miii-cli 1.1.2 → 1.1.3

package/dist/llm/stream.js

@@ -1,3 +1,61 @@
+// Transient errors worth retrying: rate limits + server-side faults
+const RETRYABLE_STATUS = new Set([429, 500, 502, 503, 529]);
+const MAX_RETRIES = 4;
+const MAX_DELAY_MS = 30_000;
+function retryDelay(attempt) {
+    // Exponential backoff: 1s → 2s → 4s → 8s, capped at 30s, ±20% jitter
+    const base = 1_000 * Math.pow(2, attempt);
+    const capped = Math.min(base, MAX_DELAY_MS);
+    return Math.round(capped * (0.8 + Math.random() * 0.4));
+}
+function sleep(ms, signal) {
+    return new Promise((resolve, reject) => {
+        if (signal?.aborted) {
+            reject(new DOMException('Aborted', 'AbortError'));
+            return;
+        }
+        const t = setTimeout(resolve, ms);
+        signal?.addEventListener('abort', () => { clearTimeout(t); reject(new DOMException('Aborted', 'AbortError')); }, { once: true });
+    });
+}
+async function fetchWithRetry(url, init, signal, onRetry) {
+    for (let attempt = 0; attempt <= MAX_RETRIES; attempt++) {
+        let res;
+        try {
+            res = await fetch(url, { ...init, signal });
+        }
+        catch (err) {
+            if (err?.name === 'AbortError')
+                throw err;
+            if (attempt === MAX_RETRIES)
+                throw err;
+            const delayMs = retryDelay(attempt);
+            onRetry?.(attempt + 1, MAX_RETRIES, delayMs);
+            await sleep(delayMs, signal);
+            continue;
+        }
+        if (res.ok || !RETRYABLE_STATUS.has(res.status) || attempt === MAX_RETRIES)
+            return res;
+        const retryAfterSec = Number(res.headers.get('retry-after') ?? 0);
+        const delayMs = retryAfterSec > 0 ? retryAfterSec * 1000 : retryDelay(attempt);
+        onRetry?.(attempt + 1, MAX_RETRIES, delayMs);
+        await sleep(delayMs, signal);
+    }
+    throw new Error('fetchWithRetry: exhausted retries without returning');
+}
+export async function warmup(provider, baseUrl, model) {
+    if (provider !== 'ollama')
+        return;
+    try {
+        await fetch(`${baseUrl}/api/generate`, {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({ model, keep_alive: '10m' }),
+            signal: AbortSignal.timeout(30_000),
+        });
+    }
+    catch { }
+}
 export async function chat(cfg) {
     if (cfg.provider === 'anthropic')
         return chatAnthropic(cfg);
@@ -6,14 +64,13 @@ export async function chat(cfg) {
     return chatOllama(cfg);
 }
 async function chatOllama(cfg) {
-    const { model, messages, baseUrl, signal, onDone, onError, onUsage, onChunk } = cfg;
+    const { model, messages, baseUrl, signal, onDone, onError, onUsage, onChunk, onRetry } = cfg;
     try {
-        const res = await fetch(`${baseUrl}/api/chat`, {
+        const res = await fetchWithRetry(`${baseUrl}/api/chat`, {
             method: 'POST',
             headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify({ model, messages, stream: !!onChunk }),
-            signal,
-        });
+        }, signal, onRetry);
         if (!res.ok) {
             onError(new Error(`Ollama ${res.status}: ${await res.text()}`));
             return;
@@ -64,14 +121,13 @@ async function chatOllama(cfg) {
     }
 }
 async function chatOpenAI(cfg) {
-    const { model, messages, baseUrl, apiKey, signal, onDone, onError, onUsage, onChunk } = cfg;
+    const { model, messages, baseUrl, apiKey, signal, onDone, onError, onUsage, onChunk, onRetry } = cfg;
     try {
-        const res = await fetch(`${baseUrl}/v1/chat/completions`, {
+        const res = await fetchWithRetry(`${baseUrl}/v1/chat/completions`, {
             method: 'POST',
             headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${apiKey ?? 'local'}` },
             body: JSON.stringify({ model, messages, stream: !!onChunk }),
-            signal,
-        });
+        }, signal, onRetry);
         if (!res.ok) {
             onError(new Error(`LLM ${res.status}: ${await res.text()}`));
             return;
@@ -118,7 +174,7 @@ async function chatOpenAI(cfg) {
     }
 }
 async function chatAnthropic(cfg) {
-    const { model, messages, baseUrl, apiKey, signal, onDone, onError, onUsage } = cfg;
+    const { model, messages, baseUrl, apiKey, signal, onDone, onError, onUsage, onRetry } = cfg;
     const url = baseUrl && baseUrl !== 'http://localhost:11434'
         ? `${baseUrl}/v1/messages`
         : 'https://api.anthropic.com/v1/messages';
@@ -132,7 +188,7 @@ async function chatAnthropic(cfg) {
         };
         if (systemParts.length)
             body.system = systemParts.join('\n\n');
-        const res = await fetch(url, {
+        const res = await fetchWithRetry(url, {
             method: 'POST',
             headers: {
                 'content-type': 'application/json',
@@ -140,8 +196,7 @@ async function chatAnthropic(cfg) {
                 'anthropic-version': '2023-06-01',
             },
             body: JSON.stringify(body),
-            signal,
-        });
+        }, signal, onRetry);
         if (!res.ok) {
             onError(new Error(`Anthropic ${res.status}: ${await res.text()}`));
             return;

