Architect

• 

  Not approved status

• 

  Approved status along with MVP feature specs

Inspiration

The inspiration for Architect sprouted from the pulsating energy of hackathon enthusiasts eager to conquer the challenges of rapid prototyping and idea validation. In the frenzy of these high-stakes events, participants often find themselves grappling with the need to quickly materialize their concepts into tangible projects. The project's visionary creators recognized this struggle and set out on a quest to harness the power of language models, those enigmatic AI beings, to provide invaluable assistance in this exhilarating endeavor.

The ultimate goal was to create a digital ally that would seamlessly guide and support hackathon participants through the chaotic storm of ideation. Armed with the uncanny abilities of language models, this assistant would generate awe-inspiring frontend designs and craft ingenious backend architectures, all while diligently analyzing project feasibility. The inspiration was nothing short of electrifying—to empower dreamers and innovators alike in transforming their ideas into reality, all within the adrenaline-charged atmosphere of a hackathon.

What it does

Architect is a tool that utilizes three separate language models (LLM chains) to handle different aspects of the ideation process. It generates potential frontend features, technologies, and design ideas, as well as backend architectures, database solutions, and other backend technologies suitable for the user's project idea.

How we built it

The system is built on a pipeline that integrates three distinct language models: Frontend Design Language Model (FLLM), Backend Architecture Language Model (BLLM), and Feasibility Analysis Language Model (FALM). The user provides their initial project idea and specifies their skills/target categories through a simple user interface.

Challenges we ran into

• Accurate Feasibility Evaluation: Ensuring the FALM correctly assesses project feasibility based on technologies, skills, and time constraints is a critical challenge.
• Integration of Multiple LLMs: Managing the interaction and data flow between FLLM, BLLM, and FALM requires careful coordination to incorporate feedback effectively.
• User Experience: Designing an intuitive and user-friendly interface that provides constructive feedback for infeasible projects is a challenge.

Accomplishments that we're proud of

• Developing a functional pipeline that utilizes language models to generate frontend and backend ideas for hackathon projects.
• Implementing the feasibility analysis to evaluate the viability of the proposed ideas.
• Creating a user interface that simplifies the input process and displays generated ideas and feasibility analysis clearly.

What we learned

• Integrating multiple language models in a pipeline to tackle distinct tasks efficiently.
• Designing a user-friendly interface that enhances the overall experience.
• Understanding the complexities of evaluating project feasibility using language models.

What's next for Architect

• An "Inspirational Mode" could be added that generates novel hackathon ideas based on trending technologies and popular themes, offering a starting point for users who don't have a specific idea in mind.
• Integration with educational resources: If a user's skillset does not match the requirements of their desired project, the tool could recommend resources for learning the necessary skills.

Built With

• git
• github
• langchain
• llm
• python
• streamlit