[WIP] feat: implement streaming methods for chat models

[radu-mocanu]

TODO:

• implement true streaming

Development Package

• Add this package as a dependency in your pyproject.toml:

[project]
dependencies = [
  # Exact version:
  "uipath-langchain==0.0.136.dev1001970617",

  # Any version from PR
  "uipath-langchain>=0.0.136.dev1001970000,<0.0.136.dev1001980000"
]

[[tool.uv.index]]
name = "testpypi"
url = "https://test.pypi.org/simple/"
publish-url = "https://test.pypi.org/legacy/"
explicit = true

[tool.uv.sources]
uipath-langchain = { index = "testpypi" }

[Copilot]

Pull Request Overview

This PR implements streaming methods for UiPath LangChain chat models by adding _stream and _astream methods to two chat model classes. This enables real-time response streaming capabilities for both synchronous and asynchronous usage patterns.

• Adds streaming method implementations to two chat model classes
• Includes proper type annotations and documentation for the new streaming methods
• Updates version number to reflect the new functionality

Reviewed Changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/uipath_langchain/chat/models.py	Implements _stream and _astream methods for two chat model classes with proper typing and documentation
src/uipath_langchain/_cli/_runtime/_conversation.py	Adds handling for string content in message mapping
pyproject.toml	Updates version from 0.0.134 to 0.0.135

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The streaming implementation returns a single chunk instead of true streaming. Consider implementing actual streaming by making a streaming request to the API or clearly document this as a fallback implementation.

[Copilot]

The async streaming implementation also returns a single chunk instead of true streaming. This duplicates the same non-streaming behavior as the sync version.

Suggested change

	response = await self._acall(self.url, payload, self.auth_headers)
	# For non-streaming response, yield single chunk
	chat_result = self._create_chat_result(response)
	chunk = ChatGenerationChunk(
	message=AIMessageChunk(
	content=chat_result.generations[0].message.content,
	additional_kwargs=chat_result.generations[0].message.additional_kwargs,
	response_metadata=chat_result.generations[0].message.response_metadata,
	usage_metadata=chat_result.generations[0].message.usage_metadata, # type: ignore
	)
	)
	yield chunk
	# Assume _acall returns an async iterator over streamed response chunks
	async for chunk_data in self._acall(self.url, payload, self.auth_headers, stream=True):
	# Convert each streamed chunk to ChatGenerationChunk
	chat_result = self._create_chat_result(chunk_data)
	yield ChatGenerationChunk(
	message=AIMessageChunk(
	content=chat_result.generations[0].message.content,
	additional_kwargs=chat_result.generations[0].message.additional_kwargs,
	response_metadata=chat_result.generations[0].message.response_metadata,
	usage_metadata=chat_result.generations[0].message.usage_metadata, # type: ignore
	)
	)

[Copilot]

This is the third instance of duplicated non-streaming logic in streaming methods. The code pattern is repeated across multiple methods with only minor variations.

[Copilot]

Fourth instance of the same duplicated non-streaming implementation. Consider extracting this chunk creation logic into a shared helper method to reduce code duplication.