id8 | Devpost

Inspiration

Recognizing the major bottleneck for today’s product teams - the tedious manual effort required to turn scattered documentation into a clear, actionable backlog - we created id8, a smart API backend designed to streamline the process. id8 ingests all of a product’s existing knowledge base and automatically organizes it into structured epics and tasks, giving product managers more time to focus on strategy, decision-making, and team leadership, instead of wrestling with paperwork.

How id8 Works

id8 is an intelligent API designed to serve as a product manager's right hand. By taking in all your product’s key documents - PDFs, DOCX files, and text notes - it organizes this information into clear, actionable Jira user stories and epics. With a single call to its /generate-epics endpoint, id8 gathers relevant details from across your knowledge base and turns them into a prioritized list of software epics, complete with helpful descriptions and acceptance criteria. This makes planning and backlog creation fast, reliable, and easy.

Behind the Build

To build id8, we chose LangChain to coordinate our retrieval-augmented generation (RAG) pipeline, acting as the central glue for document retrieval and context synthesis. We utilized HuggingFace all-MiniLM-L6-v2 model for document embeddings and stored those vectors with ChromaDB for persistent, efficient search. Our backend uses Pydantic for speed and scalability, and FastAPI for blazing-fast RAG and LLM-generation endpoints. To generate high-quality, structured JSON from retrieved knowledge, we integrated OpenRouter, giving us flexible access to advanced LLMs (like NVIDIA Nemotron).

Technical Hurdles

Handling a wide variety of document types proved to be our biggest hurdle. PDFs, DOCX files, and plain text files all have their own quirks and internal structures. To manage this, we built a custom loading pipeline that could reliably handle and clean up these formats. Tools like PyPDFLoader and Docx2txtLoader often generate messy, non-semantic output, like broken tables or split sentences, so we invested significant effort in processing and sanitizing the extracted text.

Key Achievements

One of our proudest achievements with id8 was developing a truly agentic workflow. We created an agent that doesn't wait for just a simple question, but executes a complex chain of tasks: It begins by ingesting and understanding a wide range of unstructured documents, organizing everything within a ChromaDB vector store. When tasked with a high-level goal, like a /generate-epics request, the agent runs a full RAG pipeline. This lets it reason over all the data it has collected, pulling together details from vision docs, user feedback, and requirements to synthesize a coherent, actionable backlog. Rather than acting on single pieces of information, it creates meaningful connections across the entire knowledge base.

Insights and Takeaways

Building id8 was an incredibly fun and rewarding journey that taught us the power of a modular tech stack. We gained hands-on experience with the practical challenges of designing a RAG pipeline, from prompt engineering for consistent JSON output to orchestrating async tasks with FastAPI. This process reinforced the value of clear API design. Our goal was to build a backend "engine" that other tools and teams could easily integrate and build upon. Overall, we enjoyed seeing how each piece fit together to create something robust and flexible.

Coming Soon to id8

Our vision is to evolve id8 from an "epic generator" into a truly comprehensive, automated product management agent that supports the entire product lifecycle. Our key upcoming developments include:

Go-to-Market Generation: New endpoints allowing id8 to automatically generate release notes, stakeholder updates, and marketing copy, all based on your existing product documentation.
Wireframe Generation: An agent that reads the generated epics and creates draft test cases and wireframe descriptions, accelerating the transition from idea to design.
Proactive Backlog Grooming: An agent that can be configured to scan for conflicting requirements proactively and suggest epic prioritization as you add new documents, keeping your backlog healthy and up-to-date.

These additions will make id8 an even more powerful partner for modern product teams.