Inspiration
Medicinal-plant information exists across many sources, but it is fragmented between regulatory documents, clinical literature and traditional knowledge. We wanted to connect these sources without treating them as the same type of evidence.
What it does
PhyThera lets users (health care professionals such as consulting pharmacists) search by indication and compare medicinal plants based on modern evidence, EMA/HMPC support, safety information and source provenance. Where permitted, it also connects traditional-use information and shows cross-cultural similarities separately from clinical evidence.
How we built it
n8n automates the retrieval, extraction and validation workflows, Supabase stores the structured information, and Streamlit provides the user interface. AI is used to structure complex documents and support semantic mapping, while the final scores are calculated with transparent rule-based logic.
Challenges we ran into
Different sources use very different terminology, formats and evidence standards, so building one structured system without losing context was difficult. We also had to handle missing data correctly and check whether automated access was actually permitted for each source.
Accomplishments that we're proud of
We built a working end-to-end MVP with 11 medicinal plants, from source ingestion to searchable evidence profiles. The system keeps different evidence types separate, preserves provenance and can be expanded without rebuilding the whole architecture.
What we learned
We learned that good evidence integration is not just about collecting more data, but about preserving context and uncertainty. We also learned that AI works best here as a tool for structuring information, not for making final scientific conclusions.
What's next for PhyThera
The next step is to expand the system to more plants, indications and approved databases such as WHO monographs, pharmacopoeias or clinical trial registries. We also want to explore patterns in bioactive compounds to generate new hypotheses for future research and drug discovery.
Built With
- ai
- javascript
- n8n
- python
- streamlit
- supabase
