Chat-lab.ai Documentation
Chat-lab.ai is a suite of micro-tools that encourage the use of generative AI in academic and professional workflows. This site documents how each tool works under the hood: which services and models power it, what options you can set, and what happens to your data.
The tools
| Tool | What it does | Why it helps AI & LLM work | Sample use cases | Powered by |
|---|---|---|---|---|
| Audio Insights | Transcribe audio/video with speaker labels, then generate AI summaries | Audio is invisible to language models. A clean, speaker-labeled transcript turns recordings into text an AI can summarize, search, and reason over | Transcribe research interviews for qualitative analysis, turn recorded lectures into study guides, distill committee meetings into action items | AssemblyAI, Groq, OpenAI |
| Bibliobot | Extract bibliographic metadata from academic papers | AI-assisted research still needs accurate, structured citations. The ID-first pipeline (DOI, SSRN, arXiv) pulls verified metadata from the scholarly record instead of asking a model to guess it | Build the citation sheet for a literature review, convert CSV metadata to BibTeX, catalog a course reading list, organize source PDFs for an audit or research file | Default: CrossRef, arXiv & Mistral, with Google Gemini for scanned PDFs. Optional: OpenAI |
| PDF-MD | Convert PDFs to LLM-ready Markdown | PDFs paste terribly into chatbots: broken lines, layout junk, lost tables. Markdown preserves structure and is far more token-efficient | Prep journal articles for AI-assisted summarization, create a personal knowledge base to serve as context for RAG applications, convert accounting standards or regulations for querying, make course handouts chatbot-ready | Jina Reader |
| URL-MD | Convert web pages to LLM-ready Markdown | Copy-pasting web pages drags in ads, menus, and clutter; reader-mode Markdown gives the model just the text content | Collect articles into a research corpus, save documentation for offline AI use, pull IRS/SEC guidance pages into a prompt | Jina Reader |
| Sitemap Crawler | Map a site's pages and download them as Markdown | One page is rarely enough context. A whole-site Markdown archive is a ready-made corpus for NotebookLM-style tools, RAG pipelines, or offline study | Archive a university policy site for a compliance assistant, generate a site map with clean URLs for text data collection in research projects, turn a course website into an AI-queryable knowledge base | Firecrawl, Jina Reader |
Common principles
Bring your own key, or don't. Every tool works out of the box with the app's own keys, which rely on the providers' free tiers. These are generally safe to use, but free-tier data processing policies can conflict with your institution's vendor agreements or simply with your own preferences. Each tool therefore also accepts your own API key, which may be worthwhile if your account is on a paid tier with stronger data protections. Your API key is kept only in your Chat-lab.ai browser session and never stored anywhere. See each tool's page for specifics.
Nothing you upload is kept. Files are processed and discarded; output files live only in your browser session. Further details on the Privacy & Data Retention page.
LLM-ready output. The Markdown tools exist because clean Markdown is the most token-efficient way to feed documents and web content to language models. Paste the outputs straight into your favorite AI chat or LLM pipeline.