Second version

app/config.py

@@ -1,20 +1,23 @@
 import os
-from pydantic import BaseModel
+
 from pyaml_env import parse_config
+from pydantic import BaseModel
 
 
-class Settings(BaseModel):
-    class PyrisSettings(BaseModel):
-        class APIKeyConfig(BaseModel):
-            token: str
-            comment: str
-            llm_access: list[str]
+class LLMModelConfig(BaseModel):
+    name: str
+    description: str
+    llm_credentials: dict
+
 
-        class LLMModelConfig(BaseModel):
-            name: str
-            description: str
-            llm_credentials: dict
+class APIKeyConfig(BaseModel):
+    token: str
+    comment: str
+    llm_access: list[str]
 
+
+class Settings(BaseModel):
+    class PyrisSettings(BaseModel):
         api_keys: list[APIKeyConfig]
         llms: dict[str, LLMModelConfig]
 
@@ -31,10 +34,3 @@ def get_settings(cls):
 
 
 settings = Settings.get_settings()
-
-# get keys of settings.pyris.llms and for loop over them with print
-for key in set(settings.pyris.llms.keys()):
-    print(key)
-
-for key in settings.pyris.api_keys:
-    print(key)

app/dependencies.py

@@ -1,10 +1,10 @@
 from fastapi import Depends
 from starlette.requests import Request as StarletteRequest
-from app.config import settings
+from app.config import settings, APIKeyConfig
 
 from app.core.custom_exceptions import (
     PermissionDeniedException,
-    RequiresAuthenticationException,
+    RequiresAuthenticationException, InvalidModelException,
 )
 
 
@@ -18,10 +18,10 @@ def _get_api_key(request: StarletteRequest) -> str:
 
 
 class TokenValidator:
-    async def __call__(self, request: StarletteRequest, api_key: str = Depends(_get_api_key)) -> str:
+    async def __call__(self, request: StarletteRequest, api_key: str = Depends(_get_api_key)) -> APIKeyConfig:
         for key in settings.pyris.api_keys:
             if key.token == api_key:
-                return api_key
+                return key
         raise PermissionDeniedException
 
 
@@ -33,7 +33,7 @@ async def __call__(self, request: StarletteRequest, api_key: str = Depends(_get_
                 if body.get("preferredModel") in key.llm_access:
                     return
                 else:
-                    raise PermissionDeniedException
+                    raise InvalidModelException(str(body.get("preferredModel")))
         raise PermissionDeniedException
 
 

app/main.py

@@ -2,7 +2,9 @@
 from fastapi.responses import ORJSONResponse
 
 from app.routes.messages import router as messages_router
+from app.routes.models import router as models_router
 
 app = FastAPI(default_response_class=ORJSONResponse)
 
 app.include_router(messages_router)
+app.include_router(models_router)

app/models/dtos.py

@@ -3,12 +3,6 @@
 from datetime import datetime
 
 
-class LLMModel(str, Enum):
-    GPT35_TURBO = "GPT35_TURBO"
-    GPT35_TURBO_16K_0613 = "GPT35_TURBO_16K_0613"
-    GPT35_TURBO_0613 = "GPT35_TURBO_0613"
-
-
 class ContentType(str, Enum):
     TEXT = "text"
 
@@ -37,3 +31,9 @@ class Message(BaseModel):
 
     used_model: str = Field(..., alias="usedModel")
     message: Message
+
+
+class LLMModelResponse(BaseModel):
+    id: str
+    name: str
+    description: str

app/routes/messages.py

@@ -1,6 +1,6 @@
-from fastapi import APIRouter, Depends
 from datetime import datetime, timezone
 
+from fastapi import APIRouter, Depends
 from parsimonious.exceptions import IncompleteParseError
 
 from app.core.custom_exceptions import (
@@ -9,12 +9,10 @@
     InternalServerException,
     InvalidModelException,
 )
-from app.dependencies import PermissionsValidator
-from app.models.dtos import SendMessageRequest, SendMessageResponse, LLMModel
-from app.core.custom_exceptions import BadDataException
 from app.dependencies import TokenPermissionsValidator
 from app.models.dtos import SendMessageRequest, SendMessageResponse
 from app.services.guidance_wrapper import GuidanceWrapper
+from config import settings
 
 router = APIRouter(tags=["messages"])
 