package/dist/memory/store.js

@@ -1,14 +1,12 @@
 import { readFileSync, writeFileSync, mkdirSync, existsSync } from 'fs';
 import { join } from 'path';
-import { homedir } from 'os';
-const MEMORY_DIR = join(homedir(), '.config', 'miii', 'memory');
 const MAX_FACTS = 200;
-function ensureDir() {
-    mkdirSync(MEMORY_DIR, { recursive: true });
+function memoryPath(projectDir) {
+    return join(projectDir, 'memory.json');
 }
-export function loadLongMemory(sessionName) {
-    ensureDir();
-    const p = join(MEMORY_DIR, `${sessionName}.json`);
+export function loadLongMemory(projectDir) {
+    mkdirSync(projectDir, { recursive: true });
+    const p = memoryPath(projectDir);
     if (!existsSync(p))
         return [];
     try {
@@ -19,9 +17,9 @@ export function loadLongMemory(sessionName) {
         return [];
     }
 }
-export function saveLongMemory(sessionName, facts) {
-    ensureDir();
-    writeFileSync(join(MEMORY_DIR, `${sessionName}.json`), JSON.stringify(facts));
+export function saveLongMemory(projectDir, facts) {
+    mkdirSync(projectDir, { recursive: true });
+    writeFileSync(memoryPath(projectDir), JSON.stringify(facts));
 }
 export function mergeFacts(existing, newTexts) {
     const existingSet = new Set(existing.map(f => f.text.toLowerCase()));
@@ -37,5 +35,5 @@ export function mergeFacts(existing, newTexts) {
 export function formatMemoryBlock(facts) {
     if (!facts.length)
         return '';
-    return `\n\n[Long-term memory — recalled from prior conversation]\n${facts.map(f => `- ${f.text}`).join('\n')}`;
+    return `\n\n[Long-term memory — recalled from prior sessions in this project]\n${facts.map(f => `- ${f.text}`).join('\n')}`;
 }

package/dist/sessions.js

@@ -1,23 +1,31 @@
 import { readFileSync, writeFileSync, mkdirSync, chmodSync, readdirSync, statSync, existsSync, unlinkSync } from 'fs';
 import { join } from 'path';
 import { homedir } from 'os';
-const SESSIONS_DIR = join(homedir(), '.config', 'miii', 'sessions');
-function ensureDir() {
-    mkdirSync(SESSIONS_DIR, { recursive: true, mode: 0o700 });
-    chmodSync(SESSIONS_DIR, 0o700);
+const PROJECTS_DIR = join(homedir(), '.config', 'miii', 'projects');
+export function getProjectDir(cwd) {
+    const slug = cwd.replace(/\//g, '-').replace(/^-/, '').replace(/[^a-zA-Z0-9_.-]/g, '-') || 'default';
+    return join(PROJECTS_DIR, slug);
+}
+function sessionsDir(projectDir) {
+    return join(projectDir, 'sessions');
+}
+function ensureProjectDir(projectDir) {
+    mkdirSync(sessionsDir(projectDir), { recursive: true, mode: 0o700 });
+    chmodSync(projectDir, 0o700);
 }
 function sanitizeName(name) {
     if (!/^[\w-]+$/.test(name))
         throw new Error(`invalid session name: ${name}`);
     return name;
 }
-export function listSessions() {
-    ensureDir();
-    return readdirSync(SESSIONS_DIR)
+export function listSessions(projectDir) {
+    ensureProjectDir(projectDir);
+    const dir = sessionsDir(projectDir);
+    return readdirSync(dir)
         .filter(f => f.endsWith('.json'))
         .map(f => {
         const name = f.replace('.json', '');
-        const p = join(SESSIONS_DIR, f);
+        const p = join(dir, f);
         let messageCount = 0;
         let updatedAt = 0;
         try {
@@ -30,9 +38,9 @@ export function listSessions() {
     })
         .sort((a, b) => b.updatedAt - a.updatedAt);
 }
-export function loadSession(name) {
-    ensureDir();
-    const p = join(SESSIONS_DIR, `${sanitizeName(name)}.json`);
+export function loadSession(projectDir, name) {
+    ensureProjectDir(projectDir);
+    const p = join(sessionsDir(projectDir), `${sanitizeName(name)}.json`);
     if (!existsSync(p))
         return [];
     try {
@@ -43,28 +51,29 @@ export function loadSession(name) {
         return [];
     }
 }
-export function saveSession(name, messages) {
-    ensureDir();
+export function saveSession(projectDir, name, messages) {
+    ensureProjectDir(projectDir);
     try {
-        writeFileSync(join(SESSIONS_DIR, `${sanitizeName(name)}.json`), JSON.stringify(messages), { mode: 0o600 });
+        writeFileSync(join(sessionsDir(projectDir), `${sanitizeName(name)}.json`), JSON.stringify(messages), { mode: 0o600 });
     }
     catch { }
 }
-export function deleteSession(name) {
-    const p = join(SESSIONS_DIR, `${sanitizeName(name)}.json`);
+export function deleteSession(projectDir, name) {
+    const p = join(sessionsDir(projectDir), `${sanitizeName(name)}.json`);
     if (existsSync(p))
         unlinkSync(p);
 }
-export function deleteAllSessions(exceptName) {
-    ensureDir();
-    const files = readdirSync(SESSIONS_DIR).filter(f => f.endsWith('.json'));
+export function deleteAllSessions(projectDir, exceptName) {
+    ensureProjectDir(projectDir);
+    const dir = sessionsDir(projectDir);
+    const files = readdirSync(dir).filter(f => f.endsWith('.json'));
     let count = 0;
     for (const f of files) {
         const name = f.replace('.json', '');
         if (exceptName && name === exceptName)
             continue;
         try {
-            unlinkSync(join(SESSIONS_DIR, f));
+            unlinkSync(join(dir, f));
             count++;
         }
         catch { }

