Troubleshooting

This page covers common issues encountered when using AI Core and how to resolve them. Issues are organized by category for quick reference.

Provider and connection issues

“No active provider configured” error

Symptom: The AI chat panel shows an error about no provider being available.

Cause: No LLM provider is configured or all configured providers are inactive.

Solution:

  1. Navigate to AI Core ‣ Configuration ‣ AI Providers.

  2. Ensure at least one provider is listed and marked as Active.

  3. If no providers exist, create one following the Provider Setup guide.

“Provider connection failed” or timeout

Symptom: The AI returns an error about failing to reach the provider.

Causes and solutions:

  1. Invalid API key. Click Test Connection on the provider to verify. If the test fails, check that the key is correct and has not expired.

  2. Provider outage. Check the provider’s status page:

  3. Network issues. Verify your Progrid server can reach the provider’s API endpoint. Ensure api.groq.com, api.openai.com, or api.anthropic.com are not blocked by a firewall.

  4. All providers failed. If you have multiple providers configured, all of them failed in sequence. Check the Last Error field on each provider for specific error messages.

Rate limit errors

Symptom: The AI responds with “Rate limit exceeded. Please wait before sending another message.”

Cause: You have exceeded the per-minute or per-hour message limit.

Solution: Wait for the cooldown period to expire. If rate limits are consistently hit, ask your administrator to increase the limits in Settings ‣ AI Core ‣ Rate Limiting.

Access and permission issues

“AI is not enabled” message

Symptom: The chat panel shows that the AI assistant is disabled.

Cause: AI Core has been disabled globally in the settings.

Solution: Ask your system administrator to enable the AI Assistant toggle in Settings ‣ AI Core.

“Access denied” for a model

Symptom: The AI says it cannot access a specific model (e.g., “I don’t have access to the invoices model”).

Cause: The model is not in the AI whitelist, or the required operation (read/write/create) is not enabled for that model.

Solution:

  1. Navigate to AI Core ‣ Configuration ‣ Model Access.

  2. Check if the model is listed. If not, add it.

  3. Verify the required operation toggles are enabled.

  4. Click Save.

See also

Model Access Control

“I cannot modify that field” response

Symptom: The AI says it cannot read or write a specific field.

Cause: The field is in the model’s blacklist or the global default blacklist.

Solution: If the field should be accessible, remove it from the Field Blacklist in the model access configuration. Be cautious about removing security-sensitive fields like passwords or tokens.

Chat panel not opening

Symptom: Pressing Ctrl+Space does nothing.

Causes and solutions:

  1. User not in AI group. Verify the user is assigned the AI User group at Settings ‣ Users & Companies ‣ Users.

  2. Keyboard shortcut conflict. Another browser extension or application may be capturing Ctrl+Space. Try disabling other extensions temporarily.

  3. Module not installed. Verify progrid_ai_core is installed at Settings ‣ Apps.

Conversation issues

AI gives incorrect or generic answers

Symptom: The AI provides answers that do not reflect your actual data.

Causes and solutions:

  • Model not whitelisted. The AI may be guessing instead of querying if the relevant model is not in the whitelist. Add the model following Model Access Control.

  • Knowledge sources outdated. If the AI references incorrect procedures, update the knowledge base at AI Core ‣ Configuration ‣ Knowledge Sources.

  • Model metadata stale. After adding fields or models, regenerate metadata by clicking Refresh All Metadata in the knowledge sources view.

AI stops responding mid-conversation

Symptom: The AI sends a partial response or stops mid-sentence.

Cause: The LLM provider may have reached its max token limit or experienced a timeout.

Solution:

  1. Try sending another message to continue the conversation.

  2. If the issue persists, start a new conversation (Ctrl+Space and click the new conversation button).

  3. Check if the provider’s Max Tokens setting is too low in AI Core ‣ Configuration ‣ AI Providers. Increase it if needed.

Confirmation prompt does not appear

Symptom: The AI executes write operations without asking for confirmation.

Cause: The confirmation requirement has been disabled either globally or for the specific model.

Solution:

  1. Check the global setting in Settings ‣ AI Core ‣ Security & Guardrails ‣ Require Confirmation.

  2. Check the per-model setting in the model access configuration.

  3. Enable Require Confirmation where needed.

Audit and monitoring

Audit logs growing too large

Symptom: The progrid.ai.audit.log table is consuming significant database storage.

Cause: The default retention period (90 days) may be too long for high-usage environments.

Solution:

  1. Navigate to Settings ‣ AI Core ‣ Audit Logs.

  2. Reduce the Retention Days value.

  3. Click Cleanup Now to immediately remove logs older than the new retention period.

Cannot find specific AI actions in audit logs

Symptom: You need to find what the AI did for a specific user or model.

Solution: Navigate to AI Core ‣ Configuration ‣ Audit Logs and use the search filters:

  • Today or This Week for date filtering

  • Group By –> User to see actions by user

  • Group By –> Model to see actions by model

  • Errors filter to see only failed actions

Getting help

If none of the above solutions resolve your issue:

  1. Check the audit logs at AI Core ‣ Configuration ‣ Audit Logs for error details.

  2. Review the provider status at AI Core ‣ Configuration ‣ AI Providers.

  3. Check the Progrid server logs for detailed error traces.

See also