@@ -24,7 +22,7 @@
 )
 def send_message(body: SendMessageRequest) -> SendMessageResponse:
     try:
-        model = LLMModel(body.preferred_model)
+        model = settings.pyris.llms[body.preferred_model]
     except ValueError as e:
         raise InvalidModelException(str(e))
 

app/routes/models.py

@@ -0,0 +1,18 @@
+from fastapi import APIRouter, Depends
+
+from app.dependencies import TokenValidator
+from config import settings, APIKeyConfig
+from models.dtos import LLMModelResponse
+
+router = APIRouter(tags=["models"])
+
+
+@router.get(
+    "/api/v1/models", dependencies=[Depends(TokenValidator())]
+)
+def send_message(api_key_config: APIKeyConfig = Depends(TokenValidator())) -> list[LLMModelResponse]:
+    llm_ids = api_key_config.llm_access
+
+    # Convert the result of this to a list of LLMModelResponse
+    return [LLMModelResponse(id=key, name=config.name, description=config.description) for key, config in
+            settings.pyris.llms.items() if key in llm_ids]

app/services/guidance_wrapper.py

@@ -1,14 +1,14 @@
 import guidance
 
-from app.config import settings
-from app.models.dtos import Content, ContentType, LLMModel
+from app.config import settings, LLMModelConfig
+from app.models.dtos import Content, ContentType
 
 
 class GuidanceWrapper:
     """A wrapper service to all guidance package's methods."""
 
     def __init__(
-        self, model: LLMModel, handlebars: str, parameters=None
+        self, model: LLMModelConfig, handlebars: str, parameters=None
     ) -> None:
         if parameters is None:
             parameters = {}
@@ -39,5 +39,5 @@ def query(self) -> Content:
         return Content(type=ContentType.TEXT, textContent=result["response"])
 
     def _get_llm(self):
-        llm_credentials = settings.pyris.llms[self.model].llm_credentials
+        llm_credentials = self.model.llm_credentials
         return guidance.llms.OpenAI(**llm_credentials)

application.example.yml

@@ -1,18 +1,18 @@
 pyris:
-  api_key: secret
-  llm:
-    GPT35_TURBO:
-      model:
-      token:
-      api_base:
-      api_type:
-      api_version:
-      deployment_id:
-    GPT35_TURBO_16K_0613:
-      model:
-      token:
-      chat_mode:
-    GPT35_TURBO_0613:
-      model:
-      token:
-      chat_mode:
+  api_keys:
+    - token: "secret"
+      comment: "DUMMY"
+      llm_access:
+        - DUMMY
+
+  llms:
+    DUMMY:
+      name: "Dummy model"
+      description: "Dummy model for testing"
+      llm_credentials:
+        api_base: ""
+        api_type: ""
+        api_version: ""
+        deployment_id: ""
+        model: ""
+        token: ""

application.test.yml

@@ -1,18 +1,18 @@
 pyris:
-  api_key: secret
-  llm:
-    GPT35_TURBO:
-      model:
-      token:
-      api_base:
-      api_type:
-      api_version:
-      deployment_id:
-    GPT35_TURBO_16K_0613:
-      model:
-      token:
-      chat_mode:
-    GPT35_TURBO_0613:
-      model:
-      token:
-      chat_mode:
+  api_keys:
+    - token: "secret"
+      comment: "DUMMY"
+      llm_access:
+        - DUMMY
+
+  llms:
+    DUMMY:
+      name: "Dummy model"
+      description: "Dummy model for testing"
+      llm_credentials:
+        api_base: ""
+        api_type: ""
+        api_version: ""
+        deployment_id: ""
+        model: ""
+        token: ""