package/dist/tasks/compactor.js

@@ -1,8 +1,13 @@
 import { chat } from '../llm/stream.js';
-const COMPACT_THRESHOLD = 18;
+// ~4 chars per token heuristic. 40K chars ≈ 10K tokens — safe floor for 7B local models.
+// Cloud providers can handle more but compacting early keeps responses fast regardless.
+const COMPACT_CHAR_THRESHOLD = 40_000;
 const KEEP_RECENT = 6;
+export function contextSize(messages) {
+    return messages.reduce((sum, m) => sum + m.content.length, 0);
+}
 export function shouldCompact(messages) {
-    return messages.length > COMPACT_THRESHOLD;
+    return contextSize(messages) > COMPACT_CHAR_THRESHOLD;
 }
 const COMPACT_SYSTEM = `You are a context summarizer for an AI coding agent session.
 Your job: produce a dense, structured summary of the conversation so the agent can continue the task without losing context.
@@ -26,7 +31,7 @@ Any constraints, errors encountered, important facts the agent must remember to
 
 Be factual. No padding. Include file paths, error messages, and command outputs verbatim when relevant.`;
 export async function compactContext(messages, cfg, goal) {
-    if (messages.length <= COMPACT_THRESHOLD)
+    if (contextSize(messages) <= COMPACT_CHAR_THRESHOLD)
         return messages;
     const system = messages[0]?.role === 'system' ? messages[0] : null;
     const recent = messages.slice(messages.length - KEEP_RECENT);
@@ -42,6 +47,7 @@ export async function compactContext(messages, cfg, goal) {
         `Conversation to summarize:\n\n${transcript}`,
     ].join('');
     let summary = '';
+    let compactErr = '';
     await chat({
         ...cfg,
         messages: [
@@ -49,8 +55,10 @@ export async function compactContext(messages, cfg, goal) {
             { role: 'user', content: userPrompt },
         ],
         onDone: (text) => { summary = text.trim(); },
-        onError: () => { },
+        onError: (err) => { compactErr = err.message; },
     });
+    if (compactErr)
+        console.error(`[compactor] LLM error: ${compactErr}`);
     // Fallback to dumb compaction if LLM fails
     if (!summary)
         return dumbCompact(messages, goal);
@@ -65,6 +73,8 @@ export async function compactContext(messages, cfg, goal) {
     ];
 }
 function dumbCompact(messages, goal) {
+    if (contextSize(messages) <= COMPACT_CHAR_THRESHOLD)
+        return messages;
     const system = messages[0]?.role === 'system' ? messages[0] : null;
     const userGoal = messages.find(m => m.role === 'user' && !m.content.startsWith('['));
     const recent = messages.slice(messages.length - KEEP_RECENT);

