Troubleshooting the AI Center in rConfig V8

Most AI Center problems come down to one of four things: the feature flag, the terms, the active provider, or the provider itself. Work through the checks below in order; each one rules out a layer.

Quick checks

Check	Where	Fix
Is AI Center in the left sidebar?	Any page	No: set `AI_CENTER_ENABLED=true` in `.env` and run `php artisan rconfig:clear-all`.
Does opening the AI Center show the terms page?	AI Center	Yes: accept the terms. Each user accepts once.
Does AI Settings show No AI provider is activated?	AI Center → AI Settings	Click Activate on a provider.
Does Test on the provider card succeed?	AI Center → AI Settings	No: see Provider errors.
Is the capability switched on?	AI Center → AI Configuration	No: turn it on and reload the page.

AI buttons are missing

AI buttons (Ask AI, Analyze with AI, Narrate with AI, AI Generate Rules, AI Assistant) are hidden, not greyed out, when they can’t work. A button is missing when any of these is true:

AI_CENTER_ENABLED isn’t true on the server.
You haven’t accepted the AI Center terms.
That capability is off in AI Configuration.
You don’t have permission for the page or device itself.

Fix the cause, then reload the page; capability state is loaded when the page opens.

The AI Center returns a 403 page

For everyone: AI_CENTER_ENABLED is off, or the change hasn’t been picked up. Run php artisan rconfig:clear-all after editing .env.
For one user: their role doesn’t include aicenter.view. See Roles.

“No AI provider is activated.”

Shown in chat panels, narrations, and AI Settings when providers exist but none is active, or no provider exists. Saving a provider doesn’t activate it.

Admins see a link: Open AI Center → Settings to add and activate a provider.
Other users see Ask an administrator to activate an AI provider in AI Center → Settings., and the chat input reads No AI provider activated — messages cannot be sent.

Provider errors

These messages come from the provider, through rConfig. They appear in chat replies, narration panels, and the Test toast.

Message	Likely cause	Fix
Authentication with the AI provider failed. The configured API key appears to be invalid or missing.	Wrong, revoked, or missing API key	Edit the provider and paste a new key.
The configured AI credentials are not permitted to perform this request.	Key lacks access to the model or project	Check the key’s permissions in the provider console.
The configured AI model could not be found. Please verify the model name.	Model name typo, or a model your account can’t use	Correct the Model field. For Ollama, match `ollama list` exactly; for Azure, use the deployment name.
The AI provider is receiving too many requests right now. Please wait a moment and try again.	Provider rate limit	Wait and retry, or raise your limits with the provider.
The AI provider account has insufficient credits or quota available.	Out of credit or quota	Top up or raise the quota in the provider console.
The AI provider is temporarily overloaded. Please try again shortly.	Provider-side capacity issue	Retry later.
The request is too large for the AI provider to process. Try a smaller section.	Config or question too big for the model’s context window	Select a smaller block of config, or use a model with a larger context window.
The AI provider could not be reached. Please check the connection details and try again.	Network, DNS, firewall, or wrong base URL	From the rConfig server, run `curl <base-url>` (for Ollama, `curl http://<host>:11434/api/tags`). For Azure, check the provider’s Endpoint (or `AZURE_OPENAI_URL` in `.env` for providers saved before V8.4.2).
The AI provider took too long to respond. Please try again.	Slow model, often a large local model on CPU	Use a smaller or faster model, or better hardware.
The AI provider returned an empty response. Please try again.	Model returned nothing	Retry; if it repeats, try another model.

In chat panels, a provider rate limit may instead show as Rate limit exceeded. Please wait a moment before sending another message.

Generic failures

Message	Where	What to do
Sorry, I encountered an error. Please try again.	Chat panels	Usually the capability was switched off, or your terms acceptance is missing, since the page loaded. Reload the page. If it persists, test the provider on AI Settings.
Analysis failed. Please try again.	Drift AI Analysis panel	Reload and retry. Also check the capability is still on and the config files exist.
Failed to generate narration.	Compliance narration	Reload and retry, then test the provider.
Failed to generate rules.	AI Generate Rules	Make sure the definition has a name, then retry.
Authentication required. Please refresh the page and try again.	Chat panels	Your rConfig session expired. Reload and log in.

Local models (Ollama)

Works from your laptop, fails from rConfig: Ollama listens on 127.0.0.1 by default. Start it with OLLAMA_HOST=0.0.0.0 and open port 11434 to the rConfig server.
Very slow or times out: CPU-only inference on a large model is slow. Use a smaller model, or run Ollama on a machine with a GPU.
Answers are shallow or off-topic: small models struggle with long configs. Ask about a selected block rather than the whole file, or use a larger model.

MCP connections

Symptom	Fix
Generic “failed to connect”	Usually certificate trust with a self-signed or internal-CA certificate. Trust the certificate in your client (for Claude Code, `NODE_EXTRA_CA_CERTS`).
`401 Unauthorized`	The token expired (check its Expires column), was revoked, was mistyped, or is a REST API token. Create an MCP token on the MCP Server tab.
No New MCP Token button	Your role doesn’t have `aicenter.mcptoken`. Ask an admin to add it, or to create a token for you under Token for.
Can’t pick a user under Token for	You can only create tokens for users you manage: everyone if you’re an Admin, otherwise users whose role ranks below yours.
`405` from a curl test	You sent a GET. The endpoint only accepts POST; see Verify the connection.
No devices returned	The token owner’s role can’t see those devices.
Hosted agent can’t connect	Cloud-hosted agents can’t reach an internal-only address. Use a locally run client.

Still stuck?

Collect the exact error message, the provider type and model, and the time it happened, and contact support@rconfig.com. The AI audit trail (GET /api/ai/audit-logs) and the Laravel log at storage/logs/laravel.log show whether the request reached the provider.