Release v0.1.10 · kingeke/justcode

v0.1.10

Session stats now survive reloads, chat titles show up instantly, and requests are session-aware for better prompt caching on OpenRouter and OpenAI.

Session stats persist across reloads — token usage, cost, TTFT, tok/s, and the session average are saved with each conversation and restored when you reopen it (CLI and VS Code).
Conversations are tracked as sessions on OpenRouter, so chats appear grouped in its dashboard and get sticky routing for much better prompt-cache hit rates.
OpenAI requests carry a per-conversation cache key, steering repeated requests onto the same cache shard.
Scroll to top: a floating button in the extension and a "Jump to top" pill in the CLI, shown once you scroll up and never interrupting streaming.

Clicking a file in the extension's Changes panel opens VS Code's diff editor directly instead of an inline diff.
Session titles are generated before the first response starts and no longer trigger extended reasoning on reasoning models, which made titles slow and pricey.
The metrics footer labels the new-token counter "in" instead of "new".

The AVG tok/s no longer drops to 0 while a response streams.
Queued steering messages appear at their real position in the conversation immediately, instead of jumping above the in-flight turn (VS Code).
Cancelling a turn during title generation no longer hangs the request.
A title generated in another window is reused instead of being regenerated or overwritten.