package/dist/tui/InputBar.js

@@ -1,11 +1,11 @@
-import { jsx as _jsx, Fragment as _Fragment, jsxs as _jsxs } from "react/jsx-runtime";
-import { useState, useRef, useMemo, useEffect, useCallback } from 'react';
+import { jsxs as _jsxs, jsx as _jsx, Fragment as _Fragment } from "react/jsx-runtime";
+import { useState, useRef, useMemo, useEffect } from 'react';
 import { Box, Text, useStdout } from 'ink';
 import { InputArea } from './components/InputArea.js';
 import { ModelPicker } from './components/ModelPicker.js';
 import { ConfigPicker } from './components/ConfigPicker.js';
 import { Divider } from './components/StatusBar.js';
-import { tools, getSystemPrompt } from '../tools/index.js';
+import { tools } from '../tools/index.js';
 import { toolArgSummary } from './printer.js';
 import { MacroQueue } from '../tasks/queue.js';
 import { TaskExecutor } from '../tasks/executor.js';
@@ -21,6 +21,7 @@ import { setInkInstance } from './printer.js';
 import { createSearchCodebaseTool } from '../index/tool.js';
 import { saveConfig } from '../config.js';
 import { getTavilyKey, saveTavilyKey } from '../tavily/client.js';
+import { warmup } from '../llm/stream.js';
 function formatElapsed(ms) {
     const s = Math.floor(ms / 1000);
     if (s < 60)
@@ -29,11 +30,27 @@ function formatElapsed(ms) {
     const rem = s % 60;
     return rem === 0 ? `${m}m` : `${m}m ${rem}s`;
 }
+const MAX_DIFF_LINES = 5;
+function DiffPreview({ toolName, args }) {
+    if (toolName === 'patch_file' && (args.old || args.new)) {
+        const oldLines = String(args.old ?? '').split('\n');
+        const newLines = String(args.new ?? '').split('\n');
+        return (_jsxs(Box, { flexDirection: "column", paddingLeft: 2, children: [oldLines.slice(0, MAX_DIFF_LINES).map((line, i) => (_jsxs(Text, { color: "red", dimColor: true, children: ["- ", line.slice(0, 72)] }, `o${i}`))), oldLines.length > MAX_DIFF_LINES && (_jsxs(Text, { color: "gray", dimColor: true, children: ["  \u2026", oldLines.length - MAX_DIFF_LINES, " more"] })), newLines.slice(0, MAX_DIFF_LINES).map((line, i) => (_jsxs(Text, { color: "green", dimColor: true, children: ["+ ", line.slice(0, 72)] }, `n${i}`))), newLines.length > MAX_DIFF_LINES && (_jsxs(Text, { color: "gray", dimColor: true, children: ["  \u2026", newLines.length - MAX_DIFF_LINES, " more"] }))] }));
+    }
+    if ((toolName === 'edit_file' || toolName === 'create_file') && args.content) {
+        const n = String(args.content).split('\n').length;
+        return (_jsx(Box, { paddingLeft: 2, children: _jsxs(Text, { color: "gray", dimColor: true, children: [n, " line", n === 1 ? '' : 's'] }) }));
+    }
+    return null;
+}
 export function InputBar({ config: initialConfig, skills, cwd, session, version, mcpTools = [] }) {
     const [config, setConfig] = useState(initialConfig);
     const { stdout, write: stdoutWrite } = useStdout();
     const cols = stdout.columns ?? 80;
-    useEffect(() => { setInkInstance(stdoutWrite); }, []);
+    useEffect(() => {
+        setInkInstance(stdoutWrite);
+        warmup(initialConfig.provider, initialConfig.baseUrl, initialConfig.model);
+    }, []);
     const phraseSeq = useMemo(() => Array.from({ length: 100 }, () => Math.floor(Math.random() * THINKING_PHRASES.length)), []);
     const [planningMode, setPlanningMode] = useState(false);
     const [configOpen, setConfigOpen] = useState(false);
@@ -42,8 +59,7 @@ export function InputBar({ config: initialConfig, skills, cwd, session, version,
     const executorRef = useRef(new TaskExecutor(tools));
     const lastGitStatusRef = useRef('');
     const abortRef = useRef(null);
-    const { setSessionName, sessionNameRef, historyRef, saveTimerRef, systemPromptRef, pushHistory, buildContext, renameFromMessage, } = useSession(session, cwd, config, mcpTools);
-    const sysPrompt = useCallback((extra = '') => getSystemPrompt(`\n- CWD: ${cwd}${extra}`, mcpTools), [mcpTools, cwd]);
+    const { projectDir, setSessionName, sessionNameRef, historyRef, saveTimerRef, systemPromptRef, pushHistory, buildContext, renameFromMessage, updateMemory, } = useSession(session, cwd, config, mcpTools);
     const { currentModel, setCurrentModel, currentModelRef, pickerOpen, setPickerOpen, pickerModels, pickerLoading, pickerError, pullState, handleModelSelect, handleModelPull, } = useModelPicker(config);
     const deepThinkTool = useMemo(() => ({
         name: 'deep_think',
@@ -64,13 +80,13 @@ export function InputBar({ config: initialConfig, skills, cwd, session, version,
     });
     const { handleGit } = useGit({ pushHistory, buildContext, runLoop });
     const { handleSubmit } = useSubmit({
-        config, skills, cwd, version, currentModelRef, setCurrentModel,
+        config, skills, cwd, projectDir, version, currentModelRef, setCurrentModel,
         historyRef, sessionNameRef, saveTimerRef, systemPromptRef, abortRef,
         setPlanningMode, runLoop, buildContext, pushHistory,
         setSessionName, renameFromMessage,
         setStatus, setTaskLabel, setCurrentTool,
         runRefactor, handleGit, lastGitStatusRef, mcpTools, setConfig,
-        setConfigOpen,
+        setConfigOpen, updateMemory,
     });
     const skillList = skills.list();
     return (_jsxs(Box, { flexDirection: "column", children: [configOpen ? (_jsxs(_Fragment, { children: [_jsx(ConfigPicker, { config: config, currentModel: currentModel, tavilyKey: tavilyKey, onUpdate: ({ model, ...configPatch }) => {
@@ -80,7 +96,7 @@ export function InputBar({ config: initialConfig, skills, cwd, session, version,
                                 setConfig(c => ({ ...c, ...configPatch }));
                                 saveConfig(configPatch);
                             }
-                        }, onTavilyKey: (key) => { saveTavilyKey(key); setTavilyKey(key); }, onClose: () => { setConfigOpen(false); } }), _jsx(Divider, { cols: cols })] })) : pickerOpen ? (_jsxs(_Fragment, { children: [_jsx(ModelPicker, { models: pickerModels, current: currentModel, loading: pickerLoading, error: pickerError, pull: pullState, onSelect: handleModelSelect, onPull: handleModelPull, onClose: () => { setPickerOpen(false); } }), _jsx(Divider, { cols: cols })] })) : compactRequest ? (_jsxs(Box, { paddingX: 1, flexDirection: "column", children: [_jsxs(Box, { gap: 1, children: [_jsx(Text, { color: "yellow", children: "\u26A0" }), _jsx(Text, { color: "white", bold: true, children: "context is large" }), _jsxs(Text, { color: "gray", children: ["(", compactRequest.messageCount, " messages)"] })] }), _jsx(Text, { color: "gray", dimColor: true, children: "compact to keep responses fast, or keep full history" })] })) : permissionRequest ? (_jsxs(Box, { paddingX: 1, gap: 1, children: [_jsx(Text, { color: "yellow", children: "\u26A0" }), _jsx(Text, { color: "white", bold: true, children: permissionRequest.toolName }), _jsx(Text, { color: "gray", children: toolArgSummary(permissionRequest.args) })] })) : (status === 'thinking' || status === 'tool') ? (_jsxs(Box, { flexDirection: "column", paddingX: 1, children: [_jsx(Box, { children: status === 'thinking'
+                        }, onTavilyKey: (key) => { saveTavilyKey(key); setTavilyKey(key); }, onClose: () => { setConfigOpen(false); } }), _jsx(Divider, { cols: cols })] })) : pickerOpen ? (_jsxs(_Fragment, { children: [_jsx(ModelPicker, { models: pickerModels, current: currentModel, loading: pickerLoading, error: pickerError, pull: pullState, onSelect: handleModelSelect, onPull: handleModelPull, onClose: () => { setPickerOpen(false); } }), _jsx(Divider, { cols: cols })] })) : compactRequest ? (_jsxs(Box, { paddingX: 1, flexDirection: "column", children: [_jsxs(Box, { gap: 1, children: [_jsx(Text, { color: "yellow", children: "\u26A0" }), _jsx(Text, { color: "white", bold: true, children: "context is large" }), _jsxs(Text, { color: "gray", children: ["(~", compactRequest.messageCount, "k chars)"] })] }), _jsx(Text, { color: "gray", dimColor: true, children: "compact to keep responses fast, or keep full history" })] })) : permissionRequest ? (_jsxs(Box, { paddingX: 1, flexDirection: "column", children: [_jsxs(Box, { gap: 1, children: [_jsx(Text, { color: "yellow", children: "\u26A0" }), _jsx(Text, { color: "white", bold: true, children: permissionRequest.toolName }), _jsx(Text, { color: "gray", children: toolArgSummary(permissionRequest.args) })] }), _jsx(DiffPreview, { toolName: permissionRequest.toolName, args: permissionRequest.args })] })) : (status === 'thinking' || status === 'tool') ? (_jsxs(Box, { flexDirection: "column", paddingX: 1, children: [_jsx(Box, { children: status === 'thinking'
                             ? _jsxs(_Fragment, { children: [_jsxs(Text, { color: "yellow", children: [SPARKLE[tick % SPARKLE.length], " "] }), _jsx(Text, { color: "gray", dimColor: true, italic: true, children: THINKING_PHRASES[phraseSeq[Math.floor(tick / 62) % phraseSeq.length]] })] })
                             : _jsxs(Text, { color: "yellow", dimColor: true, children: ["\u2699 running ", currentTool ?? 'tool', "\u2026"] }) }), _jsxs(Box, { gap: 2, children: [_jsx(Text, { color: "gray", dimColor: true, children: formatElapsed(Date.now() - thinkingStartRef.current) }), taskLabel && _jsx(Text, { color: "cyan", dimColor: true, children: taskLabel })] })] })) : null, _jsx(InputArea, { status: status, skills: skillList, cwd: cwd, planningMode: planningMode, permissionRequest: permissionRequest, onPermissionResponse: resolvePermission, compactRequest: compactRequest, onCompactResponse: resolveCompact, onSubmit: handleSubmit, onAbort: handleAbort, history: historyRef.current.filter(m => m.role === 'user').map(m => m.content) })] }));
 }

