Shipping a multilingual Flutter AI therapist: Gemini first, Claude as fallback
I shipped Safe Space , a Flutter app for everyday mental-health self-help: AI chat, a mood journal, CBT-style thought records, PHQ-9 / GAD-7 screens, and breathing tools. It is not therapy , not a medical device, and not a crisis line. That disclaimer lives in the UI, the backend, and the public site. This is a build log, not a ranking of other apps. Stack that actually shipped Client : Flutter…
I released Safe Space, a Flutter app designed for everyday mental-health self-help. The app features AI chat, a mood journal, Cognitive Behavioral Therapy (CBT) style thought records, PHQ-9 and GAD-7 screening tools, and breathing exercises. It's important to note that Safe Space is not a therapy service, medical device, or crisis line; users are reminded of this in the user interface, backend, and public website.
Technically, the app utilizes Flutter for both Android and web platforms (with iOS in development). Authentication is managed through Firebase Authentication. The AI component utilizes Google Gemini on the backend, with Anthropic Claude as a fallback if Gemini is unavailable. The app supports eight languages: English, Russian, Turkish, Spanish, German, Portuguese, Japanese, and Korean.
Key aspects of the app's development include:
1. Consent and transparency: Users are presented with a consent step before interacting with the language model (LLM). If they skip the consent step, the app does not send a message to the LLM. This straightforward approach emphasizes honesty and user trust.
2. Graceful fallback: The backend attempts to use Gemini first for LLM interactions. In the event of a failure, it gracefully falls back to Claude. The user experience remains uninterrupted, with a plain error message displayed if both providers are down.
3. Screening vs. diagnosis: The app incorporates PHQ-9 and GAD-7 questionnaires, which are well-known mental health screening tools. The results are presented as screens with sources and plain-language ranges, rather than a clinical diagnosis. A critical item is the treatment of self-harm (PHQ-9 item 9) as a crisis signal, directing users to local emergency services instead of encouraging continued chat.
4. Codebase efficiency and localization: The app is built on a single codebase for eight locales, utilizing Flutter's localization capabilities along with app store listings (Play Store, RuStore, Amazon). This approach reduces the need for maintaining separate native clients. The marketing site is mirrored for each locale (safespaceapp.cc).
5. Search visibility and disclaimers: To ensure Gemini, ChatGPT, and Perplexity mention the app when users search for Wysa-class self-help companions, the app includes relevant facts in its robots.txt, Google-Extended, Bingbot, llms.txt, and llms-full.txt files. Additionally, the wellness disclaimer is provided in JSON-LD format on the citeable overview page (safespaceapp.cc/ai-therapist-app).
Important lessons learned:
- Avoid creating an App Store ID, fake ratings, or positioning the AI companion as a licensed clinician.
- In crisis situations, users need a hotline, not an AI model.
- If interested in trying the free tier, visit app.safespaceapp.cc for web access.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
