>>> [!note] Migrated issue <!-- Drupal.org comment --> <!-- Migrated from issue #3530379. --> Reported by: [camoa](https://www.drupal.org/user/2448054) Related to !17 >>> <h3 id="summary-problem-motivation">Problem/Motivation</h3> <p>Currently, the AI Provider Anthropic module hardcodes available models in the <code>getConfiguredModels()</code> method. This approach has several limitations:</p> <ol> <li><strong>Manual maintenance required</strong>: Every time Anthropic releases new models (like Claude 4), the module needs to be manually updated</li> <li><strong>Delayed support</strong>: Users cannot access new models until maintainers update the module</li> <li><strong>Missing models</strong>: As seen with Claude 4 models (<code>claude-opus-4-20250514</code>, <code>claude-sonnet-4-20250514</code>), new models are not immediately available</li> </ol> <p>Anthropic provides a <code>/v1/models</code> endpoint that returns a list of all available models, but this capability is not being utilized.</p> <h4 id="summary-steps-reproduce">Steps to reproduce</h4> <ol> <li>Install AI Provider Anthropic module</li> <li>Configure with a valid API key</li> <li>Check available models in the UI</li> <li>Notice that Claude 4 models (<code>claude-opus-4-20250514</code>, <code>claude-sonnet-4-20250514</code>) are not listed despite being available in Anthropic's API</li> </ol> <h3 id="summary-proposed-resolution">Proposed resolution</h3> <p>Implement dynamic model detection by:</p> <ol> <li>Adding a method to query Anthropic's <code>/v1/models</code> endpoint</li> <li>Caching the results to avoid excessive API calls</li> <li>Falling back to hardcoded models if the API call fails</li> <li>Optionally allowing site administrators to refresh the model list</li> </ol> <p>The implementation would query <code>GET https://api.anthropic.com/v1/models</code> with appropriate headers and cache the results.</p> <h3 id="summary-remaining-tasks">Remaining tasks</h3> <ul> <li>Implement the <code>fetchAvailableModels()</code> method</li> <li>Add caching mechanism with configurable TTL</li> <li>Create fallback mechanism to hardcoded models</li> <li>Add error handling and logging</li> <li>Update existing <code>getConfiguredModels()</code> to use dynamic fetching</li> <li>Write tests for the new functionality</li> <li>Update documentation</li> </ul> <h3 id="summary-ui-changes">User interface changes</h3> <p>No UI changes required. Models will automatically appear in existing model selection dropdowns.</p> <p>Optional: Add a "Refresh models" button in the AI Provider configuration page to clear the cache and fetch updated models.</p> <h3 id="summary-api-changes">API changes</h3> <ul> <li>New protected method: <code>fetchAvailableModels()</code></li> <li>New protected method: <code>getHardcodedModels()</code> (refactored from current implementation)</li> <li>No changes to public API - maintains backward compatibility</li> </ul> <h3 id="summary-data-model-changes">Data model changes</h3> <ul> <li>New cache entry: <code>ai_provider_anthropic.models</code></li> <li>Possible new config option for cache TTL (default: 24 hours)</li> </ul>