miii-agent 0.1.11 → 0.1.13

package/README.md

@@ -1,26 +1,38 @@
-# miii
+<h1 align="center">miii</h1>
 
-> small · simple · smart · strategic · semantic
->
-> Your code never leaves your machine. No API keys. No cloud. No bullshit.
+<p align="center">
+  <strong>Small. Simple. Smart. Strategic. Semantic.</strong>
+</p>
 
-**miii** is a local-first AI coding agent that lives in your terminal. Powered by [Ollama](https://ollama.com), it reads your code, writes features, runs tests, and fixes bugs — entirely on your hardware, at native speed.
+<p align="center">
+  A local-first AI coding agent that lives in your terminal.<br>
+  Your code never leaves your machine. No API keys. No cloud. No bullshit.
+</p>
 
-[![npm](https://img.shields.io/npm/v/miii-agent)](https://www.npmjs.com/package/miii-agent)
-[![license](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
-[![node](https://img.shields.io/badge/node-%3E%3D18-brightgreen)](https://nodejs.org)
+<p align="center">
+  <a href="https://www.npmjs.com/package/miii-agent"><img src="https://img.shields.io/npm/v/miii-agent" alt="npm version"></a>
+  <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg" alt="license"></a>
+  <a href="https://nodejs.org"><img src="https://img.shields.io/badge/node-%3E%3D18-brightgreen" alt="node version"></a>
+  <a href="https://ollama.com"><img src="https://img.shields.io/badge/powered%20by-Ollama-black" alt="powered by Ollama"></a>
+</p>
 
----
+miii is a local-first AI coding agent that lives in your terminal. Powered by [Ollama](https://ollama.com), it reads your code, writes features, runs tests, and fixes bugs — entirely on your hardware, at native speed.
 
-## The name
+---
 
-**miii** stands for five principles it's built around:
+## Contents
 
-- **small** — tight codebase, no bloat. You can read the whole thing.
-- **simple** — no API keys, no accounts, no config ceremony. Just run it.
-- **smart** — decomposes problems and verifies its own work like an engineer.
-- **strategic** — plans before it acts; tools are gated, paths are confined.
-- **semantic** — works from the meaning of your code, not blind text matching.
+- [Demo](#demo)
+- [The Local-First Advantage](#the-local-first-advantage)
+- [Core Philosophy](#core-philosophy)
+- [Quick Start](#quick-start)
+- [Interaction Guide](#interaction-guide)
+- [Technical Deep Dive](#technical-deep-dive)
+- [Configuration](#configuration)
+- [System Architecture](#system-architecture)
+- [Development](#development)
+- [Project Status](#project-status)
+- [License](#license)
 
 ---
 
@@ -30,25 +42,45 @@
 
 ---
 
-## Why miii?
+## The Local-First Advantage
+
+Most AI coding tools are wrappers around cloud APIs. They are slow, expensive, and require you to trust your private codebase to a third-party server.
 
-Most AI coding tools are wrappers around cloud APIs. They're slow, expensive, and send your private code to someone else's server.
+miii flips the script:
 
-miii is different:
+- **Absolute Privacy** — Powered by Ollama. Your code stays on your disk, period.
+- **Zero Friction** — No API keys, no billing, no accounts. Just `miii`.
+- **True Agency** — miii doesn't just chat; it decomposes problems, invokes tools, and verifies results like a senior engineer.
+- **Native Performance** — No network round-trips. Latency is limited by your GPU, not a CDN.
 
-- **Local-first** — Powered by Ollama. Your code stays on your disk, period.
-- **Zero ceremony** — No API keys. No billing. No accounts. Just `miii`.
-- **Actually agentic** — miii doesn't just chat. It decomposes problems, calls tools, and verifies results like an engineer would.
-- **Fast** — No network round-trips. Response time is limited by your GPU, not a CDN.
+|                | Cloud AI agents          | **miii**                     |
+|----------------|--------------------------|------------------------------|
+| Your code      | Sent to a third party    | Never leaves your machine    |
+| Cost           | Per-token billing        | Free — runs on your hardware |
+| Setup          | API keys, accounts       | `npm i -g miii-agent`        |
+| Offline        | No                       | Yes                          |
+| Latency        | Network + queue          | Your GPU only                |
 
