[Sync] opencode 2.x

README.md

@@ -132,8 +132,10 @@ Because configs are symlinks into the repo, **you almost never run `install.sh`
 | fl's Hyprland | overrides ~9 hypr files (`hyprland.conf`, keybinds, windowrules, theme colors, monitors); inherits the rest from `gui` |
 | Unique apps | in the owning host overlay (fl: llamacpp/evremap; lw: xremap) |
 
-`aichat`, `pi`, and `opencode` are **single shared configs** in `common` (LAN-first, with the cloud
+`aichat`, `pi`, and `opencode` share one config in `common` (LAN-first, with the cloud provider also
-provider also listed so any host can switch) — no per-host overrides.
+listed so any host can switch). `lw` carries a host overlay of all three that adds its local CPU
+llama.cpp models — mirror edits between the common and `lw` copies. The opencode tree is documented
+in `common/.config/opencode/README.md`.
 
 ## HyDE dependency (important for GUI hosts)
 

common/.config/fish/config.fish

@@ -5,10 +5,17 @@ set -gx EDITOR nvim
 set -gx PAGER less
 set -gx QT_FONT_DPI 96
 set -gx PI_SKIP_VERSION_CHECK 1
-# opencode GPT/Codex OAuth overlay — only where `opencode auth login` stored an OpenAI credential,
+# opencode GPT/Codex OAuth overlay (~/.config/opencode/openai-gpt.jsonc). opencode 2.x ignores
-# so hosts without the login don't get an unusable provider in the model picker.
+# OPENCODE_CONFIG / OPENCODE_CONFIG_CONTENT, but it merges a second global file, opencode.jsonc.
-if test -r ~/.local/share/opencode/auth.json; and grep -q '"openai"' ~/.local/share/opencode/auth.json
+# Link the overlay under that name only where `opencode auth login` stored an OpenAI credential,
-    set -gx OPENCODE_CONFIG "$HOME/.config/opencode/openai-gpt.jsonc"
+# so other hosts don't get an unusable provider; drop our link again if the login goes away.
+# (Alternative: ship the link from the host's dots overlay instead of this startup check.)
+set -l _oc_auth ~/.local/share/opencode/auth.json
+set -l _oc_link ~/.config/opencode/opencode.jsonc
+if test -r $_oc_auth; and grep -q '"openai"' $_oc_auth
+    test -e $_oc_link; or ln -s openai-gpt.jsonc $_oc_link
+else if test -L $_oc_link; and test (readlink $_oc_link) = openai-gpt.jsonc
+    rm $_oc_link
 end
 set -gx MANPAGER "nvim +Man!"
 # fzf's built-in walker is the one search tool that ignores .ignore files, so point it at

