vramble: open-source VRAM scheduler auto-swaps local AI models on one workstation
FeydRowan · reddit · 2026-10-13
Reddit user FeydRowan released vramble (GitHub: ValerioDolci/vramble, MIT license), an open-source tool that fixes VRAM contention on local AI workstations.
- Problem: local LLMs, ComfyUI, TTS and coding agents each want all available VRAM
- Solution: vramble acts as a request dispatching layer that automatically loads/unloads models and manages a priority queue — no more manual model swapping (or burning Claude tokens to orchestrate)
- Example from the diagram: on a 16GB card with one model using 15GB, four services queue for the card; handover takes 4.7s and waiters are told who holds the card and for how long
- Bonus test: Qwen 27B at 200+ t/s performed better than expected
The project is new, contributions welcome, no guarantee it works out of the box on every machine.
More from coding & agent
- Indie dev ships hosted scraping MCP: pay nothing when sites block you, 1000 free credits — SidMishraDev · 2026-10-13
- Cursor's new pricing: Claude and other third-party models billed at API rates, plans include only Grok — iannuttall · 2026-10-13
- Dev: vibe coders with Claude can't yet beat 15-year principal engineers — but he expects yes within a year — facontidavide · 2026-10-13
- Open-Source MCP Security Gate Blocks Prompt Injections in Under 5ms — EstablishmentTough18 · 2026-10-13
- Open-source Claude skill writes working prompts on first try for 18+ AI tools — nikola_mr64990 · 2026-10-13
- When does an internal AI agent tool deserve to become a standalone app? — Moist_Deal_9995 · 2026-10-13