Merge branch 'dargilco/ai-model-inference' into trangevi/structured-o…

…utputs-java
Azure · Jan 10, 2025 · daf385d · daf385d
2 parents cba2d23 + 3409632
commit daf385d
Show file tree

Hide file tree

Showing 19 changed files with 335 additions and 229 deletions.
diff --git a/specification/ai/ModelClient/client.tsp b/specification/ai/ModelClient/client.tsp
@@ -16,20 +16,23 @@ namespace Customizations; // The actual name here doesn't matter and is here for
 @@usage(AI.Model.StreamingChatResponseMessageUpdate, Usage.output);
 @@usage(AI.Model.StreamingChatResponseToolCallUpdate, Usage.output);
 
-// Since we made all operator methods internal, we need to expliclty
-// say we still want the models they use to be public.
+// The operators need to be hidden, since we hand-write the public versions of those
+@@access(AI.Model.getChatCompletions, Access.internal);
+@@access(AI.Model.getEmbeddings, Access.internal);
+@@access(AI.Model.getImageEmbeddings, Access.internal);
+@@access(AI.Model.getModelInfo, Access.internal, "python");
+
+// Since we made all operator methods internal, we need to explicity
+// say we still want the models they use to be public, since they will be used by hand-written operator methods.
 @@access(AI.Model.ChatChoice, Access.public);
 @@access(AI.Model.ChatCompletions, Access.public);
 @@access(AI.Model.ChatCompletionsToolCall, Access.public);
 @@access(AI.Model.ChatCompletionsToolDefinition, Access.public);
-@@access(AI.Model.ChatCompletionsNamedToolSelection, Access.public);
-@@access(AI.Model.ChatCompletionsFunctionToolSelection, Access.public);
-@@access(AI.Model.ChatCompletionsResponseFormat, Access.public);
-@@access(AI.Model.ChatCompletionsResponseFormatJSON, Access.public);
-@@access(AI.Model.ChatCompletionsResponseFormatText, Access.public);
+@@access(AI.Model.ChatCompletionsNamedToolChoice, Access.public);
+@@access(AI.Model.ChatCompletionsNamedToolChoiceFunction, Access.public);
 @@access(AI.Model.ChatCompletionsToolCall, Access.public);
 @@access(AI.Model.ChatCompletionsToolDefinition, Access.public);
-@@access(AI.Model.ChatCompletionsToolSelectionPreset, Access.public);
+@@access(AI.Model.ChatCompletionsToolChoicePreset, Access.public);
 @@access(AI.Model.ChatRequestAssistantMessage, Access.public);
 @@access(AI.Model.ChatRequestMessage, Access.public);
 @@access(AI.Model.ChatRequestSystemMessage, Access.public);
@@ -39,12 +42,12 @@ namespace Customizations; // The actual name here doesn't matter and is here for
 @@access(AI.Model.ChatRole, Access.public);
 @@access(AI.Model.CompletionsFinishReason, Access.public);
 @@access(AI.Model.CompletionsUsage, Access.public);
-@@access(AI.Model.EmbeddingEncodingFormat, Access.public);
-@@access(AI.Model.EmbeddingInput, Access.public);
-@@access(AI.Model.EmbeddingInputType, Access.public);
-@@access(AI.Model.EmbeddingItem, Access.public);
-@@access(AI.Model.EmbeddingsResult, Access.public);
-@@access(AI.Model.EmbeddingsUsage, Access.public);
+@@access(AI.Model.EmbeddingEncodingFormat, Access.public, "python");
+@@access(AI.Model.ImageEmbeddingInput, Access.public, "python");
+@@access(AI.Model.EmbeddingInputType, Access.public, "python");
+@@access(AI.Model.EmbeddingItem, Access.public, "python");
+@@access(AI.Model.EmbeddingsResult, Access.public, "python");
+@@access(AI.Model.EmbeddingsUsage, Access.public, "python");
 @@access(AI.Model.ExtraParameters, Access.public, "java");
 @@access(AI.Model.FunctionCall, Access.public);
 @@access(AI.Model.FunctionDefinition, Access.public);
@@ -55,19 +58,19 @@ namespace Customizations; // The actual name here doesn't matter and is here for
 @@access(AI.Model.ChatMessageImageContentItem, Access.public);
 @@access(AI.Model.ChatMessageImageUrl, Access.public);
 @@access(AI.Model.ChatMessageImageDetailLevel, Access.public);
-@@access(AI.Model.StreamingChatCompletionsUpdate, Access.public);
-@@access(AI.Model.StreamingChatChoiceUpdate, Access.public);
-@@access(AI.Model.StreamingChatResponseMessageUpdate, Access.public);
-@@access(AI.Model.StreamingChatResponseToolCallUpdate, Access.public);
+@@access(AI.Model.StreamingChatCompletionsUpdate, Access.public, "python");
+@@access(AI.Model.StreamingChatChoiceUpdate, Access.public, "python");
+@@access(AI.Model.StreamingChatResponseMessageUpdate, Access.public, "python");
+@@access(AI.Model.StreamingChatResponseToolCallUpdate, Access.public, "python");
+@@access(AI.Model.ChatCompletionsResponseFormatJsonSchemaDefinition,
+  Access.public,
+  "python"
+);
 