 ---
 
-## Installation
+## Core Philosophy
+
+miii is built on five foundational principles:
+
+- **small** — A tight, bloat-free codebase. You can read the entire project in an afternoon.
+- **simple** — No configuration ceremony. Install and run.
+- **smart** — Decomposes complex tasks and verifies its own work.
+- **strategic** — Plans before acting; tools are gated and paths are confined.
+- **semantic** — Operates on the meaning of your code, not blind text matching.
+
+---
+
+## Quick Start
 
 ### Prerequisites
 
 - **Node.js** ≥ 18
-- **Ollama** running locally — [install here](https://ollama.com/download)
+- **Ollama** running locally — [Download here](https://ollama.com/download)
 - A coding model pulled locally:
 
 ```bash
@@ -57,25 +89,18 @@ ollama pull qwen2.5-coder:14b
 ollama pull deepseek-coder-v2
 ```
 
-### Install miii
+### Installation & Launch
 
 ```bash
 npm install -g miii-agent
-```
-
-### Launch
-
-```bash
 miii
 ```
 
-That's it.
-
 ---
 
-## Usage
+## Interaction Guide
 
-Once inside the TUI, just type naturally:
+Inside the TUI, interact naturally:
 
 ```
 > refactor the auth module to use async/await
@@ -92,58 +117,53 @@ Once inside the TUI, just type naturally:
 | `/models` | Switch active Ollama model |
 | `/clear` | Reset conversation history |
 | `Esc` | Stop current generation or tool run |
+| `Ctrl+O` | Toggle full tool output view |
 | `Ctrl+C` | Quit |
 
 ---
 
-## Configuration
-
-Settings live in `~/.miii/config.json` and are created on first run.
-
-```json
-{
-  "model": "qwen2.5-coder:14b",
-  "ollamaHost": "http://localhost:11434",
-  "effort": "medium"
-}
-```
-
-| Field | Description | Values |
-|-------|-------------|--------|
-| `model` | Default Ollama model | any `ollama list` model |
-| `ollamaHost` | Ollama API endpoint | URL string |
-| `effort` | Controls temperature & limits | `low` \| `medium` \| `high` |
-
----
+## Technical Deep Dive
 
-## Capabilities
+### Capabilities
 
-miii ships with a built-in tool suite the agent can invoke autonomously:
+miii ships with a built-in tool suite that the agent invokes autonomously:
 
-| Tool | What it does |
-|------|-------------|
+| Tool | Function |
+|------|----------|
 | `read_file` | Read any file in your workspace |
 | `write_file` | Create new files |
-| `edit_file` | Precise string-level edits (no rewrites) |
+| `edit_file` | Precise string-level edits with whitespace tolerance (no rewrites) |
 | `glob` | Pattern-match files across the project |
 | `grep` | Regex search across files |
 | `run_bash` | Execute shell commands |
 
-Every sensitive operation is gated by a permission system — you approve what the agent can touch, and "always" approvals persist to `~/.miii/permissions.json` so you're never asked twice. File tools are confined to your working directory; `../` traversal and absolute paths outside it are rejected.
+**Security & Safety:** Every sensitive operation is gated by a permission system. You approve what the agent can touch, and "always" approvals persist to `~/.miii/permissions.json`. File tools are strictly confined to your working directory; `../` traversal and absolute paths outside the workspace are rejected.
 
----
+### Lossless Output Spill
 
-## Checking your setup
+Big tool outputs (like 50K-line test logs) usually get truncated, leaving the model to guess. miii doesn't truncate; it **spills**.
+
+When a tool result exceeds the inline budget (~10K bytes), the full output is written to `~/.miii/output/<id>.txt`. Only a head + tail **preview** is inlined, followed by a pointer:
+
+```
+[command output truncated: 5184 lines / 412900 bytes.
+ Full output at ~/.miii/output/9f3a1c.txt — read it with
+ read_file offset/limit to see the elided middle.]
+```
 
-miii is model-agnostic — but not every local model can actually drive an agent. A model that can't emit clean tool calls will chat at you instead of editing files. `miii doctor` tells you which of *your* installed models are up to the job, before you waste time wondering why nothing happens.
+If the model needs the elided middle, it pages through it using `read_file` ranged reads. Nothing is ever lost. Spill files are garbage-collected after 24 hours.
+
+### The Model Doctor
+
+Not every local model can drive an agent. A model that cannot emit clean tool calls will simply chat at you instead of editing files. `miii doctor` validates your installed models against concrete engineering tasks.
 
