Impove function types and address PM requirements (Azure#38356)

* PM comment: remove leading empty lines and spaces for string prompt template

* Improve types for complete function

* Fix pylint error

* Fix types for messages

* Update release log and docstring

* Fix linter issue

Author: YusakuNo1

sdk/ai/azure-ai-inference/CHANGELOG.md

@@ -8,6 +8,7 @@
   * Method `AIInferenceInstrumentor().instrument()` updated with an input argument `enable_content_recording`.
   * Calling `AIInferenceInstrumentor().instrument()` twice no longer results in an exception.
   * Added method `AIInferenceInstrumentor().is_content_recording_enabled()`
+* Support [Prompty](https://github.com/microsoft/prompty) and prompt template from string. PromptTemplate class outputs an array of dictionary with OpenAI compatible message format.
 
 ### Bugs Fixed
 

sdk/ai/azure-ai-inference/azure/ai/inference/_patch.py

@@ -265,7 +265,7 @@ def __init__(
     def complete(
         self,
         *,
-        messages: List[_models.ChatRequestMessage],
+        messages: Union[List[_models.ChatRequestMessage], List[Dict[str, Any]]],
         stream: Literal[False] = False,
         frequency_penalty: Optional[float] = None,
         presence_penalty: Optional[float] = None,
@@ -288,7 +288,7 @@ def complete(
     def complete(
         self,
         *,
-        messages: List[_models.ChatRequestMessage],
+        messages: Union[List[_models.ChatRequestMessage], List[Dict[str, Any]]],
         stream: Literal[True],
         frequency_penalty: Optional[float] = None,
         presence_penalty: Optional[float] = None,
@@ -311,7 +311,7 @@ def complete(
     def complete(
         self,
         *,
-        messages: List[_models.ChatRequestMessage],
+        messages: Union[List[_models.ChatRequestMessage], List[Dict[str, Any]]],
         stream: Optional[bool] = None,
         frequency_penalty: Optional[float] = None,
         presence_penalty: Optional[float] = None,
@@ -344,7 +344,7 @@ def complete(
          Typical usage begins with a chat message for the System role that provides instructions for
          the behavior of the assistant, followed by alternating messages between the User and
          Assistant roles. Required.
-        :paramtype messages: list[~azure.ai.inference.models.ChatRequestMessage]
+        :paramtype messages: list[~azure.ai.inference.models.ChatRequestMessage] or list[dict[str, Any]]
         :keyword stream: A value indicating whether chat completions should be streamed for this request.
          Default value is False. If streaming is enabled, the response will be a StreamingChatCompletions.
          Otherwise the response will be a ChatCompletions.
@@ -473,7 +473,7 @@ def complete(
         self,
         body: Union[JSON, IO[bytes]] = _Unset,
         *,
-        messages: List[_models.ChatRequestMessage] = _Unset,
+        messages: Union[List[_models.ChatRequestMessage], List[Dict[str, Any]]] = _Unset,
         stream: Optional[bool] = None,
         frequency_penalty: Optional[float] = None,
         presence_penalty: Optional[float] = None,
@@ -507,7 +507,7 @@ def complete(
          Typical usage begins with a chat message for the System role that provides instructions for
          the behavior of the assistant, followed by alternating messages between the User and
          Assistant roles. Required.
-        :paramtype messages: list[~azure.ai.inference.models.ChatRequestMessage]
+        :paramtype messages: list[~azure.ai.inference.models.ChatRequestMessage] or list[dict[str, Any]]
         :keyword stream: A value indicating whether chat completions should be streamed for this request.
          Default value is False. If streaming is enabled, the response will be a StreamingChatCompletions.
          Otherwise the response will be a ChatCompletions.

