Repository Issue Activity (beta)

#3737 User selected reasoning effort it not being passed to local LLM services (lmstudio etc)

#3728 QoL - Display last user prompt when scrolling it offscreen "sticky prompt".

#3734 Agent loop: empty assistant message (no content, no tool_calls) poisons the session; all subsequent requests fail with 400 Bad Request

#3733 A single uncapped 1.7MB tool result deadlocked a 131K-window session (412K-token request, unrecoverable)

#3732 Making Crush easy to install on Windows: a short list of small fixes

#1698 Corrupted signature in Gemini Flash 3

#3729 Crush is not passing `top_k` to the LLM service

#3726 QoL - Small model UI display

#3408 Tool call failed: arguments were not valid JSON. Please check your tool call format and try again.

#3706 OpenCode Go will require an `x-opencode-session` header starting Sep 6

#3060 Session branches

#3689 v0.92 sending invalid content to llama.cpp

#745 Native mouse scrolling

#3658 Commit 0f403fe introduced a hardcoded timeout

#3701 MCP OAuth: Sentry sign-in fails on an unadvertised RFC 9207 iss

#3714 Warning: MallocStackLogging: can't turn off malloc stack logging because it was not enabled.

#3687 Error using GPT Luna model from Opencode ZEN

#3707 Use Copilot model metadata to select Responses or Chat Completions

#3709 Declare per-model API transport as catalog metadata (beyond Copilot)

#3681 README License section does not mention FSL Competing Use

#3480 Feature request: cap/truncate MCP tool results before they enter model context

#3691 Missing models with `with disable_provider_auto_update: true`

#3695 sessions.updated_at not bumped on message insert

#3671 Headless `crush run` silently ignores the configured large model and falls back to the first `can_reason` catalog entry

#2965 Optimize startup speed.