 ```bash
 miii doctor                     # check every local model (from `ollama list`)
 miii doctor qwen2.5-coder:7b    # check one model
-miii doctor gemma4:e4b grep     # one model, only scenarios matching "grep"
+miii doctor gemma4:e4b grep     # check one model against "grep" scenarios
 ```
 
-It runs the real agent against a handful of concrete tasks (edit a file, read-and-answer, create a file, locate a definition) and checks the *outcome* — did the file actually change, was the answer right — then prints a verdict per model:
+It verifies outcomes (did the file actually change?) and prints a verdict:
 
 ```
 === qwen3-coder ===
@@ -157,11 +177,29 @@ PASS  grep-locate         ...
   → gemma4:e4b: 1/4 — not recommended — weak tool-calling
 ```
 
-With more than one model it also prints a compatibility matrix (`+` pass, `.` fail). Cloud models are skipped by default; name one explicitly to include it. If a model comes back `marginal` or `not recommended`, pull a stronger coding model and try again.
+---
+
+## Configuration
+
+Settings live in `~/.miii/config.json` and are created on first run.
+
+```json
+{
+  "model": "qwen2.5-coder:14b",
+  "ollamaHost": "http://localhost:11434",
+  "effort": "medium"
+}
+```
+
+| Field | Description | Values |
+|-------|-------------|--------|
+| `model` | Default Ollama model | any `ollama list` model |
+| `ollamaHost` | Ollama API endpoint | URL string |
+| `effort` | Controls temperature & limits | `low` \| `medium` \| `high` |
 
 ---
 
-## Architecture
+## System Architecture
 
 ```mermaid
 graph TD
@@ -195,6 +233,10 @@ graph TD
         ReadFile -.-> Confine["Path Confinement\n(tools/paths.ts)"]
         WriteFile -.-> Confine
         EditFile -.-> Confine
+        RunBash -->|"large output"| SpillMod["Output Spill\n(tools/spill.ts)"]
+        Grep -->|"large output"| SpillMod
+        Glob -->|"large output"| SpillMod
+        SpillMod -.->|"head+tail preview\n+ read pointer"| ToolRegistry
     end
 
     Adapter -->|"HTTP streaming"| Ollama["Ollama\n(local LLM server)"]
@@ -206,16 +248,21 @@ graph TD
     subgraph Storage ["Local Storage"]
         Config["~/.miii/config.json\n(model, host, effort)"]
         Rules["~/.miii/permissions.json\n(saved allow rules)"]
+        Spill["~/.miii/output/\n(spilled tool output)"]
     end
 
     App -.->|"reads"| Config
     Policy -.->|"reads / persists 'always'"| Rules
+    SpillMod -.->|"writes full output"| Spill
+    ReadFile -.->|"pages elided middle\n(offset/limit)"| Spill
 ```
 
 ---
 
 ## Development
 
+### Setup
+
 ```bash
 git clone https://github.com/maruakshay/miii-cli.git
 cd miii-cli
@@ -223,6 +270,8 @@ npm install
 npm run dev
 ```
 
+### Build & Test
+
 ```bash
 npm run build       # production build
 npm run start       # run built output
@@ -230,11 +279,11 @@ npm run typecheck   # type-check src + eval
 npm run eval        # run the eval harness as a CI / regression gate
 ```
 
-The eval harness lives in `eval/` and powers `miii doctor`. As `npm run eval` it doubles as a regression gate — it exits non-zero if any model fails any scenario, so a prompt or tool change that regresses a baseline model is caught in CI. Same engine, two doors: `miii doctor` for users checking their setup, `npm run eval` for maintainers gating changes.
+The eval harness in `eval/` powers `miii doctor` and serves as a regression gate. If `npm run eval` exits non-zero, a prompt or tool change has regressed a baseline model.
 
-### Testing the `miii` command against your local changes
+### Testing Local Changes
 
-The global `miii` command points at whatever was last installed with `npm install -g miii-agent` — **not** your working tree. After editing source, the global binary is stale, so `miii` (and `miii doctor`) will run the old code and may appear to ignore your changes (e.g. printing the wrong model). Two ways to run your local build:
+The global `miii` command points to the last installed version of `miii-agent`. To run your local working tree:
 