-// The operators need to be hidden, since we hand-write the public versions of those to
-// 1. Add chat completions streaming (to getChatCompletions operator)
-// 2. Add hyper-params (to getChatCompletions/Embeddings/ImageEmbeddings, all clients)
-// 3. Cache model info (to getModelInfo, all clients, Python only)
-@@access(AI.Model.getChatCompletions, Access.internal);
-@@access(AI.Model.getEmbeddings, Access.internal);
-@@access(AI.Model.getImageEmbeddings, Access.internal);
-@@access(AI.Model.getModelInfo, Access.internal);
+@@clientName(AI.Model.ChatCompletionsResponseFormatJsonSchemaDefinition,
+  "JsonSchemaFormat",
+  "python"
+);
 @@clientName(AI.Model.ChatCompletionsResponseFormatJSON, "ChatCompletionsResponseFormatJson", "java");
 
 // We use shorter names in the Python client library

diff --git a/...ication/ai/ModelClient/examples/2024-05-01-preview/GetChatCompletions_MaximumSet_Gen.json b/...ication/ai/ModelClient/examples/2024-05-01-preview/GetChatCompletions_MaximumSet_Gen.json
@@ -10,57 +10,56 @@
           "role": "ChatRequestMessage"
         }
       ],
-      "frequency_penalty": 1,
+      "frequency_penalty": -2,
       "stream": true,
-      "presence_penalty": -2,
+      "presence_penalty": -1,
       "temperature": 0,
       "top_p": 0,
       "max_tokens": 0,
       "response_format": {
         "type": "ChatCompletionsResponseFormat"
       },
       "stop": [
-        "lwlqenszpaurxntgit"
+        "dcfnxrdeumnoytdaooqkbl"
       ],
       "tools": [
         {
           "type": "function",
           "function": {
-            "name": "velupowkmhiyypklqmgzzlmcjokcs",
-            "description": "dngfctkfjoypnjfikiowvpuvisga",
-            "parameters": {}
+            "name": "ikvkykzp",
+            "description": "gofxoftbpdi"
           }
         }
       ],
-      "seed": 10,
-      "model": "bcxpoxjhymqvjo"
+      "seed": 21,
+      "model": "askiizcjob"
     }
   },
   "responses": {
     "200": {
       "body": {
-        "id": "qdjwxvfotmorpdevwdmhbqnsgi",
-        "created": 16,
-        "model": "wr",
+        "id": "kgousajxgzyhugvqekuswuqbk",
+        "created": 18,
+        "model": "zjxvtpxhzhvgjrhit",
         "usage": {
-          "completion_tokens": 9,
-          "prompt_tokens": 15,
-          "total_tokens": 1
+          "completion_tokens": 19,
+          "prompt_tokens": 28,
+          "total_tokens": 16
         },
         "choices": [
           {
-            "index": 29,
+            "index": 7,
             "finish_reason": "stop",
             "message": {
               "role": "system",
-              "content": "wynvtftvlenfzzzrfmg",
+              "content": "jnsnrwblpuokzbkrzdcwubpfz",
               "tool_calls": [
                 {
-                  "id": "zlmfpvg",
+                  "id": "yrobmilsrugmbwukmzo",
                   "type": "function",
                   "function": {
-                    "name": "velupowkmhiyypklqmgzzlmcjokcs",
-                    "arguments": "zocluceldmcazefk"
+                    "name": "ikvkykzp",
+                    "arguments": "oqxvktuduomvckic"
                   }
                 }
               ]

