perf: parallelize tool server init and reduce LLM retry overhead#139

Closed

wangbinluo wants to merge 16 commits into

MiroMindAI:mainfrom

wangbinluo:dev_wbl

Commits on Feb 4, 2026

add is_final_retry flag to optimize last retry behavior (#111)
shawnlimn
authored

Commits on Mar 11, 2026

fix: scaffold-level fixes for tool calling errors

authored and
shawnlimn
committed
rebase to main and ruff format
shawnlimn
committed
added discard-all ctx mngt
shawnlimn
committed
updated the block list
shawnlimn
committed
mirothinker-1.7 release
shawnlimn
committed
updated readme
shawnlimn
committed
update license

xingxuanli
authored and
shawnlimn
committed
update table results

xingxuanli
authored and
shawnlimn
committed
updated readme

xingxuanli
authored and
shawnlimn
committed
update figure

xingxuanli
authored and
shawnlimn
committed

Commits on Mar 13, 2026

added HF link (#136)
shawnlimn
authored

Commits on Mar 17, 2026

perf: parallelize tool server init and reduce LLM retry overhead (#137)

wangbinluo
and
claude
committed
style: fix ruff formatting in manager.py

wangbinluo
and
claude
committed
revert: restore LLM retry parameters (max_retries=10, base_wait_time=30)

wangbinluo
and
claude
committed

Commits on Mar 18, 2026

perf: persistent MCP sessions — reuse subprocess across tool calls (#137)

wangbinluo
and
claude
committed