 ```bash
 node dist/cli.js doctor <model>   # run the freshly built output directly
@@ -242,11 +291,13 @@ node dist/cli.js doctor <model>   # run the freshly built output directly
 npm run build && npm link         # point the global `miii` at this repo
 ```
 
-`npm link` symlinks the global `miii` to `dist/cli.js` in this repo, so each `npm run build` is picked up automatically. Restore the published version later with `npm install -g miii-agent`. Note: `npm run dev` / `npm run start` always run the current source and never have this staleness problem.
+`npm link` symlinks the global `miii` to `dist/cli.js` in this repo. Restore the published version later with `npm install -g miii-agent`.
+
+---
 
 ## Project Status
 
-MVP. Core agent loop works. Actively refining tool execution, streaming, and the permission model. PRs welcome — fork it, break it, improve it.
+**MVP.** Core agent loop is stable. Actively refining tool execution, streaming, and the permission model. PRs are welcome — fork it, break it, improve it.
 
 ---
 

package/dist/cli.js

@@ -170,6 +170,39 @@ var init_paths = __esm({
   }
 });
 
+// src/tools/verifyHint.ts
+function verifyHint(path) {
+  const ext = path.slice(path.lastIndexOf(".") + 1).toLowerCase();
+  const cmds = {
+    ts: `npx tsc --noEmit`,
+    tsx: `npx tsc --noEmit`,
+    js: `node --check ${path}`,
+    jsx: `node --check ${path}`,
+    mjs: `node --check ${path}`,
+    cjs: `node --check ${path}`,
+    py: `python -m py_compile ${path}`,
+    go: `go build ./...`,
+    rs: `cargo check`,
+    rb: `ruby -c ${path}`,
+    php: `php -l ${path}`,
+    sh: `bash -n ${path}`,
+    bash: `bash -n ${path}`,
+    c: `gcc -fsyntax-only ${path}`,
+    h: `gcc -fsyntax-only ${path}`,
+    cpp: `g++ -fsyntax-only ${path}`,
+    cc: `g++ -fsyntax-only ${path}`,
+    java: `javac -d /tmp ${path}`
+  };
+  const cmd2 = cmds[ext];
+  if (!cmd2) return "";
+  return ` Now verify via run_bash: ${cmd2} \u2014 fix any errors it reports before continuing.`;
+}
+var init_verifyHint = __esm({
+  "src/tools/verifyHint.ts"() {
+    "use strict";
+  }
+});
+
 // src/tools/edit_file.ts
 import { readFileSync as readFileSync3, writeFileSync as writeFileSync3 } from "fs";
 function similarity(a, b) {
@@ -182,6 +215,35 @@ function similarity(a, b) {
   for (let i = 0; i < Math.min(x.length, y.length); i++) if (x[i] === y[i]) same++;
   return same / len;
 }
+function fuzzyRange(src, old_str) {
+  const srcLines = src.split("\n");
+  const oldLines = old_str.split("\n");
+  const norm = (l) => l.trim();
+  const oldNorm = oldLines.map(norm);
+  const offsets = new Array(srcLines.length);
+  let acc = 0;
+  for (let i = 0; i < srcLines.length; i++) {
+    offsets[i] = acc;
+    acc += srcLines[i].length + 1;
+  }
+  const matches2 = [];
+  const window = oldLines.length;
+  for (let i = 0; i + window <= srcLines.length; i++) {
+    let ok = true;
+    for (let j = 0; j < window; j++) {
+      if (norm(srcLines[i + j]) !== oldNorm[j]) {
+        ok = false;
+        break;
+      }
+    }
+    if (!ok) continue;
+    const start = offsets[i];
+    const last = i + window - 1;
+    const end = offsets[last] + srcLines[last].length;
+    matches2.push([start, end]);
+  }
+  return matches2.length === 1 ? matches2[0] : null;
+}
 function nearMiss(src, old_str) {
   const srcLines = src.split("\n");
   const needle = old_str.split("\n").find((l) => l.trim()) ?? old_str;
@@ -208,6 +270,7 @@ var init_edit_file = __esm({
   "src/tools/edit_file.ts"() {
     "use strict";
     init_paths();
+    init_verifyHint();
     edit_file = {
       name: "edit_file",
       description: "Replace an exact string in a file. old_str must be unique unless replace_all is set. On no match, returns the closest text in the file.",
@@ -230,6 +293,15 @@ var init_edit_file = __esm({