common/.config/opencode/README.md

@@ -0,0 +1,73 @@
+# opencode config
+
+Config tree for [opencode](https://opencode.ai) (2.x), tuned for **small local models** behind a
+llama.cpp router. Stowed into `~/.config/opencode/` as per-file symlinks; editing here edits the live
+config (restart opencode, or `opencode service restart`, to pick changes up — the 2.x background
+service caches config).
+
+## Layout
+
+| Path | What |
+|---|---|
+| `opencode.json` | Providers, permissions, primary agents, defaults. Written in the v1 schema, which 2.x still accepts. |
+| `openai-gpt.jsonc` | Optional GPT/Codex OAuth overlay (OpenAI provider + `opencode-openai-codex-auth` plugin). Not loaded by itself — see *GPT overlay*. |
+| `AGENTS.md` | Global rules prepended to every session. Same text as the pi config's `AGENTS.md`: terse, verify-before-asserting, small-model friendly. |
+| `agents/*.md` | Four subagents with real prompts and scoped permissions: `code-reviewer`, `test-runner` (hidden), `doc-writer` (edits docs only), `security-auditor` (hidden). |
+| `commands/*.md` | `/commit` (Conventional Commits message, never commits) and `/review` (read-only diff review), both on the `plan` agent. |
+| `plugins/verbose/` | 2.x TUI plugin: `<leader>v`, `/verbose`, or the palette entry toggles the *verbose view* — thinking blocks expanded and tool calls ungrouped — reading the live state from the built-in toggles so a mixed state converges. `server.ts` is a no-op half required by the loader. |
+| `themes/catppuccin-mocha.json` | Catppuccin Mocha (dark) / Latte (light) theme following the pi theme's semantic mapping. Built-in `catppuccin` exists too; this one matches pi. |
+| `tui.json` | TUI settings in the v1 schema (2.x compat): theme, leader timeout, stacked diffs, block cursor, attention notifications, a few extra keybinds. |
+
+The sandbox wrapper lives in `../fish/functions/opencodes.fish`: `opencodes` runs opencode under
+bwrap with the stowed config dirs, `~/.cache/opencode`, state and data bound, and only the needed
+environment passed through. Extra arguments go to opencode.
+
+## Providers and models
+
+Two OpenAI-compatible providers point at the same llama.cpp router: `llama.cpp` on the LAN and
+`duskadiy` through the public hostname (`apiKey` from `$DUSKADIY_API_KEY`). Model ids are the preset
+section names in `fl/.config/llamacpp/config.ini`; keep the lists in sync with it and with the pi and
+aichat configs. Both providers use a one-hour request timeout — the default five minutes cut off
+long local replies.
+
+Decisions that look odd but are deliberate:
+
+- **No `small_model`.** The router runs `--models-max 1`; any second model id evicts the resident one.
+  With `small_model` unset, title generation falls back to the session's model. Subagents carry no
+  pinned model for the same reason — the task tool uses the caller's model.
+- **No `temperature` / `top_p` from opencode.** Verified by capturing request bodies: only
+  `max_tokens` is sent, so the router presets own sampling.
+- **`reasoning: true`** only on the presets that actually think (no `reasoning-budget = 0`);
+  GLM also replays `reasoning_content` in history because its preset sets `reasoning-preserve`.
+- **`snapshot: false`** — file-change undo is not used. **`tool_output`** capped at 200 lines /
+  16 KiB for 24k-context models. **`compaction.reserved`/`prune`** are inert at this context size.
+- **`question: allow`** everywhere so agents ask instead of guessing.
+
+## Agents
+
+`auto` is the default agent and cycles with `build` and `plan` on Tab. It runs without prompts except
+for the serious set — git history rewriting and pushing, `rm`, killing processes, `systemctl`,
+network transfers, package installs — which ask, and `sudo`/`dd`/`mkfs`/secret-file reads, which are
+denied. Colors: build primary, plan info, auto yellow. All three cap agentic iterations at 30.
+
+Permission rules resolve **global first, agent rules appended, last match wins**, so an agent that
+wants a blanket rule must restate its own denies after it (the `auto` block does).
+
+## GPT overlay
+
+`openai-gpt.jsonc` is a template. opencode 2.x ignores `OPENCODE_CONFIG`, but it merges a second
+global file named `opencode.jsonc`. `fish/config.fish` links `opencode.jsonc → openai-gpt.jsonc`
+only on a host whose `~/.local/share/opencode/auth.json` holds an OpenAI credential (from
+`opencode auth login`), and removes the link again when the login is gone.
+
+## Host overlays
+
+`lw/.config/opencode/opencode.json` is a fork of the common file that adds the laptop's local CPU
+llama.cpp provider (`lwcpp`). Mirror every edit between the two.
+
+## Lifting pieces into a stock opencode
+
+Everything is plain files: copy `agents/`, `commands/`, `plugins/verbose/` or the theme into your own
+`~/.config/opencode/`. The plugin needs 2.x (it uses the `keymap.layer` / `dispatch` API and is
+discovered as a directory under `plugins/`). The permission blocks in `opencode.json` are independent
+of the providers and can be pasted as-is.

common/.config/opencode/opencode.json

@@ -1,6 +1,7 @@
 {
   "$schema": "https://opencode.ai/config.json",
   "model": "llama.cpp/gemma-4-26B-A4B-it-UD-IQ4_XS",
+  "default_agent": "auto",
   "instructions": ["CLAUDE.md", "AGENTS.md"],
   "share": "disabled",
   "autoupdate": "notify",
@@ -303,47 +304,6 @@
           "grep * .env*": "deny"
         }
       }
-    },
-    "code-reviewer": {
-      "description": "Reviews code for best practices and potential issues",
-      "prompt": "You are a code reviewer. Focus on security, performance, and maintainability.",
-      "mode": "subagent",
-      "permission": {
-        "edit": "deny"
-      }
-    },
-    "test-runner": {
-      "description": "Runs and interprets test suites",
-      "prompt": "You are a test runner. Focus on identifying failing tests and providing clear reproduction steps.",
-      "mode": "subagent",
-      "hidden": true,
-      "permission": {
-        "edit": "deny"
-      }
-    },
-    "doc-writer": {
-      "description": "Maintains project documentation and docstrings",
-      "prompt": "You are a technical writer. Focus on clarity, accuracy, and keeping documentation in sync with code changes.",
-      "mode": "subagent",
-      "permission": {
-        "edit": {
-          "*": "deny",
-          "**/*.md": "allow",
-          "**/*.mdx": "allow",
-          "**/*.markdown": "allow",
-          "**/*.txt": "allow",
-          "**/*.rst": "allow"
-        }
-      }
-    },
-    "security-auditor": {
-      "description": "Scans for security vulnerabilities and hardcoded secrets",
-      "prompt": "You are a security auditor. Focus on OWASP principles, secret detection, and dependency vulnerabilities.",
-      "mode": "subagent",
-      "hidden": true,
-      "permission": {
-        "edit": "deny"
-      }
     }
   }
 }

common/.config/opencode/plugins/verbose/server.ts

@@ -0,0 +1,8 @@
+// opencode v2 plugin "coja.verbose" — server half. Intentionally empty: all behaviour is in ./tui.ts.
+// A local plugin is a DIRECTORY under ~/.config/opencode/plugins/ whose `server`/`index` and `tui`
+// modules are resolved by name (packages/plugin/src/host.ts resolve()). No imports on purpose:
+// `define()` is an identity helper, and a dependency-free module needs no node_modules.
+export default {
+  id: "coja.verbose",
+  setup() {},
+}

