Documate

Inspiration

Researchers often struggle with fragmented workflows and a lack of a unified platform. The core problems Documate aims to solve are:

Scattered Information: No single place to maintain citations, inspirations, and overall project progress.
Loss of Context: Difficulty in tracking past research work, understanding its evolution, and recalling the "why" behind pivots.
Inefficient Collaboration: Challenges in conducting and managing group research seamlessly.

What Documate Does

Documate is a productivity tool designed to provide researchers with a centralized platform to manage their entire research lifecycle. Key features include:

Unified Research Hub:
- Maintains all citations, inspirations, and project progress in one place.
- Tracks the entire research journey, preventing loss of previous work.
Collaborative Workspace:
- Facilitates team creation and management for seamless group research.
- Allows collaborative sharing of documents (text, images, videos, PDFs) among team members.
- Includes a collaborative chat interface for team discussions and idea sharing.
Experiment & Progress Management:
- Enables logging of multiple experiments and their different results.
- Documents all research progress comprehensively.
Retrospective View & Insights:
- Offers a "retrospective view" to visualize the research journey, including pivots and their rationale, helping maintain a mental "mind map."
AI-Powered Assistant:
- Conversational Interaction: An integrated AI agent allows users to interact with their research data (documentation, experiments) conversationally.
- Real-time Information Retrieval: Leverages tools like Perplexity to fetch and organize live information (e.g., latest advancements, sources, code repositories) from the internet.
- Actionable Insights: Allows users to directly add discovered insights to their documentation or create to-do tasks within experiment tabs.
- Intelligent Task Management: The AI agent can understand conversational commands to update task statuses (e.g., moving a task to "completed").

How we built it

we have used Nextjs to manage both the frontend and backend
Ai SDK as a framework to use ai models
perplexity to fetch realtime internet data
shadcn for the ui
convex for the database
clerk for authentication
tiptap for text editor

Challenges we ran into

absence of tool calling from the perplexity api we found quick work around by hot reloading the models based on the context , requirements and had the perplexity as a tool call inside the main agent to make it work seemlessly
excess numbering in the citations recieved than the number of source urls