           const src = readFileSync3(abs, "utf-8");
           const first = src.indexOf(old_str);
           if (first === -1) {
+            if (replace_all !== true) {
+              const fuzzy = fuzzyRange(src, old_str);
+              if (fuzzy) {
+                const [s, e] = fuzzy;
+                const out2 = src.slice(0, s) + new_str + src.slice(e);
+                writeFileSync3(abs, out2, "utf-8");
+                return { content: `Edited ${path} (whitespace-tolerant match).${verifyHint(path)}` };
+              }
+            }
             return { content: `old_str not found in ${path}.${nearMiss(src, old_str)}`, is_error: true };
           }
           const all = replace_all === true;
@@ -242,7 +314,7 @@ var init_edit_file = __esm({
           const out = all ? src.split(old_str).join(new_str) : src.slice(0, first) + new_str + src.slice(first + old_str.length);
           const n = all ? src.split(old_str).length - 1 : 1;
           writeFileSync3(abs, out, "utf-8");
-          return { content: `Edited ${path}${all ? ` (${n} occurrences)` : ""}` };
+          return { content: `Edited ${path}${all ? ` (${n} occurrences)` : ""}.${verifyHint(path)}` };
         } catch (err) {
           return { content: err instanceof Error ? err.message : String(err), is_error: true };
         }
@@ -315,6 +387,7 @@ var init_write_file = __esm({
   "src/tools/write_file.ts"() {
     "use strict";
     init_paths();
+    init_verifyHint();
     write_file = {
       name: "write_file",
       description: "Create or overwrite a file with the given content. Parent dirs auto-created.",
@@ -331,7 +404,7 @@ var init_write_file = __esm({
           const abs = confinePath(path);
           mkdirSync3(dirname(abs), { recursive: true });
           writeFileSync4(abs, content, "utf-8");
-          return { content: `Wrote ${path} (${content.length} bytes)` };
+          return { content: `Wrote ${path} (${content.length} bytes).${verifyHint(path)}` };
         } catch (err) {
           return { content: err instanceof Error ? err.message : String(err), is_error: true };
         }
@@ -1348,7 +1421,7 @@ import { createElement } from "react";
 
 // src/ui/App.tsx
 init_client();
-import { useState as useState4, useEffect as useEffect3 } from "react";
+import { useState as useState5, useEffect as useEffect4 } from "react";
 import { Box as Box10, Text as Text10, useApp } from "ink";
 import { homedir as homedir6 } from "os";
 import { sep as sep2 } from "path";
@@ -1790,6 +1863,7 @@ function FilePicker({ matches: matches2, cursor }) {
 }
 
 // src/ui/ChatView.tsx
+import { useState as useState3, useEffect as useEffect3 } from "react";
 import { Box as Box9, Text as Text9 } from "ink";
 
 // src/ui/ThinkingBlock.tsx
@@ -1850,6 +1924,24 @@ var EMPTY_STATE_TITLE = "Ask anything, or try:";
 
 // src/ui/ChatView.tsx
 import { jsx as jsx9, jsxs as jsxs9 } from "react/jsx-runtime";
+var COLLAPSED_LINES = 3;
+var globalToolExpanded = false;
+var toolExpandListeners = /* @__PURE__ */ new Set();
+function toggleToolExpanded() {
+  globalToolExpanded = !globalToolExpanded;
+  toolExpandListeners.forEach((fn) => fn());
+}
+function useToolExpanded() {
+  const [expanded, setExpanded] = useState3(globalToolExpanded);
+  useEffect3(() => {
+    const handler = () => setExpanded(globalToolExpanded);
+    toolExpandListeners.add(handler);
+    return () => {
+      toolExpandListeners.delete(handler);
+    };
+  }, []);
+  return expanded;
+}
 function formatTokens(n) {
   if (n >= 1e3) return (n / 1e3).toFixed(n >= 1e4 ? 0 : 1) + "k";
   return String(n);
@@ -1872,8 +1964,8 @@ function FileEditBlock({
   removed,
   previewLines
 }) {
-  const MAX = 16;
-  const shown = previewLines.slice(0, MAX);
+  const expanded = useToolExpanded();