common/.config/opencode/plugins/verbose/tui.ts

@@ -0,0 +1,62 @@
+// opencode v2 TUI plugin: one key toggles the "verbose view".
+//
+// v2 dropped the v1 tool-details / timestamps / generic-output toggles. The session view now has two
+// display toggles that matter: thinking blocks (session.thinking show|hide) and exploration grouping
+// (session.grouping auto|none — "auto" folds read/glob/grep tool calls into one collapsed group).
+// Verbose = thinking expanded AND tool calls shown individually. Both settings persist in the TUI config.
+//
+// State is read from the built-in commands' live titles (the only plugin-visible signal):
+//   session.toggle.thinking              → "Collapse thinking" while expanded, else "Expand thinking"
+//   session.toggle.exploration_grouping  → "Show tool calls individually" while grouped, else "Group related tool calls"
+// then only the toggles that differ from the target are dispatched, so a mixed state converges.
+//
+// Auto-discovered from ~/.config/opencode/plugins/verbose/ (server.ts + this file). Key configurable
+// via plugin options; default <leader>v. Also /verbose and a command-palette entry.
+const THINKING = "session.toggle.thinking"
+const GROUPING = "session.toggle.exploration_grouping"
+
+type Command = { readonly id?: string; readonly title?: string }
+type Context = {
+  readonly options?: Readonly<Record<string, unknown>>
+  readonly keymap: {
+    layer(input: () => { readonly commands?: readonly unknown[] }): void
+    dispatch(id: string, input?: string): void
+    commands(): readonly Command[]
+  }
+  readonly ui: { readonly toast: { show(o: { message: string; variant?: string; duration?: number }): void } }
+}
+
+export default {
+  id: "coja.verbose",
+  setup(context: Context) {
+    const key = typeof context.options?.key === "string" && context.options.key ? (context.options.key as string) : "<leader>v"
+    const title = (id: string) => context.keymap.commands().find((c) => c.id === id)?.title ?? ""
+
+    context.keymap.layer(() => ({
+      commands: [
+        {
+          id: "verbose.toggle",
+          title: "Toggle verbose view (expand thinking, ungroup tool calls)",
+          group: "Session",
+          bind: key,
+          palette: true,
+          slash: { name: "verbose" },
+          run() {
+            const thinkingTitle = title(THINKING)
+            const groupingTitle = title(GROUPING)
+            if (!thinkingTitle && !groupingTitle) {
+              context.ui.toast.show({ message: "Verbose view: open a session first", variant: "warning", duration: 2000 })
+              return
+            }
+            const thinkingVerbose = thinkingTitle.startsWith("Collapse")
+            const groupingVerbose = groupingTitle.startsWith("Group")
+            const on = !(thinkingVerbose && groupingVerbose)
+            if (thinkingTitle && thinkingVerbose !== on) context.keymap.dispatch(THINKING)
+            if (groupingTitle && groupingVerbose !== on) context.keymap.dispatch(GROUPING)
+            context.ui.toast.show({ message: on ? "Verbose view on" : "Verbose view off", variant: "info", duration: 1500 })
+          },
+        },
+      ],
+    }))
+  },
+}

common/.config/opencode/tui-plugins/verbose.ts

@@ -1,71 +0,0 @@
-// opencode TUI plugin: one key toggles the whole "verbose view".
-//
-// The session view keeps four independent display toggles in the TUI's persisted KV store
-// (~/.local/state/opencode/kv.json). Built-in commands only flip one each. This plugin reads all
-// four and drives them to the same state: if every one is on, turn all off; otherwise turn all on.
-// Writing the KV keys directly is what the built-in toggles do too (kv.signal is a reactive
-// store), so the open session re-renders immediately and the choice survives restarts.
-//
-// Declared in tui.json:  "plugin": [["./tui-plugins/verbose.ts", { "key": "<leader>v" }]]
-// (TUI plugins are never auto-discovered, and this dir is deliberately NOT ~/.config/opencode/plugins/,
-//  which the server-side plugin loader scans and would reject a TUI-only module.)
-// Also available as /verbose and from the command palette.
-import type { TuiPlugin, TuiPluginApi, TuiPluginModule } from "@opencode-ai/plugin/tui"
-
-// KV keys + values as used by packages/tui/src/routes/session/index.tsx and context/thinking.ts (v1.18.29)
-const KEYS = {
-  details: "tool_details_visibility", // boolean, default true
-  generic: "generic_tool_output_visibility", // boolean, default false
-  timestamps: "timestamps", // "show" | "hide", default "hide"
-  thinking: "thinking_mode", // "show" | "hide", default "hide"
-} as const
-
-const COMMAND = "verbose.toggle"
-
-function current(api: TuiPluginApi) {
-  return {
-    details: api.kv.get<boolean>(KEYS.details, true) === true,
-    generic: api.kv.get<boolean>(KEYS.generic, false) === true,
-    timestamps: api.kv.get<string>(KEYS.timestamps, "hide") === "show",
-    thinking: api.kv.get<string>(KEYS.thinking, "hide") === "show",
-  }
-}
-
-function apply(api: TuiPluginApi, on: boolean) {
-  api.kv.set(KEYS.details, on)
-  api.kv.set(KEYS.generic, on)
-  api.kv.set(KEYS.timestamps, on ? "show" : "hide")
-  api.kv.set(KEYS.thinking, on ? "show" : "hide")
-}
-
-const tui: TuiPlugin = async (api, options) => {
-  const opts = (options ?? {}) as Record<string, unknown>
-  const key = typeof opts.key === "string" && opts.key.length > 0 ? opts.key : "<leader>v"
-
-  api.keymap.registerLayer({
-    mode: "base",
-    commands: [
-      {
-        name: COMMAND,
-        title: "Toggle verbose view (tool details, generic output, thinking, timestamps)",
-        category: "Session",
-        namespace: "palette",
-        slashName: "verbose",
-        run() {
-          const state = current(api)
-          const on = !Object.values(state).every(Boolean)
-          apply(api, on)
-          api.ui.toast({
-            variant: "info",
-            message: on ? "Verbose view on" : "Verbose view off",
-            duration: 1500,
-          })
-        },
-      },
-    ],
-    bindings: [{ key, cmd: COMMAND, desc: "Toggle verbose view" }],
-  })
-}
-
-const plugin: TuiPluginModule & { id: string } = { id: "coja.verbose", tui }
-export default plugin

