Released on May 21, 2026, localmodelproxy v0.3.0 is a maintenance and compatibility update designed to refine how the proxy interacts with advanced model features and backend services. This release focuses on improving the stability of model reasoning chains and ensuring more reliable communication between the proxy and its various backends.

What’s new

Gemini Tool-Use Compatibility

One of the most significant additions in v0.3.0 is the improved support for Gemini models. The proxy now preserves “thought signatures” during tool-use interactions. When a Gemini model returns a tool call accompanied by a thought signature, localmodelproxy remembers that signature and automatically injects it into subsequent requests. This ensures that the model’s internal reasoning chain remains intact, leading to more coherent and accurate tool-use outcomes.

Refined Backend Communication

To ensure a smoother flow of data between the client and the LLM providers, we’ve implemented two key improvements to header handling:

  • Transparent Decompression: By stripping the Accept-Encoding header from forwarded requests, the proxy prevents backends from sending compressed responses that could cause compatibility issues, ensuring the client receives a readable response every time.
  • Precise Request Sizing: We’ve fixed a bug related to Content-Length headers. The proxy now accurately recalculates the request body size after any modifications (such as thought signature injection), preventing communication errors with strict backend APIs.

Why it matters

For users leveraging Gemini’s advanced tool-calling capabilities, these changes are critical. Preserving thought signatures means the model doesn’t “lose its train of thought” when interacting through the proxy, maintaining the high-level reasoning that makes these models powerful.

Additionally, the improvements to header management remove subtle friction points in backend communication. By ensuring correct request sizing and avoiding unexpected compression, localmodelproxy becomes more robust and transparent, reducing the likelihood of intermittent request failures.

Upgrade and Installation

Version v0.3.0 is a minor release and serves as a drop-in replacement for v0.2.0. There are no breaking changes, and upgrading is as simple as updating your binary or container image to the latest version.

This post was AI-generated using the model unsloth/gemma-4-31B-it-GGUF:UD-Q5_K_XL. It is based on the release v0.3.0 of the localmodelproxy repository, generated on May 21, 2026. Author: release-storyteller