diff --git a/...ication/ai/ModelClient/examples/2024-05-01-preview/GetChatCompletions_MinimumSet_Gen.json b/...ication/ai/ModelClient/examples/2024-05-01-preview/GetChatCompletions_MinimumSet_Gen.json
@@ -14,21 +14,21 @@
   "responses": {
     "200": {
       "body": {
-        "id": "qdjwxvfotmorpdevwdmhbqnsgi",
-        "created": 16,
-        "model": "wr",
+        "id": "kgousajxgzyhugvqekuswuqbk",
+        "created": 18,
+        "model": "zjxvtpxhzhvgjrhit",
         "usage": {
-          "completion_tokens": 9,
-          "prompt_tokens": 15,
-          "total_tokens": 1
+          "completion_tokens": 19,
+          "prompt_tokens": 28,
+          "total_tokens": 16
         },
         "choices": [
           {
-            "index": 29,
+            "index": 7,
             "finish_reason": "stop",
             "message": {
               "role": "system",
-              "content": "wynvtftvlenfzzzrfmg"
+              "content": "jnsnrwblpuokzbkrzdcwubpfz"
             }
           }
         ]

diff --git a/specification/ai/ModelClient/examples/2024-05-01-preview/GetEmbeddings_MaximumSet_Gen.json b/specification/ai/ModelClient/examples/2024-05-01-preview/GetEmbeddings_MaximumSet_Gen.json
@@ -6,27 +6,28 @@
     "extra-parameters": "error",
     "body": {
       "input": [
-        "wditkfkcn"
+        "p"
       ],
-      "dimensions": 14,
+      "dimensions": 11,
       "encoding_format": "base64",
       "input_type": "text",
-      "model": "esgcnvlwfzgrstu"
+      "model": "kwkpluujwiabfquhkaugttxut"
     }
   },
   "responses": {
     "200": {
       "body": {
+        "id": "cknxthfa",
         "data": [
           {
-            "index": 4
+            "index": 21
           }
         ],
         "usage": {
-          "prompt_tokens": 30,
-          "total_tokens": 29
+          "prompt_tokens": 4,
+          "total_tokens": 22
         },
-        "model": "rbnjxkkdrp"
+        "model": "uvrmctbnze"
       }
     }
   }

diff --git a/...ication/ai/ModelClient/examples/2024-05-01-preview/GetImageEmbeddings_MaximumSet_Gen.json b/...ication/ai/ModelClient/examples/2024-05-01-preview/GetImageEmbeddings_MaximumSet_Gen.json
@@ -7,29 +7,30 @@
     "body": {
       "input": [
         {
-          "image": "hxkwvrx",
-          "text": "qpoyljvucirvkzjhhnhpdeqkl"
+          "image": "puqkvvlvgcjyzughesnkena",
+          "text": "azrzyjsmnuefqpowpvfmyobeehqsni"
         }
       ],
-      "dimensions": 17,
+      "dimensions": 26,
       "encoding_format": "base64",
       "input_type": "text",
-      "model": "wssoguntnhwg"
+      "model": "jyb"
     }
   },
   "responses": {
     "200": {
       "body": {
+        "id": "cknxthfa",
         "data": [
           {
-            "index": 4
+            "index": 21
           }
         ],
         "usage": {
-          "prompt_tokens": 30,
-          "total_tokens": 29
+          "prompt_tokens": 4,
+          "total_tokens": 22
         },
-        "model": "rbnjxkkdrp"
+        "model": "uvrmctbnze"
       }
     }
   }

diff --git a/specification/ai/ModelClient/examples/2024-05-01-preview/GetModelInfo_MaximumSet_Gen.json b/specification/ai/ModelClient/examples/2024-05-01-preview/GetModelInfo_MaximumSet_Gen.json
@@ -7,9 +7,9 @@
   "responses": {
     "200": {
       "body": {
-        "model_name": "wzkm",
+        "model_name": "jno",
         "model_type": "embeddings",
-        "model_provider_name": "jkxwuyloxsmuhsevvytzp"
+        "model_provider_name": "ulyaphtaszwdkefpbkklnjtrhzh"
       }
     }
   }

diff --git a/specification/ai/ModelClient/examples/2024-05-01-preview/GetModelInfo_MinimumSet_Gen.json b/specification/ai/ModelClient/examples/2024-05-01-preview/GetModelInfo_MinimumSet_Gen.json
@@ -7,9 +7,9 @@
   "responses": {
     "200": {
       "body": {
-        "model_name": "wzkm",
+        "model_name": "jno",
         "model_type": "embeddings",
-        "model_provider_name": "jkxwuyloxsmuhsevvytzp"
+        "model_provider_name": "ulyaphtaszwdkefpbkklnjtrhzh"
       }
     }
   }