start of backend implementation

Author: jonahjung22

packages/jupyter-ai/jupyter_ai/extension.py

@@ -53,6 +53,7 @@
 
 
 from .model_providers.model_handlers import ChatModelEndpoint
+from .model_providers.parameters_rest_api import ModelParametersRestAPI
 
 
 class AiExtension(ExtensionApp):
@@ -62,6 +63,7 @@ class AiExtension(ExtensionApp):
         (r"api/ai/chats/stop_streaming/?", InterruptStreamingHandler),
         (r"api/ai/completion/inline/?", DefaultInlineCompletionHandler),
         (r"api/ai/models/chat/?", ChatModelEndpoint),
+        (r"api/ai/model-parameters/?", ModelParametersRestAPI),
         (r"api/ai/secrets/?", SecretsRestAPI),
         (
             r"api/ai/static/jupyternaut.svg()/?",

packages/jupyter-ai/jupyter_ai/model_providers/parameters_rest_api.py

@@ -0,0 +1,52 @@
+"""
+REST API endpoint for model parameter definitions.
+"""
+from jupyter_server.base.handlers import APIHandler as BaseAPIHandler
+from tornado.web import authenticated, HTTPError
+from pydantic import BaseModel
+from typing import List, Optional
+
+from litellm.litellm_core_utils.get_supported_openai_params import get_supported_openai_params
+
+class ModelParametersResponse(BaseModel):
+    """Response model for model parameters API."""
+    #parameters: List[]
+    count: int
+    
+class ModelParametersRestAPI(BaseAPIHandler):
+    """
+    REST API for model parameters at `/api/ai/model-parameters`
+    
+    GET /api/ai/model-parameters: Returns all supported OpenAI parameters
+    GET /api/ai/model-parameters?core=true: Returns only core parameters
+    """
+    
+    @authenticated
+    def get(self):
+        """
+        Returns list of supported model parameters for OpenAI-compatible models.
+        
+        Query Parameters:
+        - core (boolean): If true, returns only core/common parameters
+        
+        These parameters work with LiteLLM across different providers.
+        """
+        try:
+            core_only = self.get_query_argument("core", default="false").lower() == "true"
+            if core_only:
+                parameters = get_core_parameters()
+            else:
+                parameters = get_supported_openai_params()
+            
+            response = ModelParametersResponse(
+                parameters=parameters,
+                count=len(parameters)
+            )
+            
+            self.set_status(200)
+            self.set_header("Content-Type", "application/json")
+            self.finish(response.model_dump_json())
+            
+        except Exception as e:
+            self.log.exception("Failed to get model parameters")
+            raise HTTPError(500, f"Internal server error: {str(e)}")