Deployment Reference
Configuration options for the self-hosted Riza Code Interpreter
Environment Variables
RIZA_API_KEY
The API key to use when talking to the Riza API. This is only needed when using custom runtimes.
RIZA_LICENSE_KEY
The license key to validate on startup. If the license is invalid, riza serve
will return an error.
RIZA_SERVE_WORKERS
The number of worker threads to run. Defaults to the number of CPU cores divided by two.
Added in rizio/code-interpreter:0.4.6
RIZA_SERVE_QUEUE_SIZE
The size of the work queue shared among workers. Defaults to the number of workers.
Added in rizio/code-interpreter:0.5.0
RIZA_SERVE_QUEUE_SEND_TIMEOUT
The amount of time (in seconds) to wait for an item to be added to the work
queue. If the timeout is reached, the server will send a 429 Too Many Requests
response. If set to 0, return 429 Too Many Requests
if the queue is full.
Defaults to 5 (seconds).
Added in rizio/code-interpreter:0.5.0
RIZA_SERVE_CONCURRENCY
The maximum number of concurrent requests to handle. When set, reqeusts will be processed in the order in which they’re receieved. By default there is no limit set.
Liveness and Readiness Probes
The /livez
and /readyz
endpoints can be used for readiness and liveness
probes
when deploying into a Kubernetes cluster.
Prometheus Metrics
The /metrics
endpoint contains the following metrics
http_requests_total
A counter for each HTTP request, grouped by {method, path, status}
.
http_requests_inflight
A gauge for the number of current HTTP requests, including liveness and readiness probes and the metrics endpoint itself.
worker_queue_size
A gauge for the size of the work queue.
num_workers
A gauge for the number of workers.
work_items_queued
A gauge for the number of work items waiting for a spot in the queue.
work_items_inflight
A gauge for the number of inflight work items, which includes items waiting for a spot in the queue.