Longcat api not available, using local catalogue

[noobix]

I have been routing to longcat ai for months and my session allocated to an api key was exhausted and I created an api key and reconfigured the connection again. For days now my routine has not worked as expected, I have learned the hard way to just edit the connection and update the api key rather than removing the connection entirely. Please the logs don't say much and I don't know what to look for to tell my provider. This is my latest log if I retest the connection and manually try to import models

[DB] cache_size changed to 16384KB
[DB] SQLite database ready: C:\Users\nanay\AppData\Roaming\omniroute\storage.sqlite (DATA_DIR=C:\Users\nanay\AppData\Roaming\omniroute, SQLITE_FILE=C:\Users\nanay\AppData\Roaming\omniroute\storage.sqlite)
(node:21360) [DEP0190] DeprecationWarning: Passing args to a child process with shell option true can lead to security vulnerabilities, as the arguments are not escaped, only concatenated.
(Use `node --trace-deprecation ...` to show where the warning was created)
 GET /api/health/ping 200 in 15ms (next.js: 5ms, proxy.ts: 5ms, application-code: 5ms)
 GET /api/health/ping 200 in 19ms (next.js: 6ms, proxy.ts: 9ms, application-code: 5ms)
 GET /api/sync/cloud 200 in 30ms (next.js: 15ms, proxy.ts: 11ms, application-code: 4ms)
 GET /api/sync/cloud 200 in 15ms (next.js: 4ms, proxy.ts: 7ms, application-code: 4ms)
 GET /api/auth/csrf 200 in 24ms (next.js: 5ms, proxy.ts: 10ms, application-code: 10ms)
[DB] Changing cache_size from 65536KB to 16384KB
[DB] cache_size changed to 16384KB
[DB] SQLite database ready: C:\Users\nanay\AppData\Roaming\omniroute\storage.sqlite (DATA_DIR=C:\Users\nanay\AppData\Roaming\omniroute, SQLITE_FILE=C:\Users\nanay\AppData\Roaming\omniroute\storage.sqlite)
[proxyLogger] Loaded 200 proxy logs from SQLite
(node:19656) [DEP0190] DeprecationWarning: Passing args to a child process with shell option true can lead to security vulnerabilities, as the arguments are not escaped, only concatenated.
(Use `node --trace-deprecation ...` to show where the warning was created)
 POST /api/providers/3f3d00cc-c8ca-4261-947e-a27df5f0e0ec/test 200 in 2.6s (next.js: 34ms, proxy.ts: 10ms, application-code: 2.5s)
 GET /api/provider-nodes 200 in 34ms (next.js: 13ms, proxy.ts: 14ms, application-code: 7ms)
 GET /api/providers 200 in 43ms (next.js: 6ms, proxy.ts: 15ms, application-code: 22ms)
 GET /api/settings/proxy?resolve=3f3d00cc-c8ca-4261-947e-a27df5f0e0ec 200 in 25ms (next.js: 5ms, proxy.ts: 10ms, application-code: 10ms)
 GET /api/providers/3f3d00cc-c8ca-4261-947e-a27df5f0e0ec/models?refresh=true 200 in 50ms (next.js: 26ms, proxy.ts: 11ms, application-code: 13ms)

[diegosouzapw]

@noobix Thanks for the log. What I can say from it: your connection test returned 200 (POST /api/providers/…/test 200), and the "Import from /models" refresh also returned 200 in 50 ms. The "API unavailable — using local catalogue" you see is expected for LongCat: our entry for it has a single built-in model (LongCat-2.0) and no model-list endpoint, so import always falls back to that local list. That isn't the problem, and it also isn't what's making your routine fail.

The log you pasted has no chat request in it, so it can't show why your routine fails. The DEP0190 line is a harmless Node warning (#15327). Could you send, from Dashboard → Logs, one failed request to LongCat (status code and the error text, key hidden), and say which model name your routine uses? Also, about "edit the key instead of removing the connection": after you recreated the connection, did the connection show a status like "unavailable"/"expired" or a cooldown time? That would tell us if the old key's failure state is being kept.

[noobix]

Please examine this log from the dashboard

omniroute-request-logs-1h-2026-10-19(1).json

[diegosouzapw]

@noobix Thanks for the export. I read it as data. It has 4 entries, all for LongCat: three model-sync calls (all 200, each syncing 1 model) and one connection test (200), all for the account "badest_crocodile". There is no chat request in it: no call to /v1/chat/completions, no failure, no error. So in that hour, nothing from your routine reached LongCat through OmniRoute, and the provider side looks healthy (key accepted, test passes).

That points away from the key and toward how the routine reaches OmniRoute. Things to check:

1. Which model name does your routine send? For LongCat it should be longcat/LongCat-2.0 (or the alias lc/LongCat-2.0). The old name from before you recreated the connection may no longer exist.
2. Does the routine point at OmniRoute's address and use an OmniRoute API key (not the LongCat key)? If you recreated API keys in OmniRoute, the old one in the routine would be rejected before it shows up as a LongCat request.
3. Try one manual request from the dashboard playground with longcat/LongCat-2.0. If that works and shows up in Logs, the provider is fine and the difference is in the routine's settings.

If the routine does send requests and they don't appear in the log, tell us which client it uses and what error it prints.

[noobix]

There is no model to prompt with in the playground.. That is what concerns me. When I send a prompt from vscode it routes to another model.

[diegosouzapw]

@noobix Thanks for the screenshot. The Playground's model box is empty ("—") and sending gives "Missing model", so the Playground doesn't have a model selected. That is separate from your VS Code routine, which does send some model name. Two things would tell us what's going on:

1. In Dashboard → Providers → LongCat, does "Available Models" list LongCat-2.0, and is it visible (not hidden with the eye icon)? Your earlier export showed the model sync returning 1 model.
2. In VS Code, which exact model name does the routine/extension send? If it sends a name that isn't one of your LongCat models (for example a combo or another provider's name), OmniRoute will route it there, which would explain "it routes to another model".

If the Playground list stays empty after you reload the page, tell us whether the dropdown is also empty with another provider selected.