v0.3.8 - Adaptive memory.

v0.3.8 - Adaptive memory.

Highlights

• Added a much stronger local-memory and prompt flow for long-running sessions.
• Expanded hardware-aware setup and token/speed recommendation logic.
• Improved the local TUI with better status handling, model/setup flow, and Discord/community prompts.
• Broadened SDK support for hardware recommendations and token-per-second estimation.

Added

• Persistent memory reflection and better saved-memory integration into prompts.
• New app path helpers and broader prompt/context plumbing.
• Token estimator datasets for hardware and model recommendations.
• src/sdk/tok_s.py for throughput estimation and related SDK exports.
• Workflow/community docs note about shared .openjet workflow files in Discord.
• Small community note in the README for benchmarkers and testers.

Improved

• Setup/runtime provisioning flow, including better memory-based context sizing.
• llama_server behavior and local runtime handling.
• Slash-command and command-surface docs.
• Agent/context/session handling across the app and SDK.
• Skills/tool/device integration in command execution.
• README and PyPI README positioning and docs.

Internal / maintenance

• Large test expansion across app, context memory, SDK recommendations, setup memory, harness, device tags, and tool/device execution.
• Cleanup in device source handling and executor/tool registry plumbing.