UiPath LangChain Client [langchain-v1.8.1]

[1.8.1] - 2026-04-09

Changed

• Renamed aws optional dependency to bedrock for clarity

UiPath LLM Client [v1.8.0]

[1.8.0] - 2026-04-08

Added

• UiPathLiteLLM — provider-agnostic LLM client powered by LiteLLM
  • completion / acompletion for chat completions across all providers
  • embedding / aembedding for embeddings
  • Automatic model discovery from the UiPath backend — detects vendor, api_flavor, and model family
  • Optional vendor_type and api_flavor overrides (same pattern as LangChain factory)
  • Supports OpenAI (chat-completions + responses API), Gemini, Bedrock (invoke + converse), and Vertex AI Claude
  • All HTTP routed through UiPath httpx transport (auth, retry, headers) — no direct calls to Google/AWS/OpenAI
  • Explicit completion parameters with full IDE autocomplete
• litellm as an optional dependency (uv add uipath-llm-client[litellm])
• _strict_response_validation parameter to all Anthropic client classes

Changed

• Updated dependency versions: uipath-platform>=0.1.21, anthropic>=0.91.0, litellm>=1.83.4

UiPath LangChain Client [langchain-v1.8.0]

[1.8.0] - 2026-04-08

Added

• UiPathChatLiteLLM — LangChain chat model powered by LiteLLM, supporting all UiPath gateway providers
• langchain-litellm as an optional dependency for LiteLLM integration

Changed

• Updated dependency versions: anthropic[bedrock,vertex]>=0.91.0
• Version bump to match core package 1.8.0

UiPath LangChain Client [langchain-v1.7.1]

[1.7.1] - 2026-04-04

Added

• custom_class parameter in get_chat_model() and get_embedding_model() factory functions to allow instantiating a user-provided class instead of the auto-detected one

UiPath LLM Client [v1.7.0]

[1.7.0] - 2026-04-03

Added

• UiPathNormalizedClient — provider-agnostic LLM client with no optional dependencies
  • client.completions.create/acreate/stream/astream for chat completions
  • client.embeddings.create/acreate for embeddings
  • Structured output via response_format (Pydantic, TypedDict, dict, json_object)
  • Tool calling with dicts, Pydantic models, or callables
  • Streaming with SSE parsing
  • Full vendor parameter coverage: OpenAI (reasoning, logprobs, logit_bias), Anthropic (thinking, top_k), Google (thinking_level/budget, safety_settings, cached_content)
  • Typed response models: ChatCompletion, ChatCompletionChunk, EmbeddingResponse
  • Accepts both dict and Pydantic model messages

UiPath LLM Client [v1.6.0]

[1.6.0] - 2026-04-03

Fixed

• Set api_flavor to None for ANTHROPIC and AZURE vendor types
• Add ANTHROPIC/AZURE cases to validator and remove unused original_message parameter
• Fix VertexAI default_headers consistency and demo import path
• Fix LLMGateway singleton cache key to include base_url

UiPath LangChain Client [langchain-v1.6.0]

[1.6.0] - 2026-04-03

Fixed

• Version bump to match core package changes

UiPath LLM Client [v1.5.10]

[1.5.10] - 2026-03-26

Changed

• Removed X-UiPath-LLMGateway-AllowFull4xxResponse from default request headers to avoid PII leakage in logs

UiPath LLM Client [v1.5.9]

[1.5.9] - 2026-03-26

Fix

• Use availableOperationCodes field (instead of operationCodes) when validating BYOM operation codes

UiPath LLM Client [v1.5.8]

[1.5.8] - 2026-03-26

Fix

• Pass base_url to OpenAI and AsyncOpenAI constructors in UiPathOpenAI and UiPathAsyncOpenAI to ensure the correct endpoint is forwarded to the underlying SDK clients