # Don't expose the actual key
+                            "api_base": cfg.get("api_base"),
+                            "language": cfg.get("language"),
+                            "litellm_params": cfg.get("litellm_params"),
+                            "created_at": None,
+                            "search_space_id": search_space_id,
+                        }
+                return None
+
+            # It's a custom config, fetch from database
+            result = await session.execute(
+                select(LLMConfig).filter(LLMConfig.id == config_id)
+            )
+            return result.scalars().first()
+
+        # Get the configs (from DB for custom, or constructed for global)
+        long_context_llm = await get_config_for_id(preference.long_context_llm_id)
+        fast_llm = await get_config_for_id(preference.fast_llm_id)
+        strategic_llm = await get_config_for_id(preference.strategic_llm_id)
 
         # Return updated preferences
         return {
             "long_context_llm_id": preference.long_context_llm_id,
             "fast_llm_id": preference.fast_llm_id,
             "strategic_llm_id": preference.strategic_llm_id,
-            "long_context_llm": preference.long_context_llm,
-            "fast_llm": preference.fast_llm,
-            "strategic_llm": preference.strategic_llm,
+            "long_context_llm": long_context_llm,
+            "fast_llm": fast_llm,
+            "strategic_llm": strategic_llm,
         }
     except HTTPException:
         raise

surfsense_backend/app/schemas/llm_config.py

@@ -62,7 +62,11 @@ class LLMConfigUpdate(BaseModel):
 
 class LLMConfigRead(LLMConfigBase, IDModel, TimestampModel):
     id: int
-    created_at: datetime
-    search_space_id: int
+    created_at: datetime | None = Field(
+        None, description="Creation timestamp (None for global configs)"
+    )
+    search_space_id: int | None = Field(
+        None, description="Search space ID (None for global configs)"
+    )
 
     model_config = ConfigDict(from_attributes=True)

surfsense_backend/app/services/llm_service.py

@@ -6,6 +6,7 @@ from langchain_litellm import ChatLiteLLM
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
 
+from app.config import config
 from app.db import LLMConfig, UserSearchSpacePreference
 
 # Configure litellm to automatically drop unsupported parameters
@@ -20,6 +21,27 @@ class LLMRole:
     STRATEGIC = "strategic"
 
 
+def get_global_llm_config(llm_config_id: int) -> dict | None:
+    """
+    Get a global LLM configuration by ID.
+    Global configs have negative IDs.
+
+    Args:
+        llm_config_id: The ID of the global config (should be negative)
+
+    Returns:
+        dict: Global config dictionary or None if not found
+    """
+    if llm_config_id >= 0:
+        return None
+
+    for cfg in config.GLOBAL_LLM_CONFIGS:
+        if cfg.get("id") == llm_config_id:
+            return cfg
+
+    return None
+
+
 async def validate_llm_config(
     provider: str,
     model_name: str,
@@ -171,7 +193,70 @@ async def get_user_llm_instance(
             )
             return None
 
-        # Get the LLM configuration
+        # Check if this is a global config (negative ID)
+        if llm_config_id < 0:
+            global_config = get_global_llm_config(llm_config_id)
+            if not global_config:
+                logger.error(f"Global LLM config {llm_config_id} not found")
+                return None
+
+            # Build model string for global config
+            if global_config.get("custom_provider"):
+                model_string = (
+                    f"{global_config['custom_provider']}/{global_config['model_name']}"
+                )
+            else:
+                provider_map = {
+                    "OPENAI": "openai",
+                    "ANTHROPIC": "anthropic",
+                    "GROQ": "groq",
+                    "COHERE": "cohere",
+                    "GOOGLE": "gemini",
+                    "OLLAMA": "ollama",
+                    "MISTRAL": "mistral",
+                    "AZURE_OPENAI": "azure",
+                    "OPENROUTER": "openrouter",
+                    "COMETAPI": "cometapi",
+                    "XAI": "xai",
+                    "BEDROCK": "bedrock",
+                    "AWS_BEDROCK": "bedrock",
+                    "VERTEX_AI": "vertex_ai",
+                    "TOGETHER_AI": "together_ai",
+                    "FIREWORKS_AI": "fireworks_ai",
+                    "REPLICATE": "replicate",
+                    "PERPLEXITY": "perplexity",
+                    "ANYSCALE": "anyscale",
+                    "DEEPINFRA": "deepinfra",
+                    "CEREBRAS": "cerebras",
+                    "SAMBANOVA": "sambanova",
+                    "AI21": "ai21",
+                    "CLOUDFLARE": "cloudflare",
+                    "DATABRICKS": "databricks",
+                    "DEEPSEEK": "openai",
+                    "ALIBABA_QWEN": "openai",
+                    "MOONSHOT": "openai",
+                    "ZHIPU": "openai",
+                }
+                provider_prefix = provider_map.get(
+                    global_config["provider"], global_config["provider"].lower()
+                )
+                model_string = f"{provider_prefix}/{global_config['model_name']}"
+
+            # Create ChatLiteLLM instance from global config
+            litellm_kwargs = {
+                "model": model_string,
+                "api_key": global_config["api_key"],
+            }
+
+            if global_config.get("api_base"):
+                litellm_kwargs["api_base"] = global_config["api_base"]
+
+            if global_config.get("litellm_params"):
+                litellm_kwargs.update(global_config["litellm_params"])
+
+            return ChatLiteLLM(**litellm_kwargs)
+
+        # Get the LLM configuration from database (user-specific config)
         result = await session.execute(
             select(LLMConfig).where(
                 LLMConfig.id == llm_config_id,