Ai

ZeroClaw v0.7.3 Introduces Workspace Refactor, Config Migration, and Streaming Upgrades

ZeroClaw v0.7.3 Introduces Workspace Refactor, Config Migration, and Streaming Upgrades

ZeroClaw v0.7.3 marks the most significant release in the project’s history, centered on a deep architectural refactor and a cleaner operational model. The update moves the codebase from a monolithic structure into a focused Cargo workspace, introduces a new Config V2 schema with built-in migration support, improves provider responsiveness through OpenRouter streaming, and separates the web dashboard from the core binary for a more modular deployment path.

What Changed

The headline change in v0.7.3 is the complete split of ZeroClaw into a multi-crate Cargo workspace. More than a dozen focused crates now separate concerns across runtime, gateway, channels, tools, memory, providers, configuration, plugins, hardware, and the API layer. This is a major structural shift that should make the platform easier to maintain, test, and extend over time. It also reduces the foundation binary footprint to 6.6 MB when built with --no-default-features, which is a meaningful improvement for lean deployments and embedded use cases.

The release also ships a Config V2 schema with an automatic migration path. Users can run zeroclaw config migrate to upgrade existing configuration files in place while preserving comments, which lowers friction for existing installations. The old props command remains available for compatibility, but the project is clearly moving toward the newer zeroclaw config interface.

On the provider side, OpenRouter now supports token-by-token streaming rather than waiting for a full completion before returning output. That brings its behavior closer to native providers and should improve the interactive experience in live chat and assistant workflows where latency matters.

Another important platform change is the decoupling of the web dashboard from the main binary. By building the dashboard separately and embedding it later in the release process, ZeroClaw is continuing to separate frontend concerns from the core runtime. That usually leads to cleaner release engineering, smaller core binaries, and more flexible UI iteration.

Why It Matters

This version matters because it is not just adding features—it is resetting the project’s internal foundation for scale. The Cargo workspace split suggests ZeroClaw is maturing from an experimental all-in-one codebase into a more modular systems platform. That makes future development faster, reduces coupling between subsystems, and should improve contributor productivity across channels, providers, plugins, and runtime internals.

The Config V2 migration path is especially important for operators. Large internal refactors often create upgrade pain, but preserving comments and offering an in-place migration command shows attention to real-world administration. That lowers adoption risk for teams already running ZeroClaw in live environments.

The OpenRouter streaming improvement also has direct user-facing impact. Faster perceived response delivery is a practical quality-of-life upgrade for AI operations, especially in systems where OpenRouter is used as a routing or fallback layer across models.

Overall, v0.7.3 looks like a release focused on platform maturity: cleaner architecture, safer upgrades, better responsiveness, and more modular packaging. For teams evaluating AI orchestration infrastructure, the most important takeaway is that this version is heavily about what is new in the release itself: a foundational overhaul designed to make future ZeroClaw updates faster, cleaner, and easier to ship.

Official Source: https://github.com/zeroclaw-labs/zeroclaw/releases/tag/v0.7.3