Cline CLI v3.0.53 Changelog

This update is all about making the tool more reliable behind the scenes. It mostly matters to developers using Google Cloud's AI models, ensuring they don't hit annoying glitches with background tasks or chat memory after updating their software.

What's new

• New AI model supported: You can now select "Fable 5" (claude-fable-5) when using Google Cloud (Vertex AI). Because Google Cloud's pricing changes depending on where their servers are located, the tool won't guess the cost—it will just say "unknown" so it doesn't give you wrong information.
• Better custom model setup: If you type in a custom ID (a specific name code) for a Google Cloud AI model, the tool now passes that exact name through without altering it. This makes sure it connects to the right path on Google's servers.

Fixes

• Stopped reconnecting to old background helpers: When you update the main tool, there's a background helper program (called a "daemon") that keeps things running smoothly. Before, the updated tool would accidentally reconnect to an old, outdated helper still running on your computer. Now, the tool checks a "fingerprint" on the helper. If the helper is running old code, the tool shuts it down and starts a fresh, updated one instead.
• Fixed chat summaries for reasoning models: "Reasoning models" are AIs built to think through complex problems step-by-step. When chats get too long, the tool tries to summarize them to save memory (this is called "compaction"). Before, this summarizing was secretly skipped for these models because of a hidden 1,024-word limit. Now, it uses your actual settings (which default to a much larger 4,096 words) to write proper summaries, and it will warn you if a summary comes back empty.

Improvements

• The summarizer now automatically lowers its output limit if an AI model reports it can't handle that many words at once, making it much more flexible across different models.

Breaking changes

No breaking changes in this release.