connor/dotfiles
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

opencode: remove oh-my-openagent from all profiles

Browse Source 
Drops the plugin entry and the agent/* model aliases, sets plain default
models, and removes the OMO roster swap from opencode.zsh.
This commit is contained in:
connor committed 2026-10-29 10:29:34 -04:00
1 parent 3fb10dd4c3
commit 6b9b1697f2
5 files changed
+10 -536
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

common/claude/CLAUDE.md
+6 -8
View File
@@ -3,7 +3,7 @@
## LLM / inference stack
Connor runs a self-hosted inference stack: LiteLLM proxy in front of a local GPU
model (llama-swap), OpenAI, Anthropic, and Moonshot/Kimi, with OpenCode + OMO on
model (llama-swap), OpenAI, Anthropic, and Moonshot/Kimi, with OpenCode on
top. **Before touching anything under `~/config/litellm/`, `~/.config/opencode/`,
or the model routing, read `~/guides/llm/robots.md`.** It carries the design
decisions, the measured data, and — more importantly — the specific wrong
@@ -79,13 +79,11 @@ The five things most likely to waste an hour if you don't know them:
VRAM buys prefill, not decode. Don't recommend a bigger GPU without checking
which wall you're actually against.
9. **The OMO agent aliases embed their model family on purpose** —
`personal/atlas-kimi-k2.7-code`, not `personal/atlas`. OMO picks its
per-model prompt by string-matching the *last path segment* of the model
name, so a "tidy-up" rename silently disables every Kimi/GPT-specific prompt
and guard with no error. **Never shorten them.** `local/librarian` and
`local/explore` are deliberately left un-suffixed — `isGptModel` matches a
bare `"gpt"` substring and `gpt-oss-20b` would trip it.
9. **OMO (oh-my-openagent) was removed from every OpenCode profile on
2026-10-29.** It injected its own skills and prompts into non-OMO agents
and used a lot of context. The `agent/*` and `personal/atlas-*` style
aliases in LiteLLM were for OMO's roster and are unused by OpenCode now;
OMO rewrote them at request time, so they 400 if called directly.
Cost work: read `model_group`, never `model`, to attribute a request. Cost is
driven by cache-read volume and context size far more than by model choice.
common/opencode/personal.json
+1 -158
View File
@@ -230,160 +230,6 @@
"cache_write": 0.0
}
},
"agent/librarian-glm-5.3-flash": {
"name": "librarian -> glm-5.3-flash",
"limit": {
"context": 128000,
"output": 8192
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/librarian-glm-5.3-flash"
},
"agent/explore-gpt-5.6-luna": {
"name": "explore -> local/explore",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 32000.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/explore-gpt-5.6-luna"
},
"agent/sisyphus-glm-5.3": {
"name": "sisyphus -> glm-5.3",
"limit": {
"context": 200000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/sisyphus-glm-5.3"
},
"agent/atlas-glm-5.3": {
"name": "atlas -> glm-5.3",
"limit": {
"context": 200000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/atlas-glm-5.3"
},
"agent/prometheus-glm-5.3": {
"name": "prometheus -> glm-5.3",
"limit": {
"context": 200000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/prometheus-glm-5.3"
},
"agent/metis-glm-5.3": {
"name": "metis -> glm-5.3",
"limit": {
"context": 200000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/metis-glm-5.3"
},
"agent/sisyphus-junior-glm-5.3-flash": {
"name": "sisyphus-junior -> glm-5.3-flash",
"limit": {
"context": 150000,
"output": 32000
},
"cost": {
"input": 0.0,
"output": 0.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "personal/sisyphus-junior-glm-5.3-flash"
},
"agent/momus-gpt-5.5": {
"name": "momus -> gpt-5.5",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "personal/momus-gpt-5.5"
},
"agent/gpt-5.6-sol-hephaestus": {
"name": "hephaestus -> gpt-5.6-sol",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 4.0,
"output": 64000.0,
"cache_read": 0.4,
"cache_write": 5.0
},
"id": "personal/hephaestus-gpt-5.6-sol"
},
"agent/oracle-gpt-5.5": {
"name": "oracle -> gpt-5.5",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "personal/oracle-gpt-5.5"
},
"agent/looker-gpt-5.5": {
"name": "looker -> gpt-5.4 (vision)",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "personal/looker-gpt-5.5"
