Objective: To optimize the existing Dify chatbot on Gluu.Repair, enhancing its effectiveness in engaging users, streamlining workflows, providing valuable client insights, and leveraging a data-driven approach for continuous improvement. This includes building a fully coded Admin Panel and integrating Brevo email subscriber data.

Project Deliverables:

1. Data Architecture & Supabase Integration Enhancement

• Objective: Establish a robust and flexible data infrastructure using Supabase to capture and manage conversational data.
• Tasks:
• Database Schema Design: Design a Supabase database schema to store conversational data (user ID (Brevo contact ID), conversation ID, timestamp, message direction, message text, extracted intent, prompt used, response text, context variables, rating, etc.).
• Dify API Integration: Implement Dify API integration to create and append records in Supabase, capturing relevant context variables.
• Brevo Integration: Implement a Supabase function or API endpoint to retrieve relevant data from Brevo.

2. Dify Workflow Analysis & Optimization

• Objective: Streamline conversational flows to improve efficiency and accuracy in responding to client inquiries.
• Tasks:
• Workflow Mapping: Analyze existing Dify conversational flows for key user intents.
• Bottleneck Identification: Identify areas for simplification and optimization.
• A/B Testing (Optional): Implement A/B testing of different workflow variations within Dify.
• Code Refactoring: Implement DRY (Don't Repeat Yourself) principles to clean and standardize custom code in Dify.
• Decision Implementation: Implement "if" code and custom actions to direct users to additional services.

3. LLM Performance Tuning & Prompt Engineering (Within Dify)

• Objective: Enhance the accuracy and relevance of Dify chatbot responses by optimizing prompts within the Dify framework.
• Tasks:
• Prompt Review: Review existing LLM prompts within Dify for clarity, accuracy, and bias.
• Prompt Experimentation: Experiment with different prompt engineering techniques within the Dify interface.
• Feedback Incorporation: Implement a feedback mechanism (if not already present) to identify prompts performing poorly or bias.

4. Knowledge Base Enhancement & Content Strategy

• Objective: Develop dynamic and adaptive knowledge base within DIFY.
• Tasks:
• Implement a strategy of having dynamic prompts.
• Automate the workflow process to get more information and knowledge

5. User Feedback Implementation & Insight Generation

• Objective: Establish a user feedback system that the business can implement.
• Task:
• Report on trends
• Categorize to help with actions and tasks.

6. Admin Panel Development (Fully Coded)

• Objective: Build a user-friendly, fully coded Admin Panel to monitor and manage the Dify chatbot.
• Tasks:
• UI Design: Design an intuitive Admin Panel interface (Dashboard, Conversation Log, Feedback Analysis, Knowledge Base Editor, Workflow Editor, Brevo User Data).
• Implementation: Build the Admin Panel with your chosen coding languages, leveraging existing front-end frameworks and UI component libraries.
• Connect Admin Panel with the Supabase data: Access and display data from Supabase.
• Brevo Data Integration: Display key Brevo email subscriber data.
• Authentication: Implement user authentication and authorization.

7. BREVO Integration (User Tracking)

• Objective: Integrate Brevo data into your code.
• Tasks:
• Find a way to pull and show.
• Show what actions clients have been taking.

Project Tools

• Dify
• Brevo
• Supabase (postgres)
• Vercel (python)
• GIT for coding standards

Expected Outcomes:

• Well architect conversation to help the customer and business.
• Increase user integration.
• A strong user face for business.
• Admin panel with integrated Brevo data.

Project Management

• Documentation
• Code standards