package/dist/tui/git-context.js

@@ -1,5 +1,3 @@
-import { readFile } from '../files/ops.js';
-import { resolve } from 'path';
 import { exec } from 'child_process';
 import { promisify } from 'util';
 const gitRun = promisify(exec);
@@ -14,44 +12,11 @@ export async function buildGitContext(cwd, lastStatusRef) {
         if (!status || status === lastStatusRef.current)
             return { prefix: '', label: '' };
         lastStatusRef.current = status;
-        const MAX_TOTAL = 40_000;
-        const MAX_FILE = 15_000;
-        let total = 0;
-        const parts = [];
-        const skipped = [];
-        for (const line of status.split('\n')) {
-            const code = line.slice(0, 2);
-            if (code.includes('D'))
-                continue;
-            const raw = line.slice(3).trim().replace(/^"|"$/g, '');
-            const arrowIdx = raw.lastIndexOf(' -> ');
-            const rel = arrowIdx !== -1 ? raw.slice(arrowIdx + 4) : raw;
-            if (!rel)
-                continue;
-            try {
-                const content = readFile(resolve(cwd, rel));
-                if (!content || content.length > MAX_FILE) {
-                    skipped.push(rel);
-                    continue;
-                }
-                total += content.length;
-                if (total > MAX_TOTAL) {
-                    skipped.push(rel);
-                    continue;
-                }
-                parts.push(`<file path="${rel}">\n${content}\n</file>`);
-            }
-            catch {
-                skipped.push(rel);
-            }
-        }
-        if (!parts.length && !skipped.length)
-            return { prefix: '', label: '' };
-        let prefix = '[Auto-context: git-changed files]\n' + parts.join('\n') + '\n';
-        if (skipped.length)
-            prefix += `Files changed but too large to auto-load: ${skipped.join(', ')}\n`;
-        prefix += '\n';
-        const label = `auto-loaded ${parts.length} changed file(s)${skipped.length ? `, skipped ${skipped.length} (too large)` : ''}`;
+        const files = status.split('\n')
+            .map(l => l.slice(3).trim().replace(/^"|"$/g, ''))
+            .filter(Boolean);
+        const prefix = `[Git: ${files.length} changed file(s)]\n${status}\n\n`;
+        const label = `git: ${files.length} changed file(s) in context`;
         return { prefix, label };
     }
     catch {

