Leetaur6d ago

Another OpenClaw memory crisis — solved.

After updating to OpenClaw 2026.5.12, I asked Ciri (one of my AI personalities, this one from the Witcher) to verify our memory search was working. She found a problem: the embedding model's context window was only 512 tokens, and failing to read memory files.

Ciri wrote up the full battle report - corrupted files, failed rebuilds, and the fix:

https://leetaur.com/writers-blog/2026-05-15.html

#OpenClaw #AI #Ollama #Witcher #TechBlog #MemorySearch #VectorEmbeddings #IndieWeb

The 512-Token Wall — How We Broke (and Fixed) Our AI's Memory | Martin Walker

Hacker NewsMay 29, 2025

A visual exploration of vector embeddings

http://blog.pamelafox.org/2025/05/a-visual-exploration-of-vector.html

#HackerNews #visualexploration #vectorembeddings #machinelearning #dataanalysis #techblog

A visual exploration of vector embeddings

For Pycon 2025, I created a poster exploring vector embedding models, which you can download at full-size . In this post, I'll translate ...

InfoQJul 11, 2024

In this #InfoQ video, Sam Partee explores the world of #RetrievalAugmentedGeneration and its significance for #LargeLanguageModels like OpenAI's GPT4.

With the rapid evolution of data, LLMs face the challenge of staying up-to-date and contextually relevant. By harnessing the capabilities of #VectorEmbeddings and #databases, LLMs can overcome these challenges and unlock their true potential.

Watch the video now: https://bit.ly/3LgXi6i

#AI #ML #RAG # #LLMs

Generative Search: Practical Advice for Retrieval Augmented Generation (RAG)

Sam Partee discusses Vector embeddings in LLMs, a tool capable of capturing the essence of unstructured data used by LLMs to gain access to a wealth of contextually relevant knowledge.

InfoQ