v0.12.0

What's Changed

• fix(chat): stream compaction summaries and report failed compactions by @mudler-agent in #117
• feat(tui): footer activity strip with ctrl+g focus and per-agent chips by @mudler-agent in #118
• fix(chat): never send a request that leaves the reply no room by @mudler-agent in #121
• feat(agents): ask the model for a short title for each sub-agent by @mudler-agent in #122
• feat(tui): send messages to a running sub-agent from its log by @mudler-agent in #120
• fix(tui): ctrl+c sends the queue, ctrl+b targets foreground work, add /help by @mudler-agent in #124
• fix(tui): put the footer telemetry back at the right of the help line by @mudler-agent in #123
• feat(login): handle OAuth over SSH with port-forward hints, paste fallback, and device flow by @mudler-agent in #119
• fix(login): remove non-standard OpenAI device-code URL by @mudler-agent in #125
• fix(sessions): keep the model context and full transcript on resume by @mudler-agent in #126
• fix(login): use 127.0.0.1 as the codex OAuth redirect host by @mudler-agent in #127
• fix(sessions): keep tool-output pruning state across resume by @mudler-agent in #128
• fix(sessions): share one artifact store between tools and session by @mudler-agent in #130
• fix(login): let /login take a pasted redirect URL by @mudler-agent in #131
• feat(config): add strict_tools to send tool schemas in strict mode by @mudler-agent in #132
• chore(deps): pin cogito to the merged strict-schemas commit on main by @mudler-agent in #133

Full Changelog: v0.11.1...v0.12.0