thought-retriever

🧠 Thought-Retriever

Don't Just Retrieve Raw Data — Retrieve Thoughts

A self-evolving long-term memory system for LLM-based agents, based on the paper by UIUC, MIT & CMU (TMLR 2026).

v2.0.0 — Chinese Optimization

• 🇨🇳 Chinese Embedding Engine: jieba tokenization + TF-IDF, 3-5x better for Chinese
• 🇨🇳 Chinese Prompt Templates: Auto language detection
• 🧹 Smart Filtering: Skip meaningless messages, clean LLM output labels
• 🛡 Robust Parsing: Support Chinese "是/否/有效/无效"

Quick Install

pip install numpy jieba
git clone https://github.com/mhx0628/thought-retriever

Quick Start

from thought_retriever import ThoughtMemory, ThoughtConfig

config = ThoughtConfig(project_path=".", language="zh")
memory = ThoughtMemory(config=config)

memory.add_knowledge("小明今年10岁，喜欢画画和踢足球")
results = memory.retrieve("小明喜欢什么")

Key Features

• 🧠 Thought Memory: Distills Q&A into reusable "knowledge diamonds"
• 🔄 Self-Evolving: The more queries you process, the smarter the memory gets
• 🎚 Abstraction Hierarchy: Shallow facts → deep insights
• 🛡 Dual Filters: Anti-hallucination + anti-redundancy
• 🔌 Model-Agnostic: Works with any LLM
• 🌐 Offline-First: 4-tier fallback embedding engine

Links

• GitHub: https://github.com/mhx0628/thought-retriever
• Gitee: https://gitee.com/ma-hongxing-1/thought-retriever
• Paper: https://arxiv.org/abs/2604.12231
• CHANGELOG: https://github.com/mhx0628/thought-retriever/blob/main/CHANGELOG.md