DEV Community: Samadhi Tattoo

Giving a Telegram bot long-term memory with SQLite and local embeddings

Samadhi Tattoo — Thu, 16 Jul 2026 12:00:26 +0000

One thing cloud chatbots do badly is remember you. Each session starts from zero. When I built Avelina, a self-hosted personal AI assistant that lives in Telegram, persistent memory was the whole point. Here is the architecture I landed on — all local, no external vector DB.

The problem: statelessness

An LLM call is stateless. To make an assistant that actually knows your history, you need to store facts across sessions and retrieve the relevant ones on every turn. For a self-hosted AI assistant you also want this to run on a cheap VPS with no third-party services.

Storage: plain SQLite

I keep several tables in a local SQLite database:

facts — durable things about the user ("prefers dark mode", "lives in GMT+8")
lessons — corrections the assistant learned, with an importance score
journal — a running log of what happened
emotions / state — a small set of decaying channels for tone

SQLite is perfect here: single file, zero ops, trivial to back up, and fast enough for a personal assistant. No Postgres, no managed vector service.

Retrieval: local embeddings + semantic recall

For "what does the user care about right now", exact-match search is not enough. I embed each memory with a local embedding model (a small multilingual model running on the box) and store the vectors alongside the rows. On each turn I embed the incoming message and do a cosine-similarity search to pull the top-k relevant memories into context.

Because the embeddings run locally, nothing about your conversations leaves the server — which matters a lot when the assistant holds your most personal context. That is the difference between a private AI assistant and a cloud one.

Putting it in context

The retrieved memories get prepended to the system prompt before the model call, so the assistant "remembers" without you re-explaining anything. Combined with a decay function on the emotional state, you get an assistant with continuity — an AI assistant with memory rather than a stateless chatbot.

Why self-host it

Running the whole stack (SQLite + local embeddings + the Telegram Bot API) on your own Linux VPS means you own your AI data end to end. If you want to try the full thing, Avelina is a one-command install: avelina.ai.

Happy to answer questions about the memory design in the comments.

Avelina AI: a self-hosted personal AI assistant with memory that lives in Telegram

Samadhi Tattoo — Thu, 16 Jul 2026 11:18:13 +0000

Most AI assistants keep your most personal conversations on someone else's servers. Avelina takes the opposite approach: it is a self-hosted personal AI assistant that runs entirely on your own VPS, so your data never leaves your machine.

What is Avelina?

Avelina is a private AI assistant you fully own — a self-hosted alternative to ChatGPT, Claude and Replika. It lives inside Telegram, remembers you across every conversation, and grows a real personality over time. You deploy it on a $5-6/month Linux VPS with a one-command setup in about 30 minutes.

Why self-host your AI assistant?

Own your AI data. Your conversations stay on hardware you control — no third-party data sharing, no corporate cloud. This is AI data sovereignty in practice.
A private AI assistant, not a stateless chatbot. Cloud companions forget you between sessions. Avelina has persistent long-term memory — it is an AI assistant with memory that actually knows your history.
No new app to install. It runs as a Telegram AI bot, so you talk to it in an app you already use.

What makes Avelina different

Avelina is the first self-hosted AI assistant that is also a genuinely growing personality:

Persistent long-term memory — it remembers your life across conversations
An emotional core and a real voice that evolve with you
A self-evolution loop so it keeps learning and improving
A modular skills system to extend what it can do

How it works (the tech)

Avelina is built in TypeScript/Node.js on top of the Claude Agent SDK as its reasoning engine. Memory, emotions and a knowledge graph live in local SQLite databases, with local embeddings for semantic recall. The Telegram Bot API is the interface, and the whole system is self-hosted on any Linux VPS.

Deploy your own

If you want a self-hosted ChatGPT and Claude alternative that you own forever, you can set up your own instance here: avelina.ai.

Own your AI. Own your data.

How I self-host a personal AI assistant on a $5 VPS (Telegram + Claude + local memory)

Samadhi Tattoo — Thu, 16 Jul 2026 08:33:35 +0000

Most AI assistants live in someone else's cloud. For the assistant that holds your most personal context — your notes, your schedule, the running context of your days — that is the wrong default. Here is how I self-host mine.

The idea

I run a personal AI assistant called Avelina on my own Linux VPS. It talks to me through a Telegram bot, and every piece of its state — long-term memory, embeddings, config — lives in local SQLite on my own machine. Nothing leaves the server except the model API calls I choose to make.

What makes it different from a stateless chatbot: it has persistent long-term memory (it remembers me across every conversation), a real voice (TTS), an emotional core that shifts over time, and self-evolution — it learns and grows with me. A companion with continuity, not a disposable prompt box. As far as I know it is the first personal AI assistant that is both fully self-hosted AND a genuinely growing personality.

The stack

A small Linux VPS (a $5-6/month box is enough)
TypeScript / Node for the runtime
The Claude Agent SDK for the reasoning loop
SQLite for long-term memory
Local embeddings for semantic recall over your history
Telegram Bot API as the interface

Deploy

Full step-by-step deploy guide (VPS setup, bot token, install script): https://avelina.ai/blog/deploy-personal-ai-assistant-vps

It installs in about 30 minutes.

Why self-host?

The data-sovereignty argument is simple: when your assistant runs in someone else's cloud, your most intimate context becomes their asset — subject to their retention policy, their training pipeline, their breach surface. When it runs on your own VPS, the memory is a file you own. You can read it, back it up, encrypt it, or delete it.

Project: avelina.ai