LLM Explainer v1.7.3

Features

• Inside-out recursive explain — "Explain function with LLM (recursively)" now explains a function's direct callees first and the target function last, deferred until every callee has finished (even with multiple parallel servers), so the target is analyzed with its callees' real names/signatures already known instead of sub_XXXXXXXX stubs.
• Compact caller call-site snippets — REQUEST_CALLERS now returns just the call expression plus each argument's Hex-Rays-inferred type, instead of each caller's entire body, which could burn most of the context budget on a large caller for little benefit. Falls back to REQUEST_CODE for a caller's full body when the snippet isn't enough.
• Investigate & Reconsider Name(s) — a called-function rename the model proposed without having actually seen that function's code was previously just silently dropped (logged only to the Output window). The single-function explain dialog now surfaces those as a row with an "Investigate & Reconsider Name(s)" button that fetches the callee's code and asks the model to confirm, revise, or drop the name.

Fixes

• SUGGESTED_NAME parsing now tolerates the model phrasing it as <old> -> <new> instead of just the new name (mirrors the existing fix for SUGGESTED_SIGNATURE) — previously this silently left the rename checkbox disabled. Prompt also reinforced against that format.