},
"muse-glimmer-30b-low": {
"name": "Muse Glimmer 30B (local, low reasoning)",
"limit": {
@@ -442,11 +288,8 @@
"enabled_providers": [
"litellm"
],
"model": "litellm/agent/atlas-glm-5.3",
"model": "litellm/qwen3.8-27b-gsq-s-med",
"small_model": "litellm/local/librarian",
"plugin": [
"oh-my-openagent"
],
"agent": {
"dev": {
"model": "litellm/qwen3.8-27b-gsq-s-med"
common/opencode/work-strict.json
+1 -158
View File
@@ -191,160 +191,6 @@
"cache_write": 0.0
}
},
"agent/librarian-gpt-5.6-luna": {
"name": "librarian -> local/librarian",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 32000.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "work/librarian"
},
"agent/explore-gpt-5.6-luna": {
"name": "explore -> local/explore",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 32000.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "work/explore"
},
"agent/oracle-gpt-5.5": {
"name": "oracle -> work/oracle",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/oracle"
},
"agent/looker-gpt-5.5": {
"name": "looker -> work/looker",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/looker"
},
"agent/momus-gpt-5.5": {
"name": "momus -> work/momus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/momus"
},
"agent/sisyphus-junior-gpt-5.6-luna": {
"name": "sisyphus-junior -> work/sisyphus-junior",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 32000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/sisyphus-junior"
},
"agent/gpt-5.6-sol-hephaestus": {
"name": "hephaestus -> work/hephaestus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 1.75,
"output": 64000.0,
"cache_read": 0.175,
"cache_write": 0.0
},
"id": "work/hephaestus"
},
"agent/atlas-gpt-5.5": {
"name": "atlas -> work/atlas (gpt-5.4)",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/atlas"
},
"agent/prometheus-gpt-5.6-sol": {
"name": "prometheus -> work/prometheus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 4.0,
"output": 64000.0,
"cache_read": 0.4,
"cache_write": 5.0
},
"id": "work/prometheus"
},
"agent/sisyphus-gpt-5.6-sol": {
"name": "sisyphus -> work/sisyphus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 4.0,
"output": 64000.0,
"cache_read": 0.4,
"cache_write": 5.0
},
"id": "work/sisyphus"
},
"agent/metis-gpt-5.5": {
"name": "metis -> work/metis",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/metis"
},
"chatgpt/gpt-5.6-terra-low": {
"name": "GPT-5.6 Terra (reasoning_effort=low)",
"limit": {
@@ -364,11 +210,8 @@
"enabled_providers": [
"litellm"
],
"model": "litellm/agent/atlas-gpt-5.5",
"model": "litellm/chatgpt/gpt-5.6-terra-low",
"small_model": "litellm/local/librarian",
"plugin": [
"oh-my-openagent"
],
"permission": {
"read": {
"*": "ask",
common/opencode/work.json
+1 -158
View File
@@ -204,160 +204,6 @@
"cache_write": 0.0
}
},
"agent/librarian-gpt-5.6-luna": {
"name": "librarian -> local/librarian",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 32000.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "work/librarian"
},
"agent/explore-gpt-5.6-luna": {
"name": "explore -> local/explore",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 0.0,
"output": 32000.0,
"cache_read": 0.0,
"cache_write": 0.0
},
"id": "work/explore"
},
"agent/oracle-gpt-5.5": {
"name": "oracle -> work/oracle",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/oracle"
},
"agent/looker-gpt-5.5": {
"name": "looker -> work/looker",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/looker"
},
"agent/momus-gpt-5.5": {
"name": "momus -> work/momus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/momus"
},
"agent/sisyphus-junior-gpt-5.6-luna": {
"name": "sisyphus-junior -> work/sisyphus-junior",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 32000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/sisyphus-junior"
},
"agent/gpt-5.6-sol-hephaestus": {
"name": "hephaestus -> work/hephaestus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 1.75,
"output": 64000.0,
"cache_read": 0.175,
"cache_write": 0.0
},
"id": "work/hephaestus"
},
"agent/atlas-gpt-5.5": {
"name": "atlas -> work/atlas (gpt-5.4)",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/atlas"
},
"agent/prometheus-gpt-5.6-sol": {
