One Redditor built a full local AI assistant on 5080, 5060 Ti and 2x3090
Fabulous_Fact_606 · reddit · 2026-11-17
A Reddit user shared the full architecture of IVY, a self-built AI assistant running across a home PC (RTX 5080 + 5060 Ti) and a dual-RTX-3090 server, with a VPS hosting the web chat.
Key components
- Brain: local Qwen3.8-27B (Unsloth quant) via llama-swap, with DeepSeek and Claude models as cloud fallbacks; a bash CLI tool-loop as the engine, wired to web/voice through a Node gateway
- Memory: NVIDIA cuVS CAGRA GPU vector index over SQLite, 198,700 facts (73,300 active), Qwen3-Embedding-8B for embeddings and Qwen3-Reranker-4B for reranking
- Voice: Breeze TTS (int4), Qwen3-ASR, and NVIDIA TitaNet speaker ID
- Face: a Ditto-based voice-to-motion model plus a 3D Gaussian-splatting head (338,000 Gaussians), restyled per frame by a U-Net and streamed as H.264
- On-demand creative models for images, music, sound effects, and video
More from coding & agent
- LangChain hosts SF hands-on workshop on Managed Deep Agents with LangSmith — LangChain · 2026-11-18
- Vite+ 1.7.0 adds type-aware linting via tsrs, a Rust port of TypeScript 7 — cnakazawa · 2026-11-18
- Higgsfield's new AI film: 'The hardest part is making it not look like AI' — CurieuxExplorer · 2026-11-18
- Project triangle flips in the AI agent era: defining 'good' becomes the job — lak_luster · 2026-11-18
- The levels of agentic engineering: find what you still do by hand, then automate it — rseroter · 2026-11-18
- GitHub Next's Maggie Appleton on speculative design for prototyping software futures — JayAlammar · 2026-11-18