AnythingLLM

AnythingLLM's OpenAI (Generic) provider is designed for OpenAI-compatible interfaces. Tokeness can be connected this way.

Preparation

Field	Value
Base URL	https://n.tokeness.dev/v1
API Key	A key created in the Tokeness console
Model	The full model name copied from the Tokeness model catalog

AnythingLLM comes in desktop, Docker, and cloud versions. Menu locations may differ, but the configuration fields are essentially the same.

Configure the LLM Provider

1. Open AnythingLLM.
2. Go to Settings.
3. Find LLM Preference or Model Provider.
4. Select OpenAI (Generic) or Generic OpenAI.
5. Set the Base URL to https://n.tokeness.dev/v1.
6. Enter your Tokeness API key.
7. Enter the full model name from the Tokeness model catalog.
8. Save.

Field mapping:

AnythingLLM field	Tokeness value
LLM Provider	OpenAI (Generic)
Base URL / API Base	https://n.tokeness.dev/v1
API Key	Your Tokeness API key
Model	A Tokeness model name

Workspace Models

AnythingLLM supports per-workspace model settings. After saving the global provider, check whether the current workspace is actually using the model you just configured.

Recommended order:

1. Save the Tokeness provider in global settings first.
2. Open the workspace settings.
3. Confirm the workspace uses the Tokeness provider.
4. Select or enter the Tokeness model name.
5. Save and reopen the chat.

If the global save succeeds but the workspace still uses the old model, check the workspace-level override settings.

Knowledge Base and Embedding Model

AnythingLLM treats chat models and vector models as two separate configuration categories. After connecting a chat model, if you want to upload documents and do Q&A, you also need to configure embedding.

Recommended order:

1. Get standard chat working first.
2. Then configure the embedding model.
3. Upload a small document and test retrieval.
4. Only after that, import a larger knowledge base.

Do not import a large number of files right away. It makes troubleshooting slow and makes it hard to distinguish chat model issues from vector model issues.

Verification

1. Create a test workspace in AnythingLLM.
2. Send "reply with ok only".
3. Open the Tokeness usage logs and confirm a request arrived.
4. Upload a small text file.
5. Ask a question whose answer can be found directly in the file.

Troubleshooting

Symptom	Action
Chat still uses the old model after saving the provider	Check whether the workspace overrides the global model
401	Check the API key
404	The Base URL should be https://n.tokeness.dev/v1
model not found	Copy the full model name from the Tokeness model catalog
Chat works but document Q&A does not	Configure the embedding model separately
Responses are very slow	Test with a short question first, then check the model, context length, and document size

References

• AnythingLLM OpenAI Generic LLM
• AnythingLLM Configuration