"name": "prometheus -> work/prometheus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 4.0,
"output": 64000.0,
"cache_read": 0.4,
"cache_write": 5.0
},
"id": "work/prometheus"
},
"agent/sisyphus-gpt-5.6-sol": {
"name": "sisyphus -> work/sisyphus",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 4.0,
"output": 64000.0,
"cache_read": 0.4,
"cache_write": 5.0
},
"id": "work/sisyphus"
},
"agent/metis-gpt-5.5": {
"name": "metis -> work/metis",
"limit": {
"context": 272000,
"output": 128000
},
"cost": {
"input": 2.5,
"output": 64000.0,
"cache_read": 0.25,
"cache_write": 0.0
},
"id": "work/metis"
},
"chatgpt/gpt-5.6-terra-low": {
"name": "GPT-5.6 Terra (reasoning_effort=low)",
"limit": {
@@ -377,11 +223,8 @@
"enabled_providers": [
"litellm"
],
"model": "litellm/agent/atlas-gpt-5.5",
"model": "litellm/chatgpt/gpt-5.6-terra-low",
"small_model": "litellm/local/librarian",
"plugin": [
"oh-my-openagent"
],
"permission": {
"read": {
"*": "ask",
common/zsh/opencode.zsh
+1 -54
View File
@@ -70,55 +70,6 @@ _oc_launch() {
return 1
fi
# ---- install the profile's OMO roster --------------------------------
# WHY THIS EXISTS. OMO reads exactly ONE config, resolveHomeDir($HOME) +
# "/.omo/omo.jsonc". There is no path override: `profiles`/OMO_PROFILE are in
# the new schema but the OLD loader rejects them, and a config the loaders
# reject fails SILENTLY (agent model bindings just stop applying). Setting
# HOME per profile would relocate opencode's own data and cache dirs too.
#
# It matters because OMO picks its per-model prompt by string-matching the
# LAST PATH SEGMENT of the model name, so the alias has to name the model
# family (see guides/llm/robots.md 5.1). Personal runs Kimi on the thinking
# and worker roles; work may not (PRC exclusion). One shared alias name
# therefore cannot be correct for both -- work was left running Kimi prompts
# on GPT models, 6 of 11 agents mismatched.
#
# The opencode-side fix does NOT work and was tested before this was written:
# OMO builds protectedBuiltinAgentNames from its own agent keys and runs
# filterProtectedAgentOverrides over opencode's `agent` config, so an
# `agent.sisyphus.model` in a profile config is silently DROPPED.
#
# So the roster is swapped here instead. Only omo.jsonc is touched;
# codegraph/, lsp-daemon/ and runtime/ under ~/.omo stay shared.
#
# RACE: two sessions in different profiles started within the same instant
# could read the other's roster. The window is startup-only (OMO reads the
# file once), and the failure is BOUNDED AND LOUD -- the LiteLLM virtual key,
# not this file, is the policy boundary, so a work session that ends up with
# the personal roster gets a 403 on the Kimi models rather than using them.
local roster
case "$profile" in
personal) roster=personal ;;
work|work-strict) roster=work ;;
local) roster= ;; # local.json loads no OMO plugin
esac
if [[ -n "$roster" ]]; then
local src="$HOME/.omo/profiles/${roster}.jsonc"
if [[ ! -r "$src" ]]; then
print -u2 "oc: OMO roster '$src' is missing."
print -u2 " Refusing to launch -- the wrong roster silently mis-prompts every agent."
return 1
fi
# Atomic: never leave a half-written omo.jsonc if this is interrupted.
if ! { cp -- "$src" "$HOME/.omo/.omo.jsonc.$$" &&
mv -f -- "$HOME/.omo/.omo.jsonc.$$" "$HOME/.omo/omo.jsonc" }; then
rm -f -- "$HOME/.omo/.omo.jsonc.$$"
print -u2 "oc: could not install the $roster OMO roster."
return 1
fi
fi
# Make the active mode obvious: banner, terminal title, and a prompt marker
# that survives for the rest of the shell if the profile is exported.
local colour
@@ -166,11 +117,7 @@ _oc_launch() {
# have received." This is a preference, not a fix -- REMOVE THESE TWO KEYS to
# get Plan Mode back.
#
# Note this is NOT an OMO feature, though it looks like one. OMO's agents
# carry only description/model/tools and cannot trigger on prompt text; the
# `plan` agent is native to opencode. Disabling OMO would not have stopped it,
# and would have broken the default model -- `agent/atlas-glm-5.3` is rewritten
# by the OMO plugin at request time and 400s at LiteLLM without it.
# The `plan` agent is native to opencode, not part of a plugin.
#
# Personal is still wrapped under the bare name so `opencode` does the
# expected thing and gets a key.