All your AI agent sessions in one place — browse, search, resume. Rust + GPUI.
Rust
Add README screenshots and demo-data generator
Corey Chiu6adcaa21a2bb
diff --git a/README.md b/README.mdindex 5b11d75..89f9703 100644--- a/README.md+++ b/README.md@@ -4,6 +4,8 @@ A native macOS app that gathers every coding-agent session on your machine intoYour agent history is scattered across `~/.claude`, `~/.codex`, and five other private directories. Wake reads them all, read-only, and gives you one fast window into it. Everything stays local: no network requests, ever.++## Features- **Unified browsing** — all sessions grouped by agent / project, live file watching for incremental updates@@ -12,6 +14,8 @@ Your agent history is scattered across `~/.claude`, `~/.codex`, and five other p- **One-click resume** — reopens the session in Terminal/iTerm at the original project directory (`claude --resume`, `codex resume`, …)- **Manage** — star/pin (stored in Wake's own DB, original files untouched), export to Markdown, delete (system Trash + tombstone so deleted sessions stay deleted)++## Supported agents| Agent | Data source |diff --git a/imgs/screenshot-1.webp b/imgs/screenshot-1.webpnew file mode 100644index 0000000..01a6977Binary files /dev/null and b/imgs/screenshot-1.webp differdiff --git a/imgs/screenshot-2.webp b/imgs/screenshot-2.webpnew file mode 100644index 0000000..7d8c1aeBinary files /dev/null and b/imgs/screenshot-2.webp differdiff --git a/scripts/demo-home.py b/scripts/demo-home.pynew file mode 100644index 0000000..5ee8af6--- /dev/null+++ b/scripts/demo-home.py@@ -0,0 +1,277 @@+#!/usr/bin/env python3+"""生成 Wake 截图/演示用的假家目录——纯合成数据,不含任何真实会话。++用法:+ python3 scripts/demo-home.py # 生成到 /tmp/wake-demo-home+ HOME=/tmp/wake-demo-home dist/Wake.app/Contents/MacOS/Wake++app 的索引库与七家 agent 扫描目录全部落在假 HOME 内,真实数据不显示也不被读。+"""+import json+import os+import shutil+import sqlite3+import sys+import time+import uuid+from datetime import datetime, timedelta, timezone++HOME = sys.argv[1] if len(sys.argv) > 1 else "/tmp/wake-demo-home"+NOW = datetime.now(timezone.utc)+REPO = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))+FIXTURES = os.path.join(REPO, "crates", "wake-core", "tests", "fixtures")+++def iso(minutes_ago, offset_s=0):+ t = NOW - timedelta(minutes=minutes_ago) + timedelta(seconds=offset_s)+ return t.strftime("%Y-%m-%dT%H:%M:%S.000Z")+++def sid():+ return str(uuid.uuid4())+++# ---------------------------------------------------------------- Claude++def claude_session(proj, title_msg, turns, age_min):+ """turns: list of ("user"|"assistant", text) 或 ("tool", name, input, result)"""+ s = sid()+ enc = proj.replace("/", "-")+ d = os.path.join(HOME, ".claude", "projects", enc)+ os.makedirs(d, exist_ok=True)+ lines = []+ prev = None+ step = 0++ def push(obj):+ nonlocal prev, step+ u = f"u-{step}"+ obj.update({"parentUuid": prev, "isSidechain": False, "cwd": proj,+ "gitBranch": "main", "sessionId": s, "uuid": u,+ "timestamp": iso(age_min, step * 20)})+ lines.append(json.dumps(obj, ensure_ascii=False))+ prev = u+ step += 1++ push({"type": "user", "message": {"role": "user", "content": title_msg}})+ msg_n = 0+ for t in turns:+ if t[0] == "user":+ push({"type": "user", "message": {"role": "user", "content": t[1]}})+ elif t[0] == "assistant":+ msg_n += 1+ push({"type": "assistant", "message": {+ "id": f"msg_{s[:4]}_{msg_n}", "type": "message", "role": "assistant",+ "model": "claude-fable-5",+ "content": [{"type": "text", "text": t[1]}],+ "usage": {"input_tokens": 900 + step * 37, "output_tokens": 240 + step * 11}}})+ elif t[0] == "thinking":+ msg_n += 1+ push({"type": "assistant", "message": {+ "id": f"msg_{s[:4]}_{msg_n}", "type": "message", "role": "assistant",+ "model": "claude-fable-5",+ "content": [{"type": "thinking", "thinking": t[1]},+ {"type": "text", "text": t[2]}],+ "usage": {"input_tokens": 1200, "output_tokens": 300}}})+ elif t[0] == "tool":+ _, name, tin, tout = t+ msg_n += 1+ tid = f"toolu_{s[:4]}_{msg_n}"+ push({"type": "assistant", "message": {+ "id": f"msg_{s[:4]}_{msg_n}", "type": "message", "role": "assistant",+ "model": "claude-fable