Documentation
¶
Overview ¶
Command multitenant is the concurrency-model example for resolute-harness-go: one agent definition serving many tenants, where the instance id picks the tenant (per-instance system prompts via Initialize) and sessions inside an instance give independent, durably ordered conversations.
The scheduling contract on display:
- one session = one queue: submissions to the same session run strictly in admission order (head-of-line);
- different sessions — same tenant or different tenants — run concurrently.
Run it keyless (the local provider takes a fixed ~700ms per prompt, which makes the ordering visible on a stopwatch):
go run ./examples/multitenant
Then fire the demo load (four prompts: two queued into acme/support, one into acme/billing, one into globex — watch the queued one settle a beat later than everything else):
./examples/multitenant/load.sh
or by hand:
curl -s 'localhost:8489/agents/concierge/acme?wait=true' -d '{"kind":"user","body":"first in support","session":"support"}'
curl -N 'localhost:8489/agents/concierge/acme?session=support' # watch a session's records
Set GEMINI_API_KEY (and optionally MODEL) to serve every tenant with the real model; the scheduling contract is the harness's and does not change.
Source Files