Similia
AI wellness consultation chat
Similia is a chat consultation app for a holistic health platform in Germany, answering in English or German across five disciplines: Ayurveda, homeopathy, medicinal mushroom therapy, tissue salts, and psychosomatics. Each answer is grounded in the platform's own trained material instead of generic chat output.
- 2025
- Germany
- PHP admin + API
- Python RAG service
- OpenAI + Chroma
- English + German
The challenge
Similia's practitioners wanted to offer users on-demand guidance across several holistic disciplines, but a generic chatbot would answer from general internet knowledge rather than the platform's own body of material, and would mix up the tone and rules of different disciplines, for example giving remedy reasoning in a psychosomatics conversation where only discussion is appropriate. We built a retrieval-backed chat that looks up the relevant discipline's content before answering, with a separate prompt and rules per discipline.
Who uses it
- Users
- Sign up, choose a discipline, chat in English or German, and buy credits to continue consultations.
- Admins
- Manage per-discipline prompts and training content, and oversee user accounts.
What we built
A chat web app, an admin panel and a Python retrieval service behind a shared PHP API. The Python service finds the relevant passages in a Chroma vector store before calling OpenAI, so answers stay grounded in Similia's own material for each discipline. Admins manage prompts and training content per discipline; users sign up, chat and buy credits.
Retrieval-augmented answers
A Python service searches a Chroma vector store for the most relevant passages before the AI answers, grounding responses in Similia's own material.
Five disciplines
Separate prompts and content stores for Ayurveda, homeopathy, mushroom therapy, tissue salts, and psychosomatics.
Bilingual chat
Users converse in English or German; language-switching is built into the chat UI.
Discipline-aware rules
Psychosomatics conversations stay discussion-based with no remedy reasoning; other disciplines include a brief rationale when suggesting a remedy.
Small-talk detection
Short acknowledgments like "thanks" or "okay" are recognized so the AI doesn't over-answer casual replies.
Credit system
Users sign up with a referral code for free credits and purchase more to continue chatting.
Admin prompt management
Admins edit the system prompt and training content for each discipline independently.
Relevance threshold
Retrieval only returns passages above a similarity-score threshold, falling back to a general answer when nothing relevant is found.
How it works
- User signs up, optionally with a referral code for free credits.
- User picks a discipline (Ayurveda, homeopathy, mushrooms, tissue salts, or psychosomatics) and starts a chat.
- The Python service searches that discipline's Chroma vector store for relevant passages.
- The relevant passages, chat history, and discipline-specific system prompt are sent to OpenAI to generate the reply.
- The user continues the conversation, spending credits, and can switch language at any point.
Where the AI fits
Each user message is embedded and matched against a Chroma vector store scoped to the chosen discipline; only passages above a similarity threshold are used as context, and if nothing scores highly enough, the service answers without retrieved context rather than forcing a match. The matched passages, formatted chat history, and a discipline-specific system prompt are sent to OpenAI's chat model to produce the reply, so answers stay tied to Similia's own material and each discipline's tone, for example withholding remedy reasoning in psychosomatics conversations.
- OpenAI chat completions via LangChain
- Chroma vector store, one knowledge base per discipline
- Relevance-score threshold (0.7) before using retrieved context
- Discipline-specific system prompts stored and edited in the admin panel
Every part of the product
User chat app
PHP
Web app where users sign up, select a discipline, chat, and manage credits and subscriptions.
Admin panel
PHP
Manages per-discipline prompts, training content, and user accounts.
RAG service
Python/Flask
Retrieves relevant passages from Chroma and calls OpenAI to generate grounded chat replies.
API
PHP
Shared backend for authentication, credits, and chat history storage.
How it’s built
A Python/Flask service handles embeddings, Chroma retrieval, and the OpenAI call, sitting behind the shared PHP API so the web chat app and admin panel don't talk to OpenAI directly. Each discipline has its own prompt file and vector store, selected by the module the user is chatting in, which keeps homeopathy reasoning from leaking into a psychosomatics conversation. Small-talk detection avoids running a full retrieval/answer cycle for short acknowledgments, and the chat UI and all emails support English and German.
- PHP
- MySQL
- Python
- Flask
- LangChain
- OpenAI API
- Chroma
- pymysql
- JavaScript
- i18n (EN/DE)
Related services: LLM integration AI app development
What clients say about working with us
Verified Fiverr reviews from our clients, quoted word for word.
Thank you for your excellent work Venkey and team. It was a pleasure working with you together on this complex project. Highly recommended.
friends_madlenGermany · Client on this projectGreat experience working with this developer. The app was delivered clean, bug-free, and exactly as requested. Super quick to respond and very detail-oriented, highly recommend!
jamillesodreRepeat buyerBrazil · AI Mobile Apps
The team at Venzpire by Venkey are truly outstanding professionals. They did an excellent job and I am extremely satisfied with the results. Communication was clear and consistent and their consultation was spot-on. What impressed me the most is how genuinely human and approachable they are. Highly recommended!
vladpereverzyevRepeat buyerItaly · AI Technology Consulting
The team understood the requirement and was able to deliver before time. They also took care of the deployment. Great job !
debabratatha136India · AI Websites & Software
Questions about Similia
What does Similia do?
Similia is a bilingual AI chat consultation that Venzpire built for a holistic health platform in Germany, covering Ayurveda, homeopathy, medicinal mushroom therapy, tissue salts, and psychosomatics. Users chat in English or German and get answers grounded in the platform's own trained material rather than generic chatbot responses.
Which platforms does Similia run on?
Similia is a web chat app for users and a separate web admin panel for managing prompts and content, both built on PHP, with a Python retrieval service behind them. Venzpire built all three as one connected system.
How does Similia use AI?
Similia retrieves relevant passages from a Chroma vector store scoped to the chosen discipline, then sends them along with the conversation history and a discipline-specific system prompt to OpenAI to generate the reply. This retrieval-augmented approach, built by Venzpire, keeps answers grounded in Similia's own material and each discipline's rules rather than general chatbot knowledge.
Your product could be next.
Tell us what you are building. Venktesh reads every application himself and replies within one business day.
Apply for a strategy call