plai.chat
plai.chat
Get retrieval-augmented answers from your own price list, catalogue, FAQ or handbook — without building a RAG pipeline. Add a file or link to a plai.chat conversation and ask. Answers cite their sources; not stored on our servers.
Open PLAI.chat →Free account · no subscription required · free or pay-per-use models · 300+ models
RAG (retrieval-augmented generation) lets an AI answer from your documents instead of only its training data. A do-it-yourself setup means an embeddings model, a vector database and code to glue them together. In plai.chat it is three steps:
Details on formats and how retrieval works: chat with your documents.
Answer questions from your own price list, catalogue or policies without hiring a developer.
Reply to customers from an FAQ or knowledge file, grounded in your own wording.
Look up product specs and prices from a CSV export during a call or before a quote.
Check procedures and onboarding steps in a Markdown handbook.
Load a client's docs or API spec (JSON / YAML) and ask about it.
Ask questions about your own notes, in your own language.
Query a dataset in CSV or a text corpus and keep track of where each answer came from.
“What does the premium package cost with installation?” — answered from your CSV, each chunk keeps the header row.
Find products by attribute, size or availability from a catalogue export.
Returns, warranty or shipping rules from a Markdown or text file, answered from your own policy.
New hires ask about tools and processes; answers come from the handbook.
Link your website's llms-full.txt and ask about your own products; Sources link back to the right page.
Add two or three supplier offers as CSV or text and ask what differs in price and terms.
| DIY RAG pipeline | plai.chat simple RAG | |
|---|---|---|
| Setup | Write ingestion, chunking, embeddings and retrieval code | Add a file or link in the chat |
| Infrastructure | Vector database, hosting, monitoring | None to run |
| Cost | Embeddings, database and hosting costs | Indexing included; free models or pay-per-use answers; no subscription required |
| Privacy | Depends on your setup | Not stored permanently on our servers (uploads kept in your browser on a best-effort basis, links as links; server index in memory only); processed by the embedding provider and your chat model; ZDR applies to embeddings |
| Model choice | Whatever you integrate | Any of 300+ chat models, switch any time |
| Scale | As large as you build it | Up to 3 sources per chat, 2 MB each; 10 new large sources indexed per day |
| Formats | Anything you can parse | Text, Markdown, JSON/JSONL, YAML, CSV/TSV — no PDF or HTML pages yet |
| Sharing | Shared knowledge base for a team | Per conversation, not shared team-wide |
If you need a team-wide knowledge base over thousands of PDFs, a dedicated RAG pipeline is the better tool. If you want quick, cited answers from a few business files today, plai.chat is the simple option.
plai.chat does not store your documents or chats permanently on its servers. Uploaded files are kept in your browser on a best-effort basis (not in private windows) and links as links; the server indexes them in memory only and rebuilds after a restart, or asks you to re-upload a file your browser no longer has. The text is processed by the embedding provider and sent to the chat model you pick. Zero Data Retention mode applies to embeddings too. See pricing for model costs.
RAG (retrieval-augmented generation) means the AI looks up the relevant passages in your own documents and answers from them, citing numbered passages (small documents are read in full). In plai.chat this is built into every chat: you add a link or file and ask — no embeddings pipeline, vector database or code to set up.
Yes. A signed-in user (a free account is enough) adds up to 3 sources per conversation from the composer's “Chat context” (book) button — for example a price list CSV, a product catalogue, an FAQ or an onboarding handbook — and asks questions in plain language. Nothing to install or host.
There is no subscription requirement and no extra charge for embeddings or indexing — that cost is covered by plai.chat. Answers are billed like any message: free models stay free; on paid models the source text sent with each message counts as input tokens, so messages with sources cost more.
.txt, .md, .json, .jsonl, .yaml / .yml, .csv and .tsv, uploaded or fetched from a link, up to 2 MB each. PDF and HTML web pages are not supported yet; a link must point to a text, Markdown, JSON, YAML or CSV file such as your site's llms-full.txt or a CSV export.
No. Documents are not stored permanently on our servers. Uploaded files are kept in your browser (best effort) and links are stored as links; the server holds the text and index in memory only and rebuilds them after a restart (links re-fetched, uploads re-sent from your browser, or you are asked to re-upload). The text is processed by the embedding provider (via OpenRouter) and sent to the model you pick to answer. With Zero Data Retention mode on, embeddings use ZDR too.
Not yet. Sources belong to one conversation in one browser — up to 3 per chat, 2 MB each, with daily and hourly indexing caps — rather than a shared, team-wide knowledge base. Each person adds the files they need to their own chat. If free models are busy, a backup model may answer without your sources.
Any chat model in plai.chat — 300+ models, free tier included — with follow-up questions. You can add per-chat custom instructions, e.g. a tone or role for support answers.
No code. No vector database. Answers with sources from your own documents.
Open PLAI.chat →