requests to brain don't work

[ssalikoff]

Hello!
I have deployed Quivr according to the instructions and entered the ChatGPT API key in .env file
I created a brain and added a text document to it.
However, when I make a request to this brain, I receive an error:
TypeError: NetworkError when attempting to fetch resource.
And in the console output, I see the message:
backend-api | UnauthorizedError: status_code: 401, body: {'message': 'invalid api token'}
At the same time, requests to the GPT-4 model are working.
In my OpenAI account, I see that the tokens are being spent.
Could you please advise me on how to resolve this issue?

[dosubot[bot]]

Hey there, @ssalikoff! I'm here to help you with any bugs, questions, or contributions. Let's tackle this together!

To resolve the "UnauthorizedError: status_code: 401, body: {'message': 'invalid api token'}" issue, ensure the following:

1. Verify API Token: Make sure the API token used for the brain request is correct and hasn't expired. This token is different from the ChatGPT API key.

2. Check Permissions: Ensure the user associated with the API token has the necessary permissions (Viewer, Editor, Owner) for the brain.

3. Validate Authorization: Use the validate_authorization function to confirm the user has the appropriate roles for the brain ^[1].

If these steps don't resolve the issue, you might need to investigate the token generation and assignment process further.

To continue talking to Dosu, mention @dosu.

[ssalikoff]

I don't understand what this is about:

Make sure the API token used for the brain request is correct and hasn't expired. This token is different from the ChatGPT API key.

What other token is being referred to? Where can I get it and where should I enter it?"

[dosubot[bot]]

The API token for brain requests is different from the ChatGPT API key. This token is used specifically for accessing and interacting with the "brains" in Quivr.

Here's how you can manage it:

1. Storage and Retrieval: API tokens are stored in a database table named "api_keys". You can create and retrieve these tokens using the create_api_key and get_active_api_key methods ^[1].

2. Assignment: The process for generating and assigning these tokens wasn't detailed in the available information, so you might need to check the system's documentation or configuration for specifics on how to generate and assign them.

3. Usage: Once you have the correct API token, ensure it's included in your requests to the brain. This is separate from the ChatGPT API key, which is used for accessing OpenAI's services.

If you need to generate a new token, you might need to look into the system's admin interface or API for creating and managing these tokens.

To continue talking to Dosu, mention @dosu.

[ssalikoff]

@dosu Nothing is clear. What specific steps should I take? What documentation is being referred to? Where can I find it?

[dosubot[bot]]

To resolve the issue with the API token for brain requests in Quivr, follow these steps:

1. Generate an API Token:

   • Use the create_api_key method in the Quivr codebase to generate a new API token. This can be found in the api_keys.py file ^[1].

2. Assign the Token to a User:

   • Ensure the token is associated with the correct user. This involves using the handleCreateClick function to create the API key and update the user's identity if necessary ^[2].

3. Check Token Storage:

   • Verify that the token is stored correctly in the strapi_api_tokens collection with the necessary attributes like name, accessKey, and permissions ^[3].

4. Use the Token in Requests:

   • Include the generated API token in your requests to the brain. This is done by using the AuthBearer class to validate the token in the request headers ^[4].

Unfortunately, specific documentation for these steps wasn't found in the repository. You might need to explore the codebase or any available admin interfaces for more details on managing API tokens.

To continue talking to Dosu, mention @dosu.

[ssalikoff]

So, in order to use Quivr, do I need to be a Python developer myself?

[xXMrNidaXx]

Debugging brain request failures:

1. Check brain exists

curl http://localhost:5050/brains
# Should list your brains

2. Verify brain ID

# Make sure you're using the right UUID
brain_id = "your-brain-uuid-here"

3. Check API logs

docker logs quivr-backend-1
# Look for error messages

4. Test with curl

curl -X POST http://localhost:5050/chat \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKEN" \
  -d '{"brain_id": "...", "question": "test"}'

Common issues:

• Brain not indexed yet
• Missing auth token
• Incorrect brain_id format
• Backend not connected to vector DB

5. Check Supabase/Qdrant

docker logs quivr-supabase-1
docker logs quivr-qdrant-1

We debug RAG systems at RevolutionAI. Usually it's an auth or indexing issue.

What error message do you see?