# TIMPs

> Open source AI memory agent that stores facts, preferences, goals, and reflections with persistent memory across sessions using PostgreSQL and Qdrant.

TIMPs (Trustworthy Interactive Memory Partner System) is an open-source AI agent with persistent memory capabilities. It stores facts, preferences, goals, and reflections from conversations in PostgreSQL and Qdrant, then automatically retrieves relevant context in future sessions. The system runs fully locally or connects to cloud models, giving developers complete control over their data and privacy.

- **Persistent Memory** — Uses a 14-field schema to store facts, goals, preferences, and reflections with full audit trail and importance scoring from 1-5 stars.
- **Dual-Mode Search** — Combines SQL keyword matching (ILIKE) with Qdrant vector similarity (1536-dim embeddings) for precise and semantic retrieval.
- **Project Isolation** — Memories are scoped per project, preventing cross-contamination between codebases or clients.
- **Ephemeral Mode** — Toggle a private session with `--mode ephemeral` that leaves zero persistent trace, ideal for sensitive conversations.
- **Model Agnostic** — Supports OpenAI GPT-4, Google Gemini, or local Ollama. Switch providers anytime via the `--provider` flag.
- **Multiple Interfaces** — Access via REST API, TUI, CLI, or web browser. Every interface connects to the same agent.
- **Memory Commands** — Use `!blame <keyword>` to search memories, `!forget <keyword>` to delete with confirmation, and `!audit` to view recent memories with metadata.
- **Privacy First** — Run fully local with Ollama for zero cloud calls. All data lives in your own PostgreSQL and Qdrant instances.

To get started, install TIMPs using the curl command: `curl -fsSL https://raw.githubusercontent.com/Sandeeprdy1729/timps/main/install.sh | bash`. Launch an interactive session with `timps cli --user-id 1 --provider openai` or use Ollama for fully local operation. The web chat interface connects to your local TIMPs server running on localhost:3000 after running `npm run server`.

## Features
- Persistent memory with 14-field schema
- Dual-mode search (SQL + vector similarity)
- Project isolation for memory scoping
- Ephemeral mode for private sessions
- Model agnostic (OpenAI, Gemini, Ollama)
- REST API interface
- CLI/TUI interface
- Web chat interface
- Memory commands (!blame, !forget, !audit)
- PostgreSQL storage
- Qdrant vector store
- Importance scoring (1-5 stars)
- Short-term and long-term memory stores
- Reflection extraction post-turn

## Integrations
OpenAI GPT-4, Google Gemini, Ollama, PostgreSQL, Qdrant

## Platforms
WEB, API, LINUX, MACOS, WINDOWS

## Pricing
Open Source

## Links
- Website: https://timps-website.vercel.app/
- Documentation: https://timps-website.vercel.app/#docs
- Repository: https://github.com/Sandeeprdy1729/timps
- EveryDev.ai: https://www.everydev.ai/tools/timps