From 557ba762efa6d44e51d30696f179f43970ad8380 Mon Sep 17 00:00:00 2001 From: DreamCoder08 Date: Tue, 29 Sep 2026 15:52:43 -0500 Subject: [PATCH] chore(gga): use the codex provider pinned globally to gpt-6.1-sol The global gga config now forces codex with gpt-6.1-sol at medium reasoning for every repository. Align this project's .gga and drop the stale kilo notes. --- .gga | 18 +++++++----------- 1 file changed, 7 insertions(+), 11 deletions(-) diff --git a/.gga b/.gga index 23f1d3f5..d523b054 100644 --- a/.gga +++ b/.gga @@ -1,18 +1,14 @@ # Gentleman Guardian Angel Configuration -PROVIDER="kilo" +PROVIDER="codex" FILE_PATTERNS="*.sh,*.zsh,*.bash,.zshrc,.bashrc,.bash_profile,.tmux.conf" EXCLUDE_PATTERNS="*.log,*.bak,node_modules,target,.git" RULES_FILE="AGENTS.md" STRICT_MODE="false" -# Provider notes (2026-09-11): -# codex -> "You've hit your usage limit ... try again at Sep 14th, 2026 11:27 PM" -# claude -> "You've hit your weekly limit ... resets Sep 13, 5pm (America/Lima)" -# opencode-> "The usage limit has been reached" -# gemini -> requires a trusted workspace (GEMINI_CLI_TRUST_WORKSPACE=true) -# kilo -> works today (routes to DeepSeek); verified with `gga run` -# STRICT_MODE="false" keeps this gate advisory: a provider hiccup or an ambiguous -# reply never blocks a commit. Setting it to "true" makes gga block whenever the -# reply lacks exactly one "STATUS: PASSED|FAILED" line, which the kilo default -# model does not always emit. +# Provider: the global ~/.config/gga/config pins every repository to OpenAI Codex with +# GPT-6.1-Sol at medium reasoning (GGA_PROVIDER overrides this file and a shim sets the +# model), so this value only documents the intent and keeps `gga` working without it. +# STRICT_MODE="false" keeps the gate advisory: a provider hiccup or an ambiguous reply +# never blocks a commit. "true" blocks whenever the reply lacks exactly one +# "STATUS: PASSED|FAILED" line.