Lynkr giving message Operational Error Service temporarily overloaded. (service_unavailable) code 503

[jcarron2]

Hi,

Thanks for putting together this very neat software!
I managed to get it running, and am using it with a local ollama running on a different machine.
I can query the ollama with curl, and get a response back OK

EG:
`jc@claude-ubuntu-2004:$ curl http://localhost:8080/health/live
{"status":"alive","timestamp":"2026-01-14T06:25:28.728Z","uptime":8.088089386}jc@claude-ubuntu-2004:$

jc@claude-ubuntu-2004:$ curl http://localhost:8080/health/live
{"status":"alive","timestamp":"2026-01-14T06:25:28.728Z","uptime":8.088089386}jc@claude-ubuntu-2004:$
jc@claude-ubuntu-2004:$ curl http://localhost:8080/v1/messages -H "Content-Type: application/json" -H "anthropic-version: 2023-06-01" -d '{
"model": "claude-3-5-sonnet-20241022",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "Hello!"}
]
}'
{"id":"msg_1768371987581","type":"message","role":"assistant","model":"claude-3-5-sonnet-20241022","content":[{"type":"tool_use","id":"call_0duj8df1","name":"AskUserQuestion","input":{"questions":"[{"header": "Greeting", "multiSelect": false, "options": [{"label": "Hello", "description": "The standard greeting."}, {"label": "Hi", "description": "A more casual greeting."}], "question": "How are you today?"}]"}}],"stop_reason":"tool_use","stop_sequence":null,"usage":{"input_tokens":1918,"output_tokens":78,"cache_creation_input_tokens":0,"cache_read_input_tokens":0}}jc@claude-ubuntu-2004:$

`

However I run into a problem after doing a query or two, where Lynkr starts printing this message out over and over , rapidly.

[2026-01-14 06:28:27.165 +0000] WARN (claude-backend): Load shedding: Heap usage exceeded threshold env: "development" heapUsedPercent: "93.32" threshold: "90.00" [2026-01-14 06:28:27.166 +0000] WARN (claude-backend): Operational error env: "development" error: { "message": "Service temporarily overloaded. Please retry after a few seconds.", "code": "service_unavailable", "statusCode": 503 } [2026-01-14 06:28:27.543 +0000] WARN (claude-backend): Operational error env: "development" error: { "message": "Service temporarily overloaded. Please retry after a few seconds.", "code": "service_unavailable", "statusCode": 503 }

Presently using model
codemate-ai/mini-coder:latest ( 2.5gb ), if I try a bigger model I don't seem to get any response before the Lynkr messages start

Below is the pasted output of a working test!
I attached the Lynkr startup text and Lynkr error as txt files to this discussion if they're of any help.

Thank you again,
Jonathan

`
╭─── Claude Code v2.1.7 ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ │ Tips for getting started │
│ Welcome back JC! │ Run /init to create a CLAUDE.md file with instructions for Claude │
│ │ Note: You have launched claude in your home directory. For the best experience, launc… │
│ ▐▛███▜▌ │ ────────────────────────────────────────────────────────────────────────────────────── │
│ ▝▜█████▛▘ │ Recent activity │
│ ▘▘ ▝▝ │ No recent activity │
│ │ │
│ Opus 4.5 · Claude Pro · xxxxxxx@gmail.com's │ │
│ Organization │ │
│ /home/jc │ │
╰─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

❯ what is 2+2

● 2 + 2 equals 4.

✻ Worked for 1m 59s

❯ list files in this directory

✽ Sautéing… (ctrl+c to interrupt · 1m 58s · ↑ 0 tokens)
⎿  Tip: Use /agents to optimize specific tasks. Eg. Software Architect, Code Writer, Code Reviewer

`
lynkr-startup.txt
lynkr-error.txt

[vishalveerareddy123]

I just fixed some issues with ollama
Also the model which you gave seems to be very slow when I ran it on my local machine
You can try using nemotron nano on ollama cloud. Its free tier is very generous and works really well

[jcarron2]

I will re clone this locally and try the new version, thank you for putting in some fixes!!

Also thanks very much for the suggestion, will give that model a try, see they have a locally hosted one at 30b/24GB - locally hosted is really what I am interested in using.

I have been using this model 'hir0rameel/qwen-claude:latest' which seems to work the best so far, I just had earlier tested with a small model to see if the messages I got were delayed induced.

The cloud option is good, seems to be the same 30b/24GB model as you can download? I will try the cloud tier to see how it works, but want to keep local llm !

[vishalveerareddy123]

Yep try the cloud models
they are super fast and works really well. On local
the problems arises both on the hardware side and also quantization