JSPM

Persistent personal AI that actually remembers you.

LLMs like ChatGPT, Claude.ai, and Gemini tend to forget you the moment you end your session. RecallMEM doesn't. It builds a profile of who you are, extracts facts after every conversation, and runs vector search across your entire history to find relevant context. By the time you've used it for a week, it knows you better than any AI ever will.

Use it with Claude or OpenAI for fast responses and the best models (~5 minute setup). Or run everything locally with Gemma 4 for 100% privacy. You'll get the same memory framework either way. Your call.

Two chats. Different sessions. The AI remembers.

What is this

A personal AI chatbot with REAL memory. Plug in any LLM you want and RecallMEM gives it persistent memory of who you are, what you've talked about, and what's currently true vs historical.

The best part is that the LLM will never touch your memory in the database. Every retrieval is deterministic SQL + cosine similarity, assembled by TypeScript before the LLM ever sees it. The LLM only proposes new facts; a TypeScript validator decides what gets stored. Facts have timestamps and get auto-retired when you contradict them ("works at Acme" → "left Acme"). Deep dive on the architecture →

You can run it three ways:

• Cloud LLMs (recommended for most people). Add a Claude or OpenAI API key in Settings. Fast, smart, works on any computer. Your memory still stays local in your own Postgres database. Only the chat messages go to the provider.
• Local LLMs (recommended for privacy). Run Gemma 4 via Ollama. Nothing leaves your machine, ever. Slower setup (~18 GB model download) and slower responses, but truly air-gappable.
• Both. Use cloud for daily chat, switch to local for the sensitive stuff. The model dropdown lets you pick per-conversation.

Features

• Three-layer memory across every chat: synthesized profile, extracted facts table, and vector search over all past conversations
• Temporal awareness so the model knows what's current vs. historical. Auto-retires stale facts when the truth changes.
• Live fact extraction after every assistant reply, not just when the chat ends
• Memory inspector where you can view, edit, or delete every fact
• Vector search across past conversations with dated recall
• Custom rules for how you want the AI to talk to you
• File uploads (images, PDFs, code). Gemma 4 handles vision natively.
• Web search when using Anthropic or Ollama (via Brave Search)
• Wipe memory unrecoverably with DELETE + VACUUM FULL + CHECKPOINT
• Bring any LLM. Ollama, Anthropic, OpenAI, or any OpenAI-compatible API.

Quick start (Mac)

Two options. Pick whichever fits your priority.

Option A: Cloud LLM (Claude or OpenAI) — fastest, ~5 minutes

You need Node.js 20+ and Homebrew. Then:

npx recallmem

The installer sets up Postgres, pgvector, and Ollama (for the embedding model that powers memory). When the browser opens to localhost:3000:

1. Click Settings in the top right
2. Click Providers
3. Add your Claude or OpenAI API key
4. Pick that model from the dropdown in the chat header
5. Start chatting

Total time: ~5 minutes. The AI remembers everything across every chat. Your memory stays in your local Postgres database. Only the chat messages go to the cloud provider.

Option B: Local Gemma 4 — 100% private, ~15-45 minutes

Same npx recallmem command. When the app opens, click Settings → Manage models and download one of these:

• Gemma 4 E4B (4 GB, ~5 minute download) — fastest to test
• Gemma 4 26B (18 GB, ~20-30 minute download) — recommended for daily use
• Gemma 4 31B (19 GB, slower, best quality)

Then pick that model from the dropdown and chat. Nothing leaves your machine.

# Postgres + pgvector (apt example)
sudo apt install postgresql-17 postgresql-17-pgvector
sudo systemctl start postgresql

# Ollama
curl -fsSL https://ollama.com/install.sh | sh
sudo systemctl start ollama
ollama pull embeddinggemma
ollama pull gemma4:26b

# Run
npx recallmem

CLI commands

npx recallmem            # Setup if needed, then start the app
npx recallmem init       # Setup only (deps, DB, models, env)
npx recallmem start      # Start the server (assumes setup done)
npx recallmem doctor     # Check what's missing or broken
npx recallmem upgrade    # Pull latest code, run pending migrations
npx recallmem version    # Print version

Privacy

If you only use Ollama, nothing leaves your machine, ever. You can air-gap the computer and it keeps working. If you add a cloud provider, only the chat messages and your assembled system prompt go to that provider's servers. Your database, embeddings, and saved API keys stay local.

For developers

Underneath the chat UI, RecallMEM is a deterministic memory framework you can fork and use in your own AI app. The whole lib/ folder is intentionally framework-shaped.

lib/
├── memory.ts        Memory orchestrator (profile + facts + vector recall in parallel)
├── prompts.ts       System prompt assembly with all memory context
├── facts.ts         Fact extraction (LLM proposes) + validation (TypeScript decides)
├── profile.ts       Synthesizes a structured profile from active facts
├── chunks.ts        Transcript splitting, embedding, vector search
├── chats.ts         Chat CRUD + transcript serialization
├── post-chat.ts     Post-chat pipeline (title, facts, profile rebuild, embed)
├── rules.ts         Custom user rules / instructions
├── embeddings.ts    EmbeddingGemma calls via Ollama
├── llm.ts           LLM router (Ollama, Anthropic, OpenAI, OpenAI-compatible)
└── db.ts            Postgres pool + configurable user ID resolver

Wire in your own auth with two calls at startup and every lib function respects it. See the developer docs for embedding the memory layer into your own app, the database schema, testing, and optional Langfuse observability.

Docs

Limitations (v0.1)

Text only (no voice yet). No multi-user. No mobile app. OpenAI vision not fully wired. Reasoning models (o1/o3, extended thinking) may have edge cases. Fact supersession is LLM-judged and intentionally conservative. See the full limitations list.

Contributing

Forks, PRs, bug reports, ideas, all welcome. See CONTRIBUTING.md for the dev setup.

License

Apache 2.0. See LICENSE and NOTICE. Use it, modify it, fork it, ship it commercially.

Status

v0.1.2. It works. I use it every day.

I built RecallMEM because I wanted an AI that actually knows me. Not because I'm paranoid about privacy (though that's a nice bonus). The chat models you use today forget you the second you close the tab and that drives me crazy. So I fixed it.

There's no CI, no error monitoring, no SLA. If you want to use it as your daily AI tool, fork it, make it yours, and expect to read the code if something breaks. That's the deal. If this is useful to you, that's cool. If not, no hard feelings.

github.com/RealChrisSean/RecallMEM