COMPATIBILITY & ERROR FIXES_

Cause: Mistyped key string, unintended leading/trailing whitespace, or revoked key credentials in Google AI Studio / OpenAI Platform.

Remedy: Navigate to AI Setup in Abhyas. Re-copy your key directly from your provider dashboard. Verify there are no whitespace characters at the start or end, and click 'Test connection'.

Cause: You have exceeded your provider's per-minute request rate (e.g. Gemini 15 RPM free tier) or exhausted prepaid billing credits (OpenAI).

Remedy: Wait 60 seconds and submit your turn again. If using OpenAI, check your account credit balance in platform.openai.com. Abhyas shows estimated token cost before each turn.

Cause: Your browser or operating system blocked microphone access for Abhyas.

Remedy: Click the lock/settings icon in your browser URL bar, ensure Microphone is set to 'Allow', and refresh the page. You can also switch to typing: voice and text are equal first-class citizens in Abhyas.

Cause: Device disk space is nearly full, or browser incognito/private mode limits IndexedDB storage.

Remedy: Open Settings in Abhyas, export a backup of your data, and click 'Delete all data' to purge old sessions. In incognito mode, Abhyas runs in memory mode which discards state upon closing.

Cause: Internet dropped during turn generation with a cloud AI provider.

Remedy: Abhyas automatically saves your partial session progress without data loss. Reconnect to the internet and click 'Retry', or run local models with Ollama for 100% offline practice.

Cause: Browser Same-Origin Policy blocked fetch to http://localhost:11434 or local server because cross-origin headers are not enabled on the server.

Remedy: For Ollama, set the environment variable OLLAMA_ORIGINS="*" and restart Ollama (Windows: set OLLAMA_ORIGINS=* && ollama serve; macOS: launchctl setenv OLLAMA_ORIGINS "*"). For LM Studio, toggle 'Enable CORS' in Server Settings. Refer to the Local Models Guide (/guide/local-models).

Cause: No process is listening on http://localhost:11434 (Ollama) or port 1234 (LM Studio), or the wrong port is entered.

Remedy: Start your local model server before clicking 'Test connection'. For Ollama, run 'ollama serve' in your terminal. For LM Studio, click 'Start Server'. Verify the endpoint matches your active port.

Cause: Your browser does not support WebGPU, or hardware graphics acceleration is disabled in your browser flags.

Remedy: Use Google Chrome 113+ or Microsoft Edge 113+ on a PC with a GPU. Verify 'Use graphics acceleration when available' is turned ON in chrome://settings/system. Alternatively, connect via Ollama or LM Studio.