diff --git a/.env.example b/.env.example index 39fb2acc..abaac4c7 100644 --- a/.env.example +++ b/.env.example @@ -725,6 +725,15 @@ DOCREADER_TRANSPORT=grpc # (asynq.wiki_concurrency); requires a process restart. # WEKNORA_WIKI_ASYNQ_CONCURRENCY=16 +# Default per-model concurrency cap for background tasks (document ingest / +# enrichment). Shared across replicas, keyed by model ID. Only affects +# background LLM/embedding/VLM calls — interactive chat is never gated. +# 0 or negative disables the default cap (per-model overrides in model +# management still apply). Also settable in the management UI under +# Settings → System settings (model.max_concurrency); UI changes take effect +# immediately without a restart. +# WEKNORA_MODEL_MAX_CONCURRENCY=8 + # Read/write timeout (in milliseconds) the asynq client uses against Redis. # Default 500ms (writes scale to 1000ms). Bump if you see "i/o timeout" # errors during heavy multimodal counter DECRs or large batch uploads. diff --git a/docker-compose.yml b/docker-compose.yml index d2955a23..0a622401 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -194,6 +194,7 @@ services: - WEKNORA_TENANT_MAX_OWNED_PER_USER=${WEKNORA_TENANT_MAX_OWNED_PER_USER:-} - WEKNORA_ASYNQ_CONCURRENCY=${WEKNORA_ASYNQ_CONCURRENCY:-32} - WEKNORA_WIKI_ASYNQ_CONCURRENCY=${WEKNORA_WIKI_ASYNQ_CONCURRENCY:-16} + - WEKNORA_MODEL_MAX_CONCURRENCY=${WEKNORA_MODEL_MAX_CONCURRENCY:-8} - APK_MIRROR_ARG=${APK_MIRROR_ARG:-} - WEKNORA_BOOTSTRAP_SYSTEM_ADMIN_EMAIL=${WEKNORA_BOOTSTRAP_SYSTEM_ADMIN_EMAIL:-} depends_on: