pqnld: A Rust Service That Turns Your Existing LLM Into a Closed-Set Decision Model via Logprobs
Difficult_Ad_2778 · reddit · 2026-10-14
pqnld (github.com/ErikBPF/pqnld) is a small Rust service that adds closed-set decisions to existing LLM deployments: give it context, questions and allowed answers, and it returns a choice with scores read from model logprobs rather than hallucinated confidence numbers. Options get single-token labels (with an experimental extended-label mode), sub-200ms answers on a Qwen3.8 27B NVFP4 deployment, HTTP and MCP interfaces, and no extra model or training needed — vLLM is the live-tested backend.
More from coding & agent
- First hands-on with Google Antigravity: three research subagents running in parallel — ChrisUniverse · 2026-10-14
- Serverless tool calling could slash agent costs 30x by running 1000 agents per sandbox — madhavsinghal_ · 2026-10-14
- Apex Matches Frontier Scores at Lower Cost but Overthinks, Says Aiden Bai — aidenybai · 2026-10-14
- Deploy Cursor Origin Apps to Cloudflare: Auto-Build, Preview or Ship on Every Change — dinasaur_404 · 2026-10-14
- AI proofs now span hundreds of thousands of Lean lines — verification becomes a second compute meter — demian_ai · 2026-10-14
- ViteConf 2026 kicks off Oct 15 with Evan You's 'Developer Tooling in the Agentic Era' keynote — cnakazawa · 2026-10-14