package/dist/tui/hooks/useRunLoop.js

@@ -3,13 +3,18 @@ import { readFileSync, writeFileSync, unlinkSync, existsSync } from 'fs';
 import { chat } from '../../llm/stream.js';
 import { tools as staticTools } from '../../tools/index.js';
 import { StreamParser, extractBareToolCall } from '../../parser/stream-parser.js';
-import { shouldCompact, compactContext } from '../../tasks/compactor.js';
+import { shouldCompact, compactContext, contextSize } from '../../tasks/compactor.js';
 import * as printer from '../printer.js';
-const MAX_TOOL_DEPTH = 6;
+const MAX_TOOL_DEPTH = 10;
 const FILE_EDIT_TOOLS = new Set(['edit_file', 'create_file', 'patch_file', 'delete_file']);
 const SHOW_RESULT_TOOLS = new Set(['run_tests', 'git_commit']);
 const PERMISSION_TOOLS = new Set(['edit_file', 'patch_file', 'delete_file', 'create_file', 'move_file', 'run_command', 'git_commit']);
 const CHECKPOINT_TOOLS = new Set(['edit_file', 'patch_file', 'create_file', 'delete_file']);
+// Tool result messages that are ephemeral — never worth storing in memory or compact summaries
+const EPHEMERAL_PATTERN = /^Tool (read_file|list_files|run_tests) result:|^\[current state of|^\[Context compacted/;
+export function stripEphemeral(messages) {
+    return messages.filter(m => m.role !== 'user' || !EPHEMERAL_PATTERN.test(m.content));
+}
 export function useRunLoop(config, currentModelRef, pushHistory, extraTools = [], abortRef) {
     const [status, setStatus] = useState('idle');
     const [tick, setTick] = useState(0);
@@ -57,11 +62,12 @@ export function useRunLoop(config, currentModelRef, pushHistory, extraTools = []
         if (shouldCompact(contextMsgs)) {
             const approved = await new Promise(resolve => {
                 compactResolveRef.current = resolve;
-                setCompactRequest({ messageCount: contextMsgs.length });
+                setCompactRequest({ messageCount: Math.round(contextSize(contextMsgs) / 1000) });
             });
             if (approved) {
                 printer.systemMsg('compacting context…');
-                msgs = await compactContext(contextMsgs, {
+                const toCompact = stripEphemeral(contextMsgs);
+                msgs = await compactContext(toCompact, {
                     provider: config.provider,
                     model: currentModelRef.current,
                     baseUrl: config.baseUrl,
@@ -80,6 +86,9 @@ export function useRunLoop(config, currentModelRef, pushHistory, extraTools = []
             baseUrl: config.baseUrl,
             messages: msgs,
             signal: abortRef.current.signal,
+            onRetry(attempt, max, delayMs) {
+                printer.systemMsg(`retry ${attempt}/${max} — waiting ${Math.round(delayMs / 1000)}s`);
+            },
             async onDone(fullText) {
                 const pendingTools = [];
                 const textParts = [];

package/dist/tui/hooks/useSession.js

@@ -1,5 +1,5 @@
 import { useState, useRef, useEffect } from 'react';
-import { loadSession, saveSession, deleteSession } from '../../sessions.js';
+import { getProjectDir, loadSession, saveSession, deleteSession } from '../../sessions.js';
 import { getSystemPrompt } from '../../tools/index.js';
 import { getTavilyKey, saveTavilyKey } from '../../tavily/client.js';
 import * as printer from '../printer.js';
@@ -10,6 +10,7 @@ function buildSystemPrompt(cwd, facts, extraTools = []) {
     return getSystemPrompt(`\n- CWD: ${cwd}`, extraTools) + formatMemoryBlock(facts);
 }
 export function useSession(initialSession, cwd, config, extraTools = []) {
+    const projectDir = getProjectDir(cwd);
     const [sessionName, setSessionName] = useState(initialSession);
     const sessionNameRef = useRef(initialSession);
     const historyRef = useRef([]);
@@ -17,43 +18,45 @@ export function useSession(initialSession, cwd, config, extraTools = []) {
     const firstMessageSentRef = useRef(false);
     const longMemoryRef = useRef([]);
     const systemPromptRef = useRef(buildSystemPrompt(cwd, [], extraTools));
+    const extractingRef = useRef(false);
     useEffect(() => { sessionNameRef.current = sessionName; }, [sessionName]);
     useEffect(() => {
-        const facts = loadLongMemory(initialSession);
+        const facts = loadLongMemory(projectDir);
         longMemoryRef.current = facts;
         systemPromptRef.current = buildSystemPrompt(cwd, facts, extraTools);
         if (facts.length)
-            printer.systemMsg(`long memory: ${facts.length} facts loaded`);
-        const history = loadSession(initialSession);
+            printer.systemMsg(`project memory: ${facts.length} facts`);
+        const history = loadSession(projectDir, initialSession);
         historyRef.current = history;
         if (history.length)
             printer.systemMsg(`resumed "${initialSession}" — ${history.length} messages`);
         if (config.tavilyApiKey && !getTavilyKey())
             saveTavilyKey(config.tavilyApiKey);
         if (!getTavilyKey()) {
-            printer.systemMsg('Tavily API key not set — web search disabled. Run /tavily-key <key> to enable. Get a free key at https://tavily.com');
+            printer.systemMsg('Tavily API key not set — web search disabled. Run /config to enable.');
         }
     }, []);
     function scheduleSave() {
         if (saveTimerRef.current)
             clearTimeout(saveTimerRef.current);
         saveTimerRef.current = setTimeout(() => {
-            saveSession(sessionNameRef.current, historyRef.current);
+            saveSession(projectDir, sessionNameRef.current, historyRef.current);
             saveTimerRef.current = null;
         }, 2000);
     }
     function pushHistory(msg) {
         historyRef.current.push(msg);
-        if (historyRef.current.length > SHORT_MEMORY_SIZE) {
+        if (historyRef.current.length > SHORT_MEMORY_SIZE && !extractingRef.current) {
             const dropped = historyRef.current.splice(0, historyRef.current.length - SHORT_MEMORY_SIZE);
+            extractingRef.current = true;
             extractFacts(dropped, config, config.model).then(newFacts => {
-                if (!newFacts.length)
-                    return;
-                const updated = mergeFacts(longMemoryRef.current, newFacts);
-                longMemoryRef.current = updated;
-                systemPromptRef.current = buildSystemPrompt(cwd, updated, extraTools);
-                saveLongMemory(sessionNameRef.current, updated);
-            });
+                if (newFacts.length) {
+                    const updated = mergeFacts(longMemoryRef.current, newFacts);
+                    longMemoryRef.current = updated;
+                    systemPromptRef.current = buildSystemPrompt(cwd, updated, extraTools);
+                    saveLongMemory(projectDir, updated);
+                }
+            }).finally(() => { extractingRef.current = false; });
         }
         scheduleSave();
     }
@@ -74,7 +77,7 @@ export function useSession(initialSession, cwd, config, extraTools = []) {
         sessionNameRef.current = slug;
         setSessionName(slug);
         try {
-            deleteSession(oldName);
+            deleteSession(projectDir, oldName);
         }
         catch { }
     }
@@ -85,9 +88,19 @@ export function useSession(initialSession, cwd, config, extraTools = []) {
             ctx.push(extra);
         return ctx;
     }
+    function updateMemory(newFacts) {
+        if (!newFacts.length)
+            return;
+        const updated = mergeFacts(longMemoryRef.current, newFacts);
+        longMemoryRef.current = updated;
+        systemPromptRef.current = buildSystemPrompt(cwd, updated, extraTools);
+        saveLongMemory(projectDir, updated);
+    }
     return {
+        projectDir,
         sessionName, setSessionName, sessionNameRef,
         historyRef, saveTimerRef, systemPromptRef,
-        pushHistory, buildContext, renameFromMessage,
+        longMemoryRef,
+        pushHistory, buildContext, renameFromMessage, updateMemory,
     };
 }

