InboxPilot+

• 

  Dashboard

• 

  Customer memory tab

• 

  Pipeline simulator tab

• 

  Categories tab

• 

  Analytics tab

• 

  Settings tab

• 

  Category manager tab

• 

  Tutorial screen

Inspiration

As a business owner, your inbox can quickly turn chaotic. Managing a flood of angry customers, complicated refund requests, technical bug reports, and endless spam takes up precious time and mental energy. We aimed to create a co-pilot that stands at the front gate of your business, filtering out the noise, understanding customer feelings, and quickly resolving issues. This way, founders can concentrate on growing their business instead of catching up with customer support.

What it does

InboxPilot+ is an intelligent, automated control panel that intercepts, sorts, and acts on your incoming emails right away. Dynamic Categorization: Automatically directs emails into custom categories like Refund Requests, Bug Reports, Delay Complaints, or Opportunities. Sentiment Detection: Labels emails based on emotional urgency, such as Angry, Frustrated, Neutral, or Positive, to highlight high-risk issues. Autonomous Auto-Replies: Creates context-aware draft responses or sends tailored replies based on custom rules. Custom Theme Control: Features a responsive, minimalist dark/light interface designed around Google's Material UI guidelines.

How we built it

We created the frontend control panel using Streamlit, adding custom CSS to modify core components for a premium, custom-branded interface that supports different accent colors, like Google Blue or Purple Velvet. The backend runs on Python, using the Gmail OAuth2 API for secure user sign-in and mailbox access. Incoming emails pass through regex processing engines and AI sentiment models before being organized into local JSON-based files (business_profile.json, categories_config.json) for easy state maintenance.

Challenges we ran into

One major challenge was managing state persistence and UI re-renders within Streamlit. Since Streamlit scripts restart from the top on user interaction, passing Google OAuth authorization codes through URL query parameters without disrupting UI state or losing active sessions requires careful lifecycle management using @st.cache_resource and strict st.session_state checks. Additionally, normalizing chaotic HTML emails and converting inconsistent email header date formats into structured datetime objects involved a lot of regex adjustments.

Accomplishments that we're proud of

Fully Dynamic UI Engine: Developed a robust CSS injector that smoothly translates theme settings across custom markdown blocks and native Streamlit inputs/buttons alike. Robust OAuth Pipeline: Created a secure, clear Google sign-in process that adapts based on whether the user is logging in for the first time or returning. Granular Migration Logic: Designed the categorization engine to automatically fix itself and update legacy configuration profiles without affecting user data.

What we learned

We learned the complexities of building an enterprise-grade application within strict authentication guidelines like Google's OAuth2. We also gained valuable experience in structuring backend data to allow updates from a frontend toggle without causing race conditions or file-lock errors.

What's next for InboxPilot+

Our next steps include deeper Customer Memory tracking. We plan to analyze historical conversation logs saved in the Customer_Memory/ directory to build a vector database. This will enable the AI engine to remember a customer's past issues, lifetime value, and communication preferences, making auto-replies highly personalized and tailored to each specific relationship.

Built With

• gmail
• google-cloud-console
• python
• qwen-ai-api
• streamlit
• vs-code