Skip to content

docs: Add Architecture Deep Dives section to AGENTS.md#1444

Merged
aliasaria merged 1 commit intomainfrom
docs/agents-architecture-guide
Mar 4, 2026
Merged

docs: Add Architecture Deep Dives section to AGENTS.md#1444
aliasaria merged 1 commit intomainfrom
docs/agents-architecture-guide

Conversation

@aliasaria
Copy link
Copy Markdown
Member

Adds a new Architecture Deep Dives section to AGENTS.md that documents complex flows in the codebase. This section is intended to be expanded over time as we document more areas of the app.

First entry: Job Execution on Local Providers

Documents the full flow when a job is queued for a local provider, covering:

  1. Queueing — Router builds a ClusterConfig and enqueues via enqueue_local_launch()
  2. Serialized worker — Background asyncio worker pulls items one at a time and launches via the provider instance
  3. Local executionLocalProvider.launch_cluster() creates a per-job uv venv and runs the command via subprocess.Popen
  4. Error handling — Local providers use the queue worker's try/except; remote providers use tfl-remote-trap
  5. Plugin harness — Subprocess entry point for plugin-based jobs (training, eval)

@aliasaria aliasaria merged commit 4e642f0 into main Mar 4, 2026
3 of 4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant