Fix Provider And Model Issues

Provider failures can come from authentication, model availability, rate limits, network issues, local runtime state, or provider-specific settings.

Start by testing the provider outside the agent run.

Quick Checks

1. Open Settings > AI Integration.
2. Find the provider.
3. Confirm it is enabled.
4. Check credentials or login state.
5. Test the provider.
6. Confirm the selected model still exists and is available.

Then run a short manual agent message.

API Provider Issues

For API providers:

• Check the API key.
• Check provider quota and billing.
• Confirm the base URL.
• Confirm the model id.
• Check rate limits.
• Rotate the key if it may be expired or revoked.

Do not paste API keys into chats or agent instructions.

CLI Or Subscription Provider Issues

For CLI and subscription-backed providers:

• Re-run the login or setup command shown in Apprentice.
• Confirm the provider CLI is available to the runtime.
• Check whether the account session expired.
• Confirm the selected model is available for that account.

Changing provider account context can retire provider conversation continuity.

Local Runtime Issues

For LM Studio, Docker Model Runner, or Ollama:

• Start the local runtime service.
• Confirm the server URL and port.
• Confirm a model is loaded or pulled.
• Check local machine resource usage.
• Test a small prompt first.

Remember that Docker is still required for Apprentice agent execution even when the model runtime is local.

Model Quality Issues

If the provider works but the answer is poor:

• Use a stronger model.
• Narrow the prompt.
• Add clearer instructions.
• Mount only relevant folders.
• Reduce irrelevant context.
• Review Run Detail for tool behavior.

Cost Issues

If a provider is expensive:

• Set an agent budget.
• Use a cheaper model for frequent runs.
• Reduce schedule frequency.
• Lower max duration.
• Avoid broad prompts that make the agent inspect too much.

Still Failing

Collect:

• Provider name.
• Model id.
• Whether the provider test passes.
• Whether the issue affects one agent or all agents.
• Run Detail error.
• Any provider-side error message.

Use that evidence before changing unrelated agent settings.