Show HN: Harbor v0.4.19 – harbor launch –back end vLLM –web codex (github.com)

Harbor has announced the launch of version 0.4.19, which introduces new command capabilities for users to spawn pre-configured local coding tools that connect to specified inference backends. This release enhances the user experience by allowing developers to easily select different backends, such as Ollama and LlamaCPP, along with specific models like Qwen3.5-4B, when launching coding tools. Notably, the update includes an option for web-based tools that integrate features like web searching and URL reading, broadening the scope of what developers can achieve with this platform. Significantly, this launch pushes the envelope in making AI-assisted coding more accessible and customizable. The introduction of an Anthropic-compatible Messages API and OpenAI Responses API compatibility enhances the versatility of Harbor's offerings, while improvements in streaming, request validation, and rate-limiting ensure a robust and efficient experience for developers. With comprehensive documentation and enhanced guides rolling out alongside these technical upgrades, Harbor 0.4.19 positions itself as a more powerful tool in the AI/ML ecosystem, catering to the evolving needs of developers in creating sophisticated agentic coding solutions.