38d8e8e5cf
The operational _start_fleet.sh lives in a local (untracked) sandbox, so the gate + heartbeat-cadence settings weren't version-controlled anywhere. Add demo/start-fleet.example.sh: a parameterized, sanitized launcher (one agent-queue.sh run daemon per product against a live platform-service) that ships the two settings you must get right — AQ_FLEET_GATE=1 (M0 RU gate) and AQ_FLEET_LEASE_RENEW_SEC=30 (heartbeat cadence < the 90s stale threshold). No hardcoded paths/secrets; everything env-overridable. Documented in demo/README. Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
75 lines
3.3 KiB
Bash
Executable File
75 lines
3.3 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
#
|
|
# start-fleet.example.sh — reference launcher for a multi-product local fleet.
|
|
#
|
|
# Starts one detached `agent-queue.sh run` daemon (a "factory") per product, each
|
|
# routing work through the platform-service fleet coordinator. This is the tracked,
|
|
# parameterized version of the operational `_start_fleet.sh` people keep in their
|
|
# local sandbox — copy it, adjust the env vars, and run.
|
|
#
|
|
# Prereqs:
|
|
# - platform-service running on $AQ_FLEET_API (see scripts/deploy-gigafactory.sh)
|
|
# - a factory token in $FLEET_TOKEN_FILE (an admin/factory JWT for the fleet API)
|
|
# - tmux + the `longrun` helper (sourced below) for detached, logged daemons
|
|
#
|
|
# Env overrides (all optional):
|
|
# SB sandbox/state root (per-product queues live in $SB/q_<product>)
|
|
# AQ path to agent-queue.sh
|
|
# AQ_FLEET_API coordinator base URL (default http://localhost:4003/api)
|
|
# FLEET_TOKEN_FILE file holding the bearer token (default $SB/.token)
|
|
# PRODUCTS space-separated product ids (default: the ecosystem set)
|
|
# AGENT_QUEUE_MAX per-factory concurrency (default 3)
|
|
#
|
|
# Docs: ../docs/GIGAFACTORY/GIGAFACTORY_SYSTEM_OVERVIEW.md (§9 API, §14 gotchas) and
|
|
# ../docs/GIGAFACTORY/FLEET_DISPATCH_REDESIGN.md (the M0 RU gate).
|
|
set -uo pipefail
|
|
|
|
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
|
SB="${SB:-$PWD/.fleet-sandbox}"
|
|
AQ="${AQ:-$HERE/../agent-queue.sh}"
|
|
AQ_FLEET_API="${AQ_FLEET_API:-http://localhost:4003/api}"
|
|
FLEET_TOKEN_FILE="${FLEET_TOKEN_FILE:-$SB/.token}"
|
|
PRODUCTS="${PRODUCTS:-lysnrai chronomind mindlyst nomgap}"
|
|
AGENT_QUEUE_MAX="${AGENT_QUEUE_MAX:-3}"
|
|
LONGRUN_ALIAS="${LONGRUN_ALIAS:-$HERE/../../aliases/_longrun.alias}"
|
|
|
|
[ -f "$AQ" ] || { echo "agent-queue.sh not found at $AQ (set AQ=)"; exit 1; }
|
|
[ -s "$FLEET_TOKEN_FILE" ] || { echo "fleet token not found at $FLEET_TOKEN_FILE (set FLEET_TOKEN_FILE=)"; exit 1; }
|
|
TOK="$(cat "$FLEET_TOKEN_FILE")"
|
|
|
|
mkdir -p "$SB"
|
|
export LONGRUN_LOG_DIR="$SB"
|
|
# shellcheck disable=SC1090
|
|
source "$LONGRUN_ALIAS"
|
|
|
|
for p in $PRODUCTS; do
|
|
ROOT="$SB/q_$p"
|
|
longrun "gigafactory-$p" env \
|
|
AGENT_QUEUE_ROOT="$ROOT" \
|
|
AGENT_QUEUE_ENGINE=devin \
|
|
AGENT_QUEUE_MAX="$AGENT_QUEUE_MAX" \
|
|
AQ_FLEET=1 AQ_FLEET_ROUTE=1 \
|
|
AQ_FLEET_API="$AQ_FLEET_API" \
|
|
AQ_FLEET_TOKEN="$TOK" \
|
|
AQ_PRODUCT_ID="$p" \
|
|
AQ_FACTORY_ID="mac-$p" \
|
|
AQ_FLEET_GATE=1 \
|
|
AQ_FLEET_LEASE_RENEW_SEC=30 \
|
|
"$AQ" run
|
|
echo "----"
|
|
done
|
|
|
|
# Why these two matter (both verified on a live fleet):
|
|
# AQ_FLEET_GATE=1 §M0 RU gate — the run loop point-reads the cheap
|
|
# per-product queue version (GET /fleet/queue-state) and
|
|
# SKIPS the claim while nothing changed, slashing idle
|
|
# Cosmos RU. Default OFF; safe (fails open). See
|
|
# FLEET_DISPATCH_REDESIGN.md §8/§12.
|
|
# AQ_FLEET_LEASE_RENEW_SEC=30 heartbeat/renew cadence. MUST stay well under the
|
|
# coordinator's 90s stale threshold, or a healthy
|
|
# factory flaps to "stale"/"no live factory" between
|
|
# beats (the 300s default caused exactly that).
|
|
#
|
|
# Stop a factory: tmux kill-session -t gigafactory-<product>
|
|
# Tail a factory: tail -f "$SB"/longrun-gigafactory-<product>-*.log
|