Using tabby without creating an account

[fruffy]

Please describe the feature you want

Hi, we are looking into integration of tabby with Sublime-Text, but we ran into problems with the server requiring accounts and/or authentication. Is there a way to disable this workflow? We would just like to connect to the server using the Language Server Protocol.

Additional context
sublimelsp/LSP#2520 (comment)
This is the workflow we set up where we ran into this problem.

---

Please reply with a 👍 if you want this feature.

[wsxiaoys]

If you want to run Tabby without the user management feature, you can disable the entire "webserver feature" by passing the --no-webserver argument.

[jwortmann]

I have tried that, but it still does not work. It seems that tabby still expects some authorization token from tabby-agent (I assume).

tabby version 0.18.0, started via tabby.exe serve --no-webserver

It always returns an empty response to textDocument/inlineCompletion (no items).

I've tried with this Python file:

# find max element in array
def findMaxElement(arr):
    

And this is what tabby-agent logs via window/logMessage:

INFO: [TabbyAgent] Initializing...
INFO: [TabbyAgent] Applying updated config...
DEBUG: [CaCert] Loading extra certs from win-ca.
DEBUG: [TabbyAgent] Health check request: GET http://localhost:8080/v1/health. [fbaafe8f-cb03-4b93-969d-a028cdca3f24]
DEBUG: [TabbyAgent] Health check response status: 200. [fbaafe8f-cb03-4b93-969d-a028cdca3f24]
INFO: [TabbyAgent] Status changed: ready.
INFO: [TabbyAgent] Completed applying updated config.
INFO: [TabbyAgent] Initialized.
INFO: [CodeSearch] Created code search engine for recently changed files.
DEBUG: [LspServer] Collecting snippets from recently changed files...
DEBUG: [LspServer] Completed collecting snippets from recently changed files.
DEBUG: [TabbyAgent] Function providedCompletions called.
INFO: [TabbyAgent] Fetching completion...
DEBUG: [TabbyAgent] Completion request: POST http://localhost:8080/v1/completions. [a65a9c0d-9f9b-4a40-aeb3-c8bc9f9d4adb]
DEBUG: [TabbyAgent] Completion response status: 501. [a65a9c0d-9f9b-4a40-aeb3-c8bc9f9d4adb]
ERROR: [TabbyAgent] Completion request failed. [a65a9c0d-9f9b-4a40-aeb3-c8bc9f9d4adb] {
  "error": {
    "name": "HttpError",
    "status": 501,
    "statusText": "Not Implemented",
    "response": {}
  }
}
DEBUG: [TabbyAgent] Health check request: GET http://localhost:8080/v1/health. [db6fc335-85a7-49f5-8270-f0b7743056ce]
DEBUG: [CompletionCache] Updating completion cache, cache number before updating: 0
DEBUG: [CompletionCache] Updated entries number: 1
DEBUG: [CompletionCache] Completion cache updated, cache number: 1
INFO: [TabbyAgent] Completed processing completions, choices returned: 0.
DEBUG: [TabbyAgent] Health check response status: 200. [db6fc335-85a7-49f5-8270-f0b7743056ce]

[wsxiaoys]

Based on the error message you received (501 - Not Implemented), it appears that the issue may stem from not initiating Tabby with the --model parameter specified. Can you share the command you used to start Tabby?

[jwortmann]

I see. I did not pass the model explicitly, but now I've adjusted the command to start tabby via tabby.exe serve --model StarCoder-1B --no-webserver.

Now I get a timeout after 30 seconds (and therefore no items in the response). I'm using the CPU version just for testing to implement the inline completions on the client side.

INFO: [CompletionProvider] Fetching completion...
DEBUG: [TabbyApiClient] Completion request: POST http://localhost:8080/v1/completions. [d03b3e72-8d22-4faf-80be-70371abda04f]
DEBUG: [TabbyApiClient] Completion response status: 408. [d03b3e72-8d22-4faf-80be-70371abda04f]
DEBUG: [TabbyApiClient] Completion request timed out. [d03b3e72-8d22-4faf-80be-70371abda04f]

I guess I would need to use the CUDA version of tabby instead. Is the timeout for the CPU version expected?

[jwortmann]

Update: I've installed the CUDA version (...-cuda122.zip) and I can confirm that the inline completion request works correctly now:

{
    "isIncomplete": true,
    "items": [
        {
            "insertText": "max = arr[0]\n    for i in range(1, len(arr)):\n        if arr[i] > max:\n            max = arr[i]\n    return max",
            "range": {
                "start": {
                    "line": 2,
                    "character": 4
                },
                "end": {
                    "line": 2,
                    "character": 4
                }
            },
            "data": {
                "eventId": {
                    "completionId": "cmpl-8aafcc0e-812f-4cac-bac6-3bb88bc3f6fa",
                    "choiceIndex": 0
                }
            }
        }
    ]
}

Thanks for your help!