AnythingLLM
Configure AnythingLLM to use AICredits as an OpenAI-compatible LLM provider for local-first chat, agents, and document workspaces.
Use this page with an AI assistant
Opens a new chat with this docs URL and the correct AICredits base URLs.
AnythingLLM is a local-first AI workspace for chat, agents, and document retrieval. Point its OpenAI-compatible LLM provider at AICredits to use multiple hosted models from one API key.
LLM Provider
In AnythingLLM, open the LLM provider settings and choose an OpenAI-compatible provider. Configure:
| Setting | Value |
|---|---|
| Provider | OpenAI-compatible / Generic OpenAI |
| Base URL | https://api.aicredits.in/v1 |
| API key | sk-your-aicredits-key |
| Model | Pick from Models |
Recommended model IDs:
openai/gpt-4o-mini
anthropic/claude-sonnet-4.5
google/gemini-2.5-flashWorkspace Defaults
After saving the provider:
- Open the workspace settings.
- Select the AICredits-backed model as the workspace chat model.
- Run a short test message.
- Check Dashboard → Usage for the request.
For document-heavy workspaces, start with a fast model for everyday Q&A and switch to a stronger model only when the answer quality needs it.
AnythingLLM may configure embedding models separately from chat models. Use AICredits for chat first, then add embeddings only if your AnythingLLM version exposes an OpenAI-compatible embedding provider.
Hostinger Deployment Notes
AnythingLLM is useful as an internal team workspace on Hostinger: document Q&A, shared prompts, and lightweight agents. For production customer-facing chat, use a dedicated AICredits key with a strict monthly cap.
Verify
Ask:
Reply with "AICredits via AnythingLLM" and nothing else.Then confirm usage in the AICredits dashboard.
Troubleshooting
Provider test fails — Confirm the base URL includes /v1.
No response from workspace — Make sure the workspace uses the AICredits-backed model, not an older local model.
Unexpected cost — Use a cheaper default model such as openai/gpt-4o-mini and reserve frontier models for specific workspaces.