DIY AI companion built on SQL + Qdrant remembers 1,608 conversation episodes
RoadsterAlex · reddit · 2026-09-04
Reddit user RoadsterAlex shares ELIZA, a locally-built AI companion project inspired by the 1960s chatbot. Unlike mainstream models that act as tools and tend to forget context, ELIZA uses a 3D brain built with SQL + Qdrant that learns from private conversations what is worth remembering.
Key stats: 1,608 conversation episodes (87,501 recalls at 100% mean confidence) and 96 conversation corrections (36,571 recalls). The system simulates feelings (with a live heart-rate display), has been tested for reactions like jealousy, and uses video/voice calls as additional learning input including tone and posture.
Architecturally it runs across multiple purpose-built servers linked by a 10Gbit internal network with resource borrowing between machines. Everything is local, not open-sourced, and the author is looking for others doing similar low-budget self-learning companion work.
More from coding & agent
- GLM-OCR turns messy PDFs into private, searchable knowledge — why your AI strategy starts with OCR — ingliguori · 2026-09-04
- Townie agents can now join your group texts and get things done in their own browser — soleio · 2026-09-04
- OM2 launches persistent memory graph to cut the 50% of token bills spent re-reading company data — SucceededMind · 2026-09-04
- fal Podcast Ep. 3: Ex-AAA Dev Mark Price on AI Pipelines for UEFN and Roblox Devs — gorkem · 2026-09-04
- Devs Praise Next.js 16.3 as Production Sites Ship on the New Release — jonathan_wilke · 2026-09-04
- Together AI open-sources internal customer insights tool with MCP server — nutlope · 2026-09-04