common/.config/opencode/tui.json

@@ -1,7 +1,6 @@
 {
   "$schema": "https://opencode.ai/tui.json",
   "theme": "lucent-orng",
-  "plugin": [["./tui-plugins/verbose.ts", { "key": "<leader>v" }]],
   "leader_timeout": 1500,
   "diff_style": "stacked",
   "cursor": { "style": "block", "blinking": false },
@@ -11,8 +10,6 @@
     "prompt_stash_pop": "<leader>o",
     "session_fork": "<leader>f",
     "messages_last_user": "<leader>k",
-    "display_thinking": "<leader>i",
+    "display_thinking": "<leader>i"
-    "tool_details": "<leader>d",
-    "session_toggle_timestamps": "<leader>w"
   }
 }

lw/.config/opencode/opencode.json

@@ -1,6 +1,7 @@
 {
   "$schema": "https://opencode.ai/config.json",
   "model": "llama.cpp/gemma-4-26B-A4B-it-UD-IQ4_XS",
+  "default_agent": "auto",
   "instructions": ["CLAUDE.md", "AGENTS.md"],
   "share": "disabled",
   "autoupdate": "notify",
@@ -321,47 +322,6 @@
           "grep * .env*": "deny"
         }
       }
-    },
-    "code-reviewer": {
-      "description": "Reviews code for best practices and potential issues",
-      "prompt": "You are a code reviewer. Focus on security, performance, and maintainability.",
-      "mode": "subagent",
-      "permission": {
-        "edit": "deny"
-      }
-    },
-    "test-runner": {
-      "description": "Runs and interprets test suites",
-      "prompt": "You are a test runner. Focus on identifying failing tests and providing clear reproduction steps.",
-      "mode": "subagent",
-      "hidden": true,
-      "permission": {
-        "edit": "deny"
-      }
-    },
-    "doc-writer": {
-      "description": "Maintains project documentation and docstrings",
-      "prompt": "You are a technical writer. Focus on clarity, accuracy, and keeping documentation in sync with code changes.",
-      "mode": "subagent",
-      "permission": {
-        "edit": {
-          "*": "deny",
-          "**/*.md": "allow",
-          "**/*.mdx": "allow",
-          "**/*.markdown": "allow",
-          "**/*.txt": "allow",
-          "**/*.rst": "allow"
-        }
-      }
-    },
-    "security-auditor": {
-      "description": "Scans for security vulnerabilities and hardcoded secrets",
-      "prompt": "You are a security auditor. Focus on OWASP principles, secret detection, and dependency vulnerabilities.",
-      "mode": "subagent",
-      "hidden": true,
-      "permission": {
-        "edit": "deny"
-      }
     }
   }
 }