+  const shown = expanded ? previewLines : previewLines.slice(0, COLLAPSED_LINES);
   const extra = previewLines.length - shown.length;
   return /* @__PURE__ */ jsxs9(Box9, { flexDirection: "column", marginLeft: 2, children: [
     /* @__PURE__ */ jsxs9(Box9, { children: [
@@ -1890,15 +1982,24 @@ function FileEditBlock({
       "\u23BF  ",
       removed > 0 ? `Added ${added} lines, removed ${removed} lines` : `Added ${added} lines`
     ] }) }),
-    shown.map((ln, i) => /* @__PURE__ */ jsx9(Box9, { marginLeft: 4, children: /* @__PURE__ */ jsxs9(Text9, { color: ln.sign === "+" ? "green" : ln.sign === "-" ? "red" : void 0, dimColor: ln.sign === " ", children: [
-      ln.sign,
-      " ",
-      ln.text
-    ] }) }, i)),
+    shown.map((ln, i) => {
+      const width = (process.stdout.columns ?? 80) - 6 - 20;
+      const raw = `${ln.sign} ${ln.text}`;
+      const content = raw.length > width ? raw.slice(0, width) : raw.padEnd(width);
+      return /* @__PURE__ */ jsx9(Box9, { marginLeft: 4, children: /* @__PURE__ */ jsx9(
+        Text9,
+        {
+          wrap: "truncate",
+          backgroundColor: ln.sign === "+" ? "#13351f" : ln.sign === "-" ? "#3b1414" : void 0,
+          dimColor: ln.sign === " ",
+          children: content
+        }
+      ) }, i);
+    }),
     extra > 0 && /* @__PURE__ */ jsx9(Box9, { marginLeft: 4, children: /* @__PURE__ */ jsxs9(Text9, { dimColor: true, children: [
       "\u2026 ",
       extra,
-      " more lines"
+      " more lines \xB7 ctrl+o to expand"
     ] }) })
   ] });
 }
@@ -1964,6 +2065,7 @@ function summarizeResult(res, toolName) {
   return extra > 0 ? `${head} (+${extra} lines)` : head;
 }
 function ToolResultBlock({ result, toolName }) {
+  const expanded = useToolExpanded();
   const content = result.content ?? "";
   const lines = content.split("\n");
   const showMulti = (toolName === "run_bash" || toolName === "grep" || toolName === "glob" || result.is_error) && lines.length > 1;
@@ -1973,9 +2075,9 @@ function ToolResultBlock({ result, toolName }) {
       summarizeResult(result, toolName)
     ] }) });
   }
-  const MAX_LINES = 10;
   const MAX_LINE_WIDTH = 200;
-  const shown = lines.slice(0, MAX_LINES).map((l) => truncate(l, MAX_LINE_WIDTH));
+  const visible = expanded ? lines : lines.slice(0, COLLAPSED_LINES);
+  const shown = visible.map((l) => truncate(l, MAX_LINE_WIDTH));
   const extra = lines.length - shown.length;
   return /* @__PURE__ */ jsxs9(Box9, { flexDirection: "column", marginLeft: 2, children: [
     /* @__PURE__ */ jsxs9(Text9, { color: result.is_error ? "red" : void 0, dimColor: !result.is_error, children: [
@@ -1986,7 +2088,7 @@ function ToolResultBlock({ result, toolName }) {
     extra > 0 && /* @__PURE__ */ jsx9(Box9, { marginLeft: 4, children: /* @__PURE__ */ jsxs9(Text9, { dimColor: true, children: [
       "\u2026 ",
       extra,
-      " more lines"
+      " more lines \xB7 ctrl+o to expand"
     ] }) })
   ] });
 }
@@ -2127,22 +2229,22 @@ function ChatView({
 
 // src/ui/hooks/useAgentRunner.ts
 init_loop();
-import { useState as useState3, useRef } from "react";
+import { useState as useState4, useRef } from "react";
 var FLUSH_MS = 100;
 function useAgentRunner(model, activeCtx) {
-  const [messages, setMessages] = useState3([]);
-  const [thinking, setThinking] = useState3(false);
-  const [thinkingContent, setThinkingContent] = useState3("");
-  const [streaming, setStreaming] = useState3(false);
-  const [streamingContent, setStreamingContent] = useState3("");
-  const [error, setError] = useState3(null);
-  const [busy, setBusy] = useState3(false);
-  const [processingLabel, setProcessingLabel] = useState3(void 0);
-  const [agentHistory, setAgentHistory] = useState3([]);
-  const [pendingPermission, setPendingPermission] = useState3(null);
