Steps:
1: From my client to Plexus:
curl -v "http://localhost:4000/v1/chat/completions" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ..." \
-d '{
"model": "deepseek-v4-flash",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant that provides concise answers."
},
{
"role": "user",
"content": "Say Hello"
}
],
"stream": false,
"temperature": 0,
"max_tokens": 1024
}'
2: Plexus configuration
I am attaching screenshots, config files follows:
/ui/config

/ui/models

With this configuration - I expect it to failover to opencode-go-1 when nvidia-nim timeout
3: Timeout in Plexus, no failover
/ui/logs

Transformed request (from /ui/debug):
{
"model": "deepseek-ai/deepseek-v4-flash",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant that provides concise answers."
},
{
"role": "user",
"content": "Say Hello"
}
],
"stream": false,
"temperature": 0,
"max_tokens": 1024,
"reasoning_effort": "none"
}
Raw response: <empty>
And, Plexus printed this into stdout:
May 22 12:50:10 p51 plexus[108369]: 2026-05-22 12:50:10 [info ]: Dispatcher: Selected API type 'chat' for model 'deepseek-ai/deepseek-v4-flash'. Reason: matched incoming request type 'chat'
May 22 12:50:10 p51 plexus[108369]: 2026-05-22 12:50:10 [info ]: Dispatching deepseek-v4-flash to nvidia-nim:deepseek-ai/deepseek-v4-flash chat <-> chat
May 22 12:50:50 p51 plexus[108369]: 2026-05-22 12:50:50 [info ]: Request 0076f2f6-01d4-4cf7-bb5b-0b091533d4ef: Upstream timeout, usage recorded as timeout
4: my client gets from Plexus HTTP 200
Curl command (from step 1) output this
PROBLEMS:
1 - no failover to opencode-go-1 happens!
2 - timeout and HTTP 200 are two different things - I would expect either failover or SOME error code from Plexus, not HTTP 200 !
Steps:
1: From my client to Plexus:
2: Plexus configuration
I am attaching screenshots, config files follows:
/ui/config

/ui/models

With this configuration - I expect it to failover to
opencode-go-1whennvidia-nimtimeout3: Timeout in Plexus, no failover
/ui/logs

Transformed request (from /ui/debug):
Raw response:
<empty>And, Plexus printed this into stdout:
4: my client gets from Plexus HTTP 200
Curl command (from step 1) output this
PROBLEMS:
1 - no failover to
opencode-go-1happens!2 - timeout and HTTP 200 are two different things - I would expect either failover or SOME error code from Plexus, not HTTP 200 !