refactor(server): proper Python package with tool-calling, tests, uv setup#43

Closed

easel wants to merge 31 commits into

Luce-Org:mainfrom

easel:feat/setup-results-uv

Commits on Apr 27, 2026

feat(dflash): integrate TQ3_0 KV cache type

mraxai
authored and
easel
committed
chore(deps): bump llama.cpp submodule to luce-dflash merge tip

authored and
easel
committed
docs: promote TQ3_0 to default KV cache across scripts and READMEs

authored and
easel
committed
docs(dflash): bump long-context example from 128K to 256K

authored and
easel
committed
docs(dflash): drop stale "model reload per turn" limit

authored and
easel
committed
docs(dflash): drop Q5_K_M / Q6_K target roadmap bullet

authored and
easel
committed
docs: advertise TurboQuant TQ3_0 KV cache and 256K ceiling in root README

authored and
easel
committed
docs: clarify Blackwell/GB10 support alongside RTX 3090 reference

Alessandro Puppo
authored and
easel
committed
Update README.md

davide221
authored and
easel
committed
docs: add RTX 5090 / sm_120 to requirements and build comment

easel
and
claude
committed
results: add TQ3_0 short-context numbers for RTX 5090 Laptop

easel
and
claude
committed

Commits on Apr 29, 2026