Merge pull request #495 from nerdalert/smoll-docs

Adds api-server and model eval chat ENVs to native mode example

Author: vishnoianil

.env.native.example

@@ -12,3 +12,5 @@ NEXT_PUBLIC_TAXONOMY_ROOT_DIR=
 NEXT_PUBLIC_EXPERIMENTAL_FEATURES=false
 
 # IL_FILE_CONVERSION_SERVICE=http://localhost:8000 # Uncomment and fill in the http://host:port if the docling conversion service is running.
+# NEXT_PUBLIC_API_SERVER=http://localhost:8080 # Uncomment and point to the URL the api-server is running on. Native mode only and needs to be running on the same host as the UI.
+# NEXT_PUBLIC_MODEL_SERVER_URL=http://x.x.x.x # Used for model chat evaluation vLLM instances. Currently, server side rendering is not supported so the client must have access to this address for model chat evaluation to function in the UI. Currently ports, 8000 & 8001 are hardcoded and why it is not an option to set.

api-server/README.md

@@ -166,7 +166,7 @@ Starts a training job.
   {
     "modelName": "name-of-the-model",
     "branchName": "name-of-the-branch",
-    "epochs": 10  // Optional
+    "epochs": 10
   }
   ```
 
@@ -199,7 +199,7 @@ Combines data generation and training into a single pipeline job.
   {
     "modelName": "name-of-the-model",
     "branchName": "name-of-the-branch",
-    "epochs": 10  // Optional
+    "epochs": 10
   }
   ```
 
@@ -230,7 +230,7 @@ Serves the latest model checkpoint on port `8001`.
 
   ```json
   {
-    "checkpoint": "samples_12345"  // Optional
+    "checkpoint": "samples_12345"
   }
   ```
 
@@ -353,7 +353,7 @@ Unloads a specific VLLM container.
 
   ```json
   {
-    "model_name": "pre-train"  // Must be either "pre-train" or "post-train" for meow
+    "model_name": "pre-train"
   }
   ```
 
@@ -387,7 +387,7 @@ Fetches the status of a specific VLLM model.
 
   ```json
   {
-    "status": "running"  // Possible values: "running", "loading", "stopped"
+    "status": "running"
   }
   ```