Updated the ERI v1 specification & fixed spelling of the UNKNOWN role (#364)

app/MindWork AI Studio/Chat/ChatThread.cs

@@ -210,7 +210,7 @@ public sealed record ChatThread
                     ChatRole.SYSTEM => Role.SYSTEM,
                     ChatRole.NONE => Role.NONE,
                     
-                    _ => Role.UNKNOW,
+                    _ => Role.UNKNOWN,
                 },
                 
                 Content = contentData,

app/MindWork AI Studio/Tools/ERIClient/DataModel/Role.cs

@@ -6,7 +6,7 @@ namespace AIStudio.Tools.ERIClient.DataModel;
 public enum Role
 {
     NONE,
-    UNKNOW,
+    UNKNOWN,
     
     SYSTEM,
     USER,

app/MindWork AI Studio/wwwroot/changelog/v0.9.38.md

@@ -0,0 +1,2 @@
+# v0.9.38, build 213 (2025-03-xx xx:xx UTC)
+- Updated the ERI v1 specification for the ERI server assistant & fixed spelling of the `UNKNOWN` role.

app/MindWork AI Studio/wwwroot/specs/eri/v1.json

@@ -2,7 +2,7 @@
   "openapi": "3.0.1",
   "info": {
     "title": "ERI - (E)xternal (R)etrieval (I)nterface",
-    "description": "This API serves as a contract between LLM tools like AI Studio and any external data sources for RAG\n(retrieval-augmented generation). The tool, e.g., AI Studio acts as the client (the augmentation and\ngeneration parts) and the data sources act as the server (the retrieval part). The data\nsources implement some form of data retrieval and return a suitable context to the LLM tool.\nThe LLM tool, in turn, handles the integration of appropriate LLMs (augmentation & generation).\nData sources can be document or graph databases, or even a file system, for example. They\nwill likely implement an appropriate retrieval process by using some kind of embedding.\nHowever, this API does not inherently require any embedding, as data processing is\nimplemented decentralized by the data sources.",
+    "description": "This API serves as a contract between LLM tools like AI Studio and any external data sources for RAG\n(retrieval-augmented generation). The tool, e.g., AI Studio acts as the client (the augmentation and\ngeneration parts) and the data sources act as the server (the retrieval part). The data\nsources implement some form of data retrieval and return a suitable context to the LLM tool.\nThe LLM tool, in turn, handles the integration of appropriate LLMs (augmentation & generation).\nData sources can be document or graph databases, or even a file system, for example. They\nwill likely implement an appropriate retrieval process by using some kind of embedding.\nHowever, this API does not inherently require any embedding, as data processing is\nimplemented decentralized by the data sources.\n\nThe client expects that all fields in the JSON responses from an ERI server are named according\nto camel case or Pascal case conventions. The client's JSON objects for requests use camel case\nfor the field names.",
     "version": "v1"
   },
   "paths": {
@@ -494,7 +494,7 @@
       "Role": {
         "enum": [
           "NONE",
-          "UNKNOW",
+          "UNKNOWN",
           "SYSTEM",
           "USER",
           "AI",