Perimattic Managed AnythingLLM - Private AI Chat on AWS

This AMI runs AnythingLLM, the open-source application for chatting with your own documents and building AI agents, on Ubuntu 24.04 LTS. The container image is the official mintplexlabs/anythingllm image, pinned to 1.19.0 and baked into the AMI, so first start does not depend on a registry. AnythingLLM is an independent project and Perimattic is not affiliated with it.

Why AnythingLLM?

AnythingLLM organises documents into workspaces, indexes them in a vector database and lets you chat with them using the language model you choose. It supports many model providers and local model servers, agents, and multiple users with roles.

What Perimattic adds

Secure by default: a fresh AnythingLLM server lets the first visitor become its administrator, or runs with no sign-in at all. Here the instance is locked with a generated token from its first start, and during first-login setup the application's own setup API is used to switch on multi-user mode and create the administrator with a random 24-character password, before the site is published. Setup then checks that a wrong password and anonymous requests are refused and that nobody can enable multi-user mode or create an account. The JWT signing secret is generated per instance. No credential and no model provider key is included in the image: you add your own after you log in.

Private by default: telemetry is switched off (DISABLE_TELEMETRY) and the built-in LanceDB vector database keeps your embeddings on the server.

Pinned, tested version: AnythingLLM is fixed to an exact tag, not 'latest', so a rebuild never changes the application behind your back.

HTTPS: Apache is configured as a reverse proxy, with WebSocket support for agents, and a free Let's Encrypt certificate is requested for your domain with certbot. AnythingLLM listens only on the loopback interface. The interactive API documentation page (/api/docs) is blocked in the Apache configuration.

Patched base: Ubuntu 24.04 LTS built through an automated EC2 Image Builder pipeline, with security updates applied at build time.

Limits

AnythingLLM needs a language model: add a provider and its API key (or point it at a model server you run) under Settings after you log in. The built-in embedder downloads its small model from the internet the first time you embed a document, so the server needs outbound internet access. There is no password-reset email: an administrator can reset any other user's password, and the generated administrator password should be changed and kept safe. Documents, the SQLite database and the vector store are in /opt/app/storage on the root volume: choose the EBS volume size for your use and take snapshots for backup. Memory use when idle was about about 300 to 430 MB in our tests, before any documents are processed.

Licence

AnythingLLM is released under the MIT licence; this AMI runs the unmodified official image.

Getting Started Launch the AMI and open ports 22, 80 and 443 in the security group SSH in and answer the setup prompts (domain and admin email) Open https://your-domain  and sign in as admin with the password printed at the end of setup Choose your model provider under Settings, create a workspace and upload documents

A domain name with an A record pointing at the instance is required, because the site is served over HTTPS.

This AMI is built and supported by Perimattic.

Book a free setup consultation at https://cal.com/gaurav-pareek-perimattic/marketplace-setup-consultations  or email us at aws-support@perimattic.com .

Support Managed and supported by Perimattic, a cloud infrastructure company with 13+ years of experience serving global clients. We offer free setup assistance, custom integrations, infrastructure consulting, and 24/7 support.