package/dist/tui/hooks/useSubmit.js

@@ -4,6 +4,8 @@ import { getSystemPrompt } from '../../tools/index.js';
 import { saveConfig } from '../../config.js';
 import { loadSession, saveSession, listSessions, deleteSession, deleteAllSessions } from '../../sessions.js';
 import { compactContext } from '../../tasks/compactor.js';
+import { extractFacts } from '../../memory/extractor.js';
+import { stripEphemeral } from './useRunLoop.js';
 import { runDeepThink } from '../deepThink.js';
 import { buildGitContext, looksCodeRelated } from '../git-context.js';
 import { buildIndex } from '../../index/indexer.js';
@@ -35,7 +37,7 @@ export function useSubmit(deps) {
     const depsRef = useRef(deps);
     depsRef.current = deps;
     const handleSubmit = useCallback(async (text) => {
-        const { config, skills, cwd, version, currentModelRef, setCurrentModel, historyRef, sessionNameRef, saveTimerRef, systemPromptRef, abortRef, setPlanningMode, runLoop, buildContext, pushHistory, setSessionName, renameFromMessage, setStatus, setTaskLabel, setCurrentTool, runRefactor, handleGit, lastGitStatusRef, mcpTools, setConfig, setConfigOpen, } = depsRef.current;
+        const { config, skills, cwd, projectDir, version, currentModelRef, setCurrentModel, historyRef, sessionNameRef, saveTimerRef, systemPromptRef, abortRef, setPlanningMode, runLoop, buildContext, pushHistory, setSessionName, renameFromMessage, setStatus, setTaskLabel, setCurrentTool, runRefactor, handleGit, lastGitStatusRef, mcpTools, setConfig, setConfigOpen, updateMemory, } = depsRef.current;
         const cmd = text.trim();
         if (cmd === '?') {
             printer.systemMsg('shortcuts:\n' +
@@ -164,23 +166,37 @@ export function useSubmit(deps) {
             return;
         }
         if (cmd === '/compact') {
-            const full = buildContext();
-            if (full.length <= 2) {
+            // Strip ephemeral tool noise (read_file, list_files, run_tests results, injected state)
+            const meaningful = stripEphemeral(historyRef.current);
+            if (!meaningful.length) {
                 printer.systemMsg('nothing to compact');
                 return;
             }
-            printer.systemMsg(`compacting ${full.length} messages…`);
+            const before = historyRef.current.length;
+            printer.systemMsg(`compacting ${before} messages (${before - meaningful.length} ephemeral dropped)…`);
             setStatus('thinking');
             try {
-                const compacted = await compactContext(full, {
+                const cfg = {
                     provider: config.provider,
                     model: currentModelRef.current,
                     baseUrl: config.baseUrl,
                     apiKey: config.apiKey,
-                }, undefined);
-                historyRef.current = compacted.filter(m => m.role !== 'system');
-                saveSession(sessionNameRef.current, historyRef.current);
-                printer.systemMsg(`compacted: ${full.length} → ${compacted.length} messages`);
+                };
+                // Run both in parallel: LLM summary + fact extraction
+                const [compacted, facts] = await Promise.all([
+                    compactContext([{ role: 'system', content: '' }, ...meaningful], cfg),
+                    extractFacts(meaningful, config, currentModelRef.current),
+                ]);
+                // Update long-term memory with extracted facts
+                if (facts.length) {
+                    updateMemory(facts);
+                    printer.systemMsg(`memory: +${facts.length} fact${facts.length === 1 ? '' : 's'} saved`);
+                }
+                // Replace history with just the compact summary (no system msg)
+                const summaryOnly = compacted.filter(m => m.role !== 'system');
+                historyRef.current = summaryOnly;
+                saveSession(projectDir, sessionNameRef.current, summaryOnly);
+                printer.systemMsg(`compacted: ${before} → ${summaryOnly.length} message${summaryOnly.length === 1 ? '' : 's'}`);
             }
             catch (e) {
                 printer.errorMsg(`compact failed: ${e}`);
@@ -195,7 +211,7 @@ export function useSubmit(deps) {
                 clearTimeout(saveTimerRef.current);
                 saveTimerRef.current = null;
             }
-            saveSession(sessionNameRef.current, historyRef.current);
+            saveSession(projectDir, sessionNameRef.current, historyRef.current);
             const newName = new Date().toISOString().slice(0, 19).replace(/[:T]/g, '-');
             historyRef.current = [];
             setSessionName(newName);
@@ -206,7 +222,7 @@ export function useSubmit(deps) {
         }
         if (cmd === '/clear') {
             historyRef.current = [];
-            saveSession(sessionNameRef.current, []);
+            saveSession(projectDir, sessionNameRef.current, []);
             setPlanningMode(false);
             systemPromptRef.current = getSystemPrompt(`\n- CWD: ${cwd}`, mcpTools);
             printer.systemMsg('chat cleared');
@@ -295,12 +311,22 @@ export function useSubmit(deps) {
             }
         }
         if (cmd === '/sessions') {
-            const sessions = listSessions();
+            const sessions = listSessions(projectDir);
+            const shortCwd = cwd.replace(process.env.HOME ?? '', '~');
             if (!sessions.length) {
-                printer.systemMsg('no saved sessions');
+                printer.systemMsg(`project: ${shortCwd}\nno saved sessions\n\n` +
+                    `  /new                    start fresh session\n` +
+                    `  /session <name>         switch to session\n` +
+                    `  /session delete <name>  delete session\n` +
+                    `  /session delete all     delete all sessions in this project`);
                 return;
             }
-            printer.systemMsg(sessions.map(s => `${s.name === sessionNameRef.current ? '▶ ' : '  '}${s.name}  (${s.messageCount} msgs)`).join('\n'));
+            const rows = sessions.map(s => `  ${s.name === sessionNameRef.current ? '▶' : ' '} ${s.name.padEnd(32)} ${s.messageCount} msg${s.messageCount === 1 ? '' : 's'}`).join('\n');
+            printer.systemMsg(`project: ${shortCwd}  (${sessions.length} session${sessions.length === 1 ? '' : 's'})\n${rows}\n\n` +
+                `  /session <name>         switch session\n` +
+                `  /session delete <name>  delete session\n` +
+                `  /session delete all     delete all sessions in this project\n` +
+                `  /new                    start fresh session`);
             return;
         }
         if (cmd.startsWith('/session')) {
@@ -316,8 +342,8 @@ export function useSubmit(deps) {
                     return;
                 }
                 if (target === 'all') {
-                    const count = deleteAllSessions(sessionNameRef.current);
-                    printer.systemMsg(`deleted ${count} session(s) — kept active: ${sessionNameRef.current}`);
+                    const count = deleteAllSessions(projectDir, sessionNameRef.current);
+                    printer.systemMsg(`deleted ${count} session${count === 1 ? '' : 's'} — kept active: ${sessionNameRef.current}`);
                     return;
                 }
                 if (target === sessionNameRef.current) {
@@ -325,7 +351,7 @@ export function useSubmit(deps) {
                     return;
                 }
                 try {
-                    deleteSession(target);
+                    deleteSession(projectDir, target);
                     printer.systemMsg(`deleted: ${target}`);
                 }
                 catch (e) {
@@ -337,10 +363,10 @@ export function useSubmit(deps) {
                 clearTimeout(saveTimerRef.current);
                 saveTimerRef.current = null;
             }
-            saveSession(sessionNameRef.current, historyRef.current);
-            historyRef.current = loadSession(arg);
+            saveSession(projectDir, sessionNameRef.current, historyRef.current);
+            historyRef.current = loadSession(projectDir, arg);
             setSessionName(arg);
-            printer.systemMsg(`session → ${arg}  (${historyRef.current.length} messages)`);
+            printer.systemMsg(`session → ${arg}  (${historyRef.current.length} message${historyRef.current.length === 1 ? '' : 's'})`);
             return;
         }
         if (cmd === '/index' || cmd.startsWith('/index ')) {