-  const [permissionCursor, setPermissionCursor] = useState3(0);
-  const [activeToolUses, setActiveToolUses] = useState3([]);
-  const [activeToolResults, setActiveToolResults] = useState3([]);
+  const [messages, setMessages] = useState4([]);
+  const [thinking, setThinking] = useState4(false);
+  const [thinkingContent, setThinkingContent] = useState4("");
+  const [streaming, setStreaming] = useState4(false);
+  const [streamingContent, setStreamingContent] = useState4("");
+  const [error, setError] = useState4(null);
+  const [busy, setBusy] = useState4(false);
+  const [processingLabel, setProcessingLabel] = useState4(void 0);
+  const [agentHistory, setAgentHistory] = useState4([]);
+  const [pendingPermission, setPendingPermission] = useState4(null);
+  const [permissionCursor, setPermissionCursor] = useState4(0);
+  const [activeToolUses, setActiveToolUses] = useState4([]);
+  const [activeToolResults, setActiveToolResults] = useState4([]);
   const busyRef = useRef(false);
   const abortRef = useRef(null);
   const pendingPermissionRef = useRef(null);
@@ -2413,6 +2515,10 @@ function useKeyboard(opts) {
       toggleThinkingVisible();
       return;
     }
+    if (key.ctrl && char === "o") {
+      toggleToolExpanded();
+      return;
+    }
     if (key.escape && busyRef.current && abortRef.current) {
       abortRef.current.abort();
       return;
@@ -2638,30 +2744,30 @@ import { Fragment as Fragment2, jsx as jsx10, jsxs as jsxs10 } from "react/jsx-r
 function App() {
   const { exit } = useApp();
   const cwd = process.cwd().replace(homedir6(), "~").split(sep2).join("/");
-  const [cfg, setCfg] = useState4(loadConfig());
-  const [models, setModels] = useState4([]);
-  const [contexts, setContexts] = useState4({});
-  const [activeCtx, setActiveCtx] = useState4(null);
-  const [state, setState] = useState4("loading");
-  const [cursor, setCursor] = useState4(0);
-  const [updateAvailable, setUpdateAvailable] = useState4(null);
-  const [ollamaDown, setOllamaDown] = useState4(false);
-  const [sessionId, setSessionId] = useState4(() => newSessionId());
-  const [sessions, setSessions] = useState4([]);
-  const [notice, setNotice] = useState4(null);
-  const [input, setInput] = useState4("");
-  const [paletteCursor, setPaletteCursor] = useState4(0);
-  const [filePickerCursor, setFilePickerCursor] = useState4(0);
+  const [cfg, setCfg] = useState5(loadConfig());
+  const [models, setModels] = useState5([]);
+  const [contexts, setContexts] = useState5({});
+  const [activeCtx, setActiveCtx] = useState5(null);
+  const [state, setState] = useState5("loading");
+  const [cursor, setCursor] = useState5(0);
+  const [updateAvailable, setUpdateAvailable] = useState5(null);
+  const [ollamaDown, setOllamaDown] = useState5(false);
+  const [sessionId, setSessionId] = useState5(() => newSessionId());
+  const [sessions, setSessions] = useState5([]);
+  const [notice, setNotice] = useState5(null);
+  const [input, setInput] = useState5("");
+  const [paletteCursor, setPaletteCursor] = useState5(0);
+  const [filePickerCursor, setFilePickerCursor] = useState5(0);
   const agent = useAgentRunner(cfg.model, activeCtx);
-  useEffect3(() => {
+  useEffect4(() => {
     checkForUpdate().then((v) => {
       if (v) setUpdateAvailable(v);
     });
   }, []);
-  useEffect3(() => {
+  useEffect4(() => {
     if (agent.agentHistory.length) persistSession(sessionId, agent.agentHistory);
   }, [agent.agentHistory, sessionId]);
-  useEffect3(() => {
+  useEffect4(() => {
     listModels().then((m) => {
       setModels(m);
       setState(cfg.model ? "ready" : "select-model");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "miii-agent",
-  "version": "0.1.11",
+  "version": "0.1.13",
   "description": "Terminal AI coding agent powered by Ollama",
   "type": "module",
   "bin": {