Services.BgIndexingControl (fnord v0.9.29)

Session-local (per BEAM node) control plane for pausing background indexing on a per-model basis.

Why this exists

Background indexing is a convenience feature. When OpenAI starts returning throttling responses (HTTP 429 with a recognized throttling code), we can disable background indexing for the affected model(s) to reduce load.

This module is intentionally:

model-agnostic: any model string encountered can be tracked
boolean-only: models are either paused or not paused (no timers)
session-local: state is stored in Services.Globals and seeded idempotently

State

Stored in Services.Globals under the :fnord app:

:bg_indexer_paused_models => %{model => true}
:bg_indexer_throttle_counts => %{model => consecutive_throttles}
:bg_indexer_throttle_threshold => integer (default: 3)

Summary

Functions

clear_pause(model)

ensure_init()

note_success(model)

Reset the consecutive throttling count for the model back to 0.

note_throttle(model)

Increment the consecutive throttling count for the model.

pause(model)

paused?(model)

set_threshold(n)

threshold()