sdk/ai/azure-ai-inference/azure/ai/inference/prompts/_patch.py

@@ -15,6 +15,7 @@
 from ._core import Prompty
 from ._mustache import render
 from ._prompty_utils import load, prepare
+from ._utils import remove_leading_empty_space
 
 
 class PromptTemplate:
@@ -105,7 +106,8 @@ def create_messages(self, data: Optional[Dict[str, Any]] = None, **kwargs) -> Li
             parsed = prepare(self.prompty, data)
             return parsed
         elif "prompt_template" in self._config:
-            system_prompt = render(self._config["prompt_template"], data)
+            prompt_template = remove_leading_empty_space(self._config["prompt_template"])
+            system_prompt = render(prompt_template, data)
             return [{"role": "system", "content": system_prompt}]
         else:
             raise ValueError("Please provide valid prompt template")

sdk/ai/azure-ai-inference/azure/ai/inference/prompts/_utils.py

@@ -4,9 +4,10 @@
 # ------------------------------------
 # mypy: disable-error-code="import-untyped,return-value"
 # pylint: disable=line-too-long,R,wrong-import-order,global-variable-not-assigned)
+import json
 import re
+import sys
 import yaml
-import json
 from typing import Any, Dict, Union
 from pathlib import Path
 
@@ -72,3 +73,37 @@ def parse(contents):
         "body": body,
         "frontmatter": fmatter,
     }
+
+
+def remove_leading_empty_space(multiline_str: str) -> str:
+    """
+    Processes a multiline string by:
+    1. Removing empty lines
+    2. Finding the minimum leading spaces
+    3. Indenting all lines to the minimum level
+
+    :param multiline_str: The input multiline string.
+    :type multiline_str: str
+    :return: The processed multiline string.
+    :rtype: str
+    """
+    lines = multiline_str.splitlines()
+    start_index = 0
+    while start_index < len(lines) and lines[start_index].strip() == "":
+        start_index += 1
+
+    # Find the minimum number of leading spaces
+    min_spaces = sys.maxsize
+    for line in lines[start_index:]:
+        if len(line.strip()) == 0:
+            continue
+        spaces = len(line) - len(line.lstrip())
+        spaces += line.lstrip().count("\t") * 2  # Count tabs as 2 spaces
+        min_spaces = min(min_spaces, spaces)
+
+    # Remove leading spaces and indent to the minimum level
+    processed_lines = []
+    for line in lines[start_index:]:
+        processed_lines.append(line[min_spaces:])
+
+    return "\n".join(processed_lines)

sdk/ai/azure-ai-inference/samples/sample_chat_completions_from_input_prompt_string.py

@@ -23,7 +23,6 @@
         `your-azure-region` is the Azure region where your model is deployed.
     2) AZURE_AI_CHAT_KEY - Your model key (a 32-character string). Keep it secret.
 """
-# mypy: disable-error-code="union-attr,arg-type"
 # pyright: reportAttributeAccessIssue=false
 
 
@@ -73,7 +72,7 @@ def sample_chat_completions_from_input_prompt_string():
     messages = prompt_template.create_messages(input=input, rules=rules, chat_history=chat_history)
 
     client = ChatCompletionsClient(endpoint=endpoint, credential=AzureKeyCredential(key))
-    response = client.complete(messages=messages)  # type: ignore[reportCallIssue, reportArgumentType]
+    response = client.complete(messages=messages)
 
     print(response.choices[0].message.content)
 

sdk/ai/azure-ai-inference/samples/sample_chat_completions_from_input_prompty.py

@@ -23,7 +23,6 @@
         `your-azure-region` is the Azure region where your model is deployed.
     2) AZURE_AI_CHAT_KEY - Your model key (a 32-character string). Keep it secret.
 """
-# mypy: disable-error-code="union-attr"
 # pyright: reportAttributeAccessIssue=false
 
 

sdk/ai/azure-ai-inference/tests/test_prompts_utils.py

@@ -0,0 +1,26 @@
+from azure.ai.inference.prompts._utils import remove_leading_empty_space
+
+
+def test_success_with_no_changes():
+    prompt_str = """First line
+Second line"""
+    result = remove_leading_empty_space(prompt_str)
+    assert result == prompt_str
+
+
+def test_success_by_remove_leading_empty_space():
+    prompt_str = """
+    
+    First line
+
+      Second line
+        Third line
+"""
+    result = remove_leading_empty_space(prompt_str)
+    assert (
+        result
+        == """First line
+
+  Second line
+    Third line"""
+    )