Switch ingress/egress telemetry to pub/sub

[codingjoe]

No description provided.

[Copilot]

Pull request overview

This PR switches the inspector’s ingress/egress telemetry from Redis polling/time-series reads to a pub/sub-driven approach, adding live sparklines backed by an in-memory rolling buffer while keeping queue segment counts sourced from backend snapshots.

Changes:

• Introduces an in-memory TelemetryBuffer and adds inspector sparklines that refresh on a timer from pub/sub events.
• Replaces the backend telemetry polling API with queue_stats() (async counts snapshot) plus worker_telemetry() (async pub/sub stream).
• Updates the Redis backend to publish ingress/egress events via Redis pub/sub and adjusts tests accordingly.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
threadmill/inspector/telemetry.py	Adds a rolling per-second buffer to aggregate pub/sub telemetry into series/rates.
threadmill/inspector/inspector.scss	Styles new ingress/egress sparkline widgets and adjusts layout sizing.
threadmill/inspector/app.py	Wires up pub/sub listener + rolling buffer, adds sparklines, and migrates inspector refresh to async queue_stats().
threadmill/backends/redis.py	Publishes telemetry via pub/sub; adds queue_stats() + worker_telemetry() implementations.
threadmill/backends/lua/acknowledge.lua	Publishes egress telemetry on acknowledge via Redis PUBLISH.
threadmill/backends/base.py	Introduces TelemetryDirection/TelemetryEvent and replaces telemetry API with async queue_stats() + worker_telemetry().
tests/test_inspector.py	Updates inspector tests to use queue_stats() and adjusts timer expectations.
tests/backends/test_redis.py	Replaces telemetry polling tests with pub/sub publish assertions and async queue_stats() usage.
tests/backends/test_base.py	Updates base-backend contract tests for async queue_stats().
AGENTS.md	Adds docstring style guidance for mkdocs (Google style; avoid RST directives).

Comments suppressed due to low confidence (1)

threadmill/inspector/app.py:626

• If the backend telemetry stream raises an exception (network issue, decode error, etc.), the listener task will currently stop silently. Logging unexpected exceptions here would make failures visible and help avoid a “stuck at zero” UI with no clue why.

                try:
                    async for event in stream:
                        self.telemetry_buffer.record(event, now=time.monotonic())
                except asyncio.CancelledError:
                    raise