Question 1

What is QueryLlama?

Accepted Answer

QueryLlama is a hosted, multi-tenant document search platform: upload your organization's documents, search them by keyword, by meaning (semantic), or both at once, and — for enterprise customers — ask questions and get answers grounded in your own documents using your own choice of LLM. Access is governed by a 5-role permission system enforced at the database layer.

Question 2

Is QueryLlama available today?

Accepted Answer

QueryLlama is in active early development. The database schema, multi-tenant isolation, and the document-ingestion pipeline's chunking and embedding step are built; hybrid search, the question-answering layer, and bring-your-own-LLM proxying are in progress. We're taking on design partners ahead of general availability — join the waitlist and we'll loop you in as capabilities ship.

Question 3

Is this the same kind of product as the old 'natural language to SQL' pitch?

Accepted Answer

No — that was a mismatch with the real product and has been corrected. QueryLlama doesn't generate SQL and isn't a local/offline tool. It's a hosted SaaS document-search and question-answering platform. There's no local-Llama, zero-egress story here: documents and queries are processed by QueryLlama's hosted infrastructure, with strict per-organization data isolation.

Question 4

What file types can I search?

Accepted Answer

PDF, Word (DOCX), Excel (XLSX), PowerPoint (PPTX), plain text, Markdown, HTML, CSV, and common image formats are on the supported list. Image OCR and some of these formats are still being wired into the ingestion pipeline — ask us about your specific format if you're evaluating an early-access deployment.

Question 5

How is my organization's data kept separate from other customers'?

Accepted Answer

Every table in the database carries an organization ID, and Postgres Row-Level Security policies enforce that a query can only ever return rows belonging to the requesting user's organization — at the database layer, not just in application code. That means even an application bug can't leak one customer's documents into another's results.

Question 6

What does 'bring your own LLM' mean?

Accepted Answer

Enterprise customers can point QueryLlama's question-answering step at their own LLM endpoint — anything that speaks an OpenAI-compatible API — instead of the platform's default model. Your organization's API key and endpoint are stored per-organization, not shared across tenants. This capability is actively being built and is not yet available for general use.

Question 7

What are the five roles, and what can each one do?

Accepted Answer

Super admin (cross-organization platform operator), org admin (manages an organization's users, documents and settings), group admin (manages a subset of users/documents within an org), member (searches and asks questions), and viewer (read-only search). Fine-grained, group-level permission rules are still being built out.

Question 8

Is there an audit trail?

Accepted Answer

The database includes an append-only audit-log table — a database trigger blocks edits and deletes on it outright, so a log entry can't be quietly altered after the fact. Wiring every user and document action through to that log is part of the active build, not finished yet.

Question 9

What does the architecture look like?

Accepted Answer

A NestJS API, a Next.js frontend, Cloudflare Workers for document ingestion, Cloudflare R2 for file storage, and Supabase Postgres with pgvector for both relational data and embeddings. See the whitepaper for the full breakdown, including what's running today versus what's still being built.

Straight answers, including 'not yet.'

Talk to the team that actually builds the software.

Send us a message