---
title: Troubleshooting
description: Common AI extension issues and practical fixes
deprecated: true
supersededBy: ['/docs/cortex', '/docs/shield', '/docs/sentinel', '/docs/weave', '/docs/nexus']
---

## 1. "LLM manager required: register in DI or provide config.llm"

Cause:

- no compatible `aisdk.LLMManager` in container
- and extension could not build one from config

Fix:

- register an `aisdk.LLMManager` constructor before app start
- or provide complete `LLMConfiguration` and ensure manager creation succeeds

## 2. Startup falls back to memory state/vector store unexpectedly

Cause:

- configured backend creation failed during constructor execution

Fix:

- check startup logs for the original create error
- validate backend credentials and connectivity
- for vector store, remember only memory is currently implemented in factory code

## 3. Agents endpoint returns "agent factory not configured"

Cause:

- extension did not register correctly
- or controller is mounted before extension services are available

Fix:

- register extension before route setup
- start app successfully before request traffic
- verify `ai.GetAgentFactory(app.Container())` resolves

## 4. Training services not found in DI

Cause:

- `Training.Enabled` is false

Fix:

- set `ai.TrainingConfiguration{Enabled: true, ...}`
- restart app and resolve `ai.ModelTrainer` by type

## 5. Provider appears configured but model calls fail

Cause:

- `LLMConfiguration.Providers` is currently not auto-registered

Fix:

- register providers manually on `LLMManager` during app bootstrap
- verify provider registration with a simple prompt test

## 6. State is persisted but agent definitions disappear after restart

Cause:

- `AgentManager` stores definitions in in-memory maps

Fix:

- persist agent definitions in your own database/service
- recreate agents at startup from your stored metadata