Skip to content

Sukoon v2 Development Plan

Jump to bottom
Luv edited this page Oct 9, 2024 · 5 revisions

Sukoon v2 Development Plan

Week 1: Project Kick-off and Planning

Week 1: Team Reorganization and Goal Setting

  • Task: Conduct a team-wide meeting to reignite project vision and set clear goals for v2.
  • Deadline: Starts from first meeting this week

Week 1: Skill Assessment and Task Assignment

  • Task: Assess team members' skills and interests, assign roles based on preferences. Explain the current approach for building chatbot using AI agents using LangGraph. Include tutorials on LangGraph and its implementation in chatbots.
  • Role: Project Manager & Tech Lead
  • Deadline: End of this Week

Architecture Review and Planning

  • Task: Review current architecture, propose improvements for v2. Explore alternative approaches if students want to built it from ground up, clearly explain the tech tasks that are needed to be done. Tasks might include, but not limited to:
  1. Complete LangGraph flow and connect to WhatsApp API
    • Actions:
      • Complete LangGraph agentic framework & improve upon it
      • Implement WhatsApp API integration with the LangGraph framework.
  • Role: Tech Lead & Backend Developer
  • Deadline: End of Week 1-2

Week 2: Core Development Sprint

Iterate on agentic flow

  • Task: Refine and optimize the LangGraph agentic framework. This includes:
    • Improving agent decision-making processes
    • Optimizing the overall flow of information and tasks between agents
    • Implementing advanced features like memory and context management
  • Role: ML Engineering, AI engineering
  • Deadline: End of week 2

Week 3: Evals pipeline

  • Task: Improved accuracy and efficiency of the evaluation pipeline.
  • Role: ML Engineer & Backend Developer
  • Deadline: in parallel with week 2 & 3

Week 3: Database and UI Enhancements

Week 3: Database Optimization

  • Task: Optimize database schema and queries for improved performance.
  • Role: Backend Developer
  • Deadline: End of Week 3

Week 3: UI/UX Enhancements

  • Task: Implement user interface improvements based on v1 feedback.
  • Role: Frontend Developer
  • Deadline: End of Week 3

Week 4: Feature Implementation

Multi-language Support and voice support

  • Task: Implement support for multiple Indian languages.
  • Role: Integrating APIs like Bhashini
  • Deadline: End of Week 4

Week 4: Testing, Integration, and Launch Preparation

Comprehensive Testing

  • Task: Conduct thorough testing of all new features and improvements.
  • Role: QA Specialist & All Developers
  • Deadline: End of Week 4

Final Integration and Bug Fixes

  • Task: Integrate all components and address any identified issues.
  • Role: Tech Lead & All Developers
  • Deadline: End of Week 4-5

Additional Notes:

I will:

  • Hold weekly stand-up meetings to track progress and address blockers. Here they can showcase their work and receive feedback.
  • Celebrate small wins and milestones throughout the process.
  • Organise optional learning sessions on key technologies used in the project.
  • Maintain open communication channels and be readily available to address any concerns or questions from the team.

Resources for learning Langgragh:

  1. https://github.com/langchain-ai/langchain-academy
  2. https://academy.langchain.com/ (official video based series)

How Langchain is being used in our project:

Hey, we're using multi-agent system to provide supportive conversations and crisis intervention for students experiencing stress or mental health challenges. The whole workflow works like this:

  1. User Query: The interaction begins with a user input, typically expressing their concerns or state of mind.

  2. Planner Agent: This agent analyzes the user's query to determine the appropriate response path. It acts something as a triage system, deciding whether the situation requires general support or more urgent intervention.

  3. Empathetic Conversational Agent: For non-crisis situations, this agent engages in supportive dialogue. It aims to provide relief for student stress through empathetic conversation. When needed, it can use a Retrieval-Augmented Generation (RAG) tool to access relevant information to enhance its responses.

  4. Suicide Prevention Agent: If the Planner Agent detects potential crisis indicators, it routes the conversation to this specialized agent. This agent is trained to apply crisis intervention techniques like QPR (Question, Persuade, Refer) and can direct users to professional help or suicide prevention resources.

  5. RAG Tool: This tool allows the Conversational Agent to retrieve and incorporate relevant information into its responses, enhancing the quality and specificity of support provided.

This multi-agent approach ensures that users receive appropriate levels of support, from general stress relief to critical intervention, in a flexible and responsive manner.

Clone this wiki locally