# TeleMem

An agent memory management layer, positioned as a high-performance drop-in replacement for Mem0, adding semantic deduplication, long-term dialogue memory, per-character memory profiles and multimodal video reasoning.

TeleMem can replace Mem0 with a single import line and keeps its `add()` and `search()` argument and return shapes so existing Mem0 code keeps working, while being optimized for multi-turn dialogue, character modeling, long-term storage and semantic retrieval through a context-aware enhancement mechanism.

It automatically builds isolated, per-character memory profiles for role-play, companion AI, NPCs and multi-persona assistants, and adds a video understanding pipeline, video frame extraction, caption generation and vector-database construction, so an agent can store, retrieve and reason over video content the way it handles text memories, with ReAct-style multi-step video question answering.

It runs fully locally by default on Qwen and FAISS, with no cloud service, paid tier, or data leaving the machine, and has first-class support for DeepSeek Harness through an MCP server that registers its memory tools.

Website: https://teleai-uagi.github.io/telemem/