The Problem
Technical teams accumulate meeting transcripts, lecture recordings, and interview materials that remain underutilized. Extracting structured documentation from unstructured spoken content requires manual summarization, outlining, and formatting — time-intensive work that rarely happens consistently.
What This Does
Lumentis automates documentation generation from transcripts and text sources. The core pipeline lives in src/ai.ts, which orchestrates model calls across multiple stages, and src/page-generator.ts, which renders the final output. Configuration and type definitions are in src/types.ts, while src/prompts.ts manages the prompt templates used at each extraction phase. The entry point src/app.ts wires the CLI interface and handles user input flow. The system supports model switching between stages, allowing cost/quality tradeoffs — for example, using Sonnet for heavy lifting and Haiku for lighter output.
How To Use It
Setup: The repository uses package.json with no lockfile present, indicating npm or pnpm usage. Install dependencies with npm install or pnpm install.
Configuration: No environment variable configuration is documented in the repo. The README implies API keys are handled by the underlying model providers (OpenAI, Anthropic, Google) rather than requiring local config files.
Running it: From an empty directory, run npx lumentis. The tool will prompt for a transcript, doc, or notes input. After providing material, answer questions about themes and audience. The system generates an outline for review, then writes the full documentation. Deploy the resulting folder to Vercel per their deployment guide.
(Note: The README warns against running lumentis in the cloned repo directory and mentions clearing npx cache if previously used.)
Real-World Use
A product team records a 45-minute design review and uploads the transcript. Lumentis processes it through the pipeline in src/ai.ts, generating a structured technical doc in under two minutes. The team reviews the outline from src/page-generator.ts, keeps the relevant sections, and pushes the folder to Vercel for stakeholder review — no technical writer time required.
Code Health & Issues
No test files detected — repository-wide untested code paths. No Jest, Vitest, or other test framework configuration present. No CI/CD pipeline detected — no automated build/test gate in .github/ or equivalent. Pushes to main proceed without validation. Dependencies declared without a lockfile — package.json lists dependencies but no pnpm-lock.yaml, package-lock.json, or yarn.lock is tracked. This creates non-reproducible builds across environments. No license file detected in the root file listing, though a LICENSE file exists per the directory structure — verify its contents and ensure it's distributed with the package.
The Bottom Line
Lumentis delivers on its promise of one-click doc generation from transcripts, with a functional pipeline and model-switching flexibility that makes it accessible for solo creators and small teams. The absence of tests, CI, and a lockfile represents a meaningful SDLC gap — the tool works but isn't production-safe for team collaboration without those additions. It's well-suited for individual developers, researchers, or consultants who need to turn spoken material into skimmable docs quickly, but teams requiring reproducibility and quality gates should supplement with their own testing and build infrastructure.