FluentDocs: Keep Docs Accurate as Code Evolves
Source: Dev.to
Introduction
FluentDocs is a documentation maintenance system that keeps your docs accurate as your code changes. Its value is not in writing new documentation, but in eliminating the hidden cost of keeping documentation up to date. Instead of relying on memory, manual audits, or late CI warnings, FluentDocs continuously connects code changes to the documentation they affect.
With every git push, FluentDocs identifies which parts of your docs are impacted, explains why they are affected, and proposes precise, section‑level updates. Writers and developers stay in control by reviewing, editing, or rejecting suggestions in a WYSIWYG editor, then shipping the result as a standard pull request.
Live Demo
Background
FluentDocs is an upgrade to an AI Agent called AutoDoc, which won the first Dev.to challenge. Seeing its real‑world value and potential, it was turned into a fully fledged product. The goal is to help developers and technical writers keep documentation accurate as code evolves, without adding more manual work to their workflow.
Building FluentDocs also reflects a broader ambition to move from side projects to building businesses and exploring entrepreneurship through solving real problems.
Tech Stack
- Frontend & Backend: Next.js
- Editor: Tiptap (WYSIWYG) with a custom Markdown/Diff plugin
- AI Agent: FastAPI
- Authentication: NextAuth.js with GitHub provider
- Database & Search: MongoDB Atlas with Vector Search for semantic documentation mapping
- Infrastructure & Tooling:
- GitHub API: Octokit for repository interaction and PR creation
- Real‑time Updates: Server‑Sent Events (SSE)
Key Differentiators
Event‑Driven Automation
Unlike AI assistants that require manual prompts (e.g., Copilot), FluentDocs triggers automatically on every git push, proactively identifying “documentation drift” before it becomes technical debt.
Semantic Context Mapping
Uses vector embeddings and search to understand the relationship between code logic and documentation chunks, ensuring suggestions are contextually relevant rather than just keyword‑based.
Specialized WYSIWYG Diffing
The custom Tiptap‑based editor provides a visual interface for reviewing AI suggestions, allowing users to see rich‑text diffs and accept changes without manually editing raw Markdown files.
Cross‑Repository Intelligence
Supports complex architectures where documentation lives in a separate repository from the source code, maintaining synchronization across distributed systems.
Seamless PR Integration
Orchestrates the full lifecycle—from webhook ingestion to automated Pull Request creation via Octokit—fitting directly into existing CI/CD pipelines without requiring developers to change their tools.
Submission Context
This submission was made for the DEV’s Worldwide Show Tell Challenge presented by Mux.