Switch to llama.cpp for LLM

Pin Zig setup action and version for MUSL release builds

Fix MUSL release builds by using cargo-zigbuild

Add local Qwen preprocessing runtime hardening and routing coverage

Remove "failed" option from workflow steps

Only send workflow ack for multi-step workflows

Remove diagnostics command

Stabilize Slack socket sync tests under CI timing

Include message thread as context

Notify user on workflow failure with dead letter queue

Add immediate workflow-start thread acknowledgment

Fix selector output JSON

Improved error reporting on failed messages

slack: recover missed messages on socket runtime reconnect

Write selector files into the artifacts dir