Support
Frequently asked questions and contact information.
What document formats are supported?
LocalRAG! supports PDF, EPUB, DOCX, XLSX, PPTX, TXT, MD (Markdown), CSV, RTF, and HTML files. Scanned PDFs and copy-protected PDFs are also supported through OCR.
What are the available plans?
LocalRAG! offers three plans: Free (5 questions/day with Haiku), Standard (30 questions/day with Sonnet + Haiku, $4.99/month), and Pro (unlimited questions with all models including Opus, $9.99/month). You can also bring your own API key (BYOK) for free unlimited access.
How does the free trial work?
Standard and Pro plans include a 1-week free trial. After the trial ends, your subscription will automatically begin unless you cancel before the trial period expires. The Free plan is always available without a trial.
How do I cancel my subscription?
iOS: Settings > Apple ID > Subscriptions > LocalRAG! > Cancel. Android: Google Play > Profile > Payments & subscriptions > Subscriptions > LocalRAG! > Cancel. You must cancel at least 24 hours before the end of the current billing period.
Is my data private?
Yes. All document processing (text extraction, chunking, and indexing) happens entirely on your device. Only your question and relevant text snippets are sent to the AI to generate answers. Your original documents never leave your device.
Can I use my own API key?
Yes. You can enter your own Anthropic API key in Settings to use Claude directly. This gives you access to all available models and you pay Anthropic directly for usage. Your API key is stored securely in your device’s Keychain.
Which AI models are available?
Free plan: Claude Haiku 4.5. Standard plan: Claude Sonnet 4.5 and Haiku 4.5. Pro plan or BYOK: all models including Claude Opus 4. Each model offers different trade-offs between speed, accuracy, and cost.
The app is not answering correctly. What can I do?
Try re-importing the document to rebuild the search index. Make sure your question is specific and related to the document content. For scanned PDFs, ensure the OCR quality is sufficient. You can also try switching to a more capable model.
What languages are supported?
The app interface is available in English, Japanese, German, Spanish, French, Portuguese, Korean, and Chinese. You can chat with documents in any language — the AI and search pipeline work across all languages.
On-Device AI (Local LLM)
What are the limitations of on-device AI?
The on-device AI (Qwen3 4B) runs entirely on your device. Due to the smaller model size, answer quality may be lower than Claude API, especially for complex or multi-step questions. Cross-language search (e.g., asking in Japanese about English documents) requires the Search Model (E5 Multilingual) to be downloaded. Response times vary by device — typically 10 seconds to a few minutes.
Tips for better results with on-device AI?
Ask short, specific questions. Avoid multi-part questions. The on-device AI works best with single-topic queries against documents in the same language as your question. For cross-language search, make sure the Search Model is downloaded in Settings.
Why does on-device AI take longer to respond?
The local LLM runs on your device's CPU/GPU, which is much slower than cloud servers. Response times depend on your device performance, document size, and question complexity. iPhone 15 Pro and newer devices offer the best performance.