SALT stores chatbot memory in a trie and uses theme-based retrieval, but still over-retrieves
No_Sky9786 · reddit · 2026-08-04
The author describes SALT, a chatbot memory system that stores all input in a trie in DRAM and retrieves sentences using a keyword-theme dominance system with a current CELF budget of 20%.
They say the system is efficient, but it often retrieves too much irrelevant information, which makes smaller models hallucinate. They’ve tried using theme coverage as a unit for retrieval, but it still pulls in too much material within the same theme. As they add agents and more modules, they expect memory management to become an even bigger problem and ask for suggestions.
Related event: SALT Uses Trie for Agent Memory(2 posts)→
More from coding & agent
- Eve pushes enterprise agents toward one customizable assistant per company — aarthir · 2026-08-04
- Enkstein open-sources a local control plane for Codex, Claude, Gemini, and Ollama — wcoreiron · 2026-08-04
- Gemini Spark can connect MCP servers, but every tool still needs repeated approval — tristanbob · 2026-08-04
- Agent apps should expose a skill surface, then crystallize it into code — burny_tech · 2026-08-04
- Use subagents to switch models inside one session — dotey · 2026-08-04
- xAI updates Grok Build with Grok 4.5, skills, MCP, and plan mode — elonmusk · 2026-08-04