Gene Report FastAPI Streaming API

Install dependencies

pip install -r requirements.txt

create a .env file for these keys:

OPENAI_API_KEY=<open_ai_key>
PPLX_API_KEY=<perplexity_key>

Run the FastAPI server

venv/bin/uvicorn api:app  --reload

or

venv/bin/uvicorn api:app 

Usage

POST to /stream-report with JSON body:

{
  "file_content": "<file_content>",
  "user_profile": "",
  "file_type": "text_23andme"
}

Example using curl (with SSE client):

curl -N -X POST -H "Content-Type: application/json" 
  -d '{  "file_content": "<file_content>",
  "user_profile": "<add a user profile in text>",
  "file_type": "text_23andme"}' 
  http://localhost:8000/stream-report

The response will be streamed as Server-Sent Events (SSE) with agent updates.

Vue 3 + TypeScript + Vite

This template should help get you started developing with Vue 3 and TypeScript in Vite. The template uses Vue 3 <script setup> SFCs, check out the script setup docs to learn more.

Learn more about the recommended Project Setup and IDE Support in the Vue Docs TypeScript Guide.

---

For Vue UI

cd view/nume-report

npm run dev

What works

• Can only take 23andMe style text file genetic report, uploaded from local storage
• Calls Clinvar, MyVariant and Ensembl REST APIs for grounding data
• Uses perplexity to find citations and articles for the report
• Provide SNP / Category wise report
• Tries to provide Report in Structure and Easy to understand language
• Uses FASTAPI to stream the AI Agent updates
• Have dynamic LLM generated UI shown in Vue Webpage
• Takes around 4-5 minutes to finish

What doesn’t work yet

• Genetic Reports in other formats: VCF, CSV etc.
• Takes time to generate report
• Report’s structure and quality is not consistent
• The report frontend is not fully dynamic, instead of showing a section as soon as it is finished, it waits for all sections to complete
• Need to make Header and Footer dynamic with patient details and current date
• Source clicking link not working, since UI is in iframe (I am new to Vue JS)

Things I want to Improve

• More Information in the report: How Common is the Variant, Better flow for getting citations, Explanation Tab for technical words.
• Currently using external API, should build our own Vector DB for Genetic Data, will improve both speed and accuracy.
• Prompts need to be improved, structured and evaluated across different data. Maybe use BAML Framework for prompts.
• Need to build a large domain knowledge to act as a guard-rails and structure for the Report Output.
• Need to add Human-Made Reports Example to do few shots generations.
• Definitely the UI can be lot better
• More Chart, Graphs and Tables.

Stack

• Python for Backend
• Langgraph for Agent Flow and State
• Langchain for custom Tools
• OpenAI, Perplexity
• Tailwind CSS for Report UI
• Vue JS for Web page