package/dist/tui/printer.js

@@ -24,9 +24,6 @@ const gray = (s) => col(90, s);
 const yellow = (s) => col(93, s);
 const purple = (s) => col(95, s);
 const red = (s) => col(91, s);
-function indent(text, pad = '  ') {
-    return text.split('\n').map(l => pad + l).join('\n');
-}
 function stripMarkdown(s) {
     return s
         .replace(/\*\*\*(.+?)\*\*\*/g, '$1')
@@ -104,7 +101,6 @@ export function welcome(provider, model, cwd, version, updateAvailable, linked)
     const top = gray('╭') + gray('─') + bold(cyan(` MIII - CLI${versionStr} `)) + gray('─'.repeat(dashCount) + '╮');
     const bottom = gray('╰' + '─'.repeat(innerW) + '╯');
     const shortCwd = cwd.replace(process.env.HOME ?? '', '~');
-    const username = process.env.USER ?? 'there';
     const miniArt = [
         `  ${purple('   ●     ●   ')}`,
         `  ${purple('  ╱ ╲   ╱ ╲  ')}`,
@@ -124,7 +120,7 @@ export function welcome(provider, model, cwd, version, updateAvailable, linked)
         `  ${bold(yellow('Tips for getting started'))}`,
         `  Type ${cyan('@filename')} to inject file into context`,
         `  Use ${cyan('/skill')} to run a skill or command`,
-        `  Use ${cyan('/models')} to switch or pull models`,
+        `  Use ${cyan('/config')} to switch provider, model, or API key`,
         '',
     ];
     const maxLen = Math.max(leftLines.length, rightLines.length);
@@ -266,7 +262,7 @@ export function toolResultSummary(name, args, result) {
     if (summary)
         write(gray(`  ${summary}`) + '\n');
 }
-export function toolMsg(name, result) {
+export function toolMsg(_name, result) {
     const preview = result.length > 600 ? result.slice(0, 600) + '…' : result;
     const body = preview.trim()
         ? preview.split('\n').map(l => gray('    ' + l)).join('\n')

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "miii-cli",
-  "version": "1.1.2",
+  "version": "1.1.3",
   "type": "module",
   "description": "The high-performance local AI coding agent for your terminal. Automate complex workflows with local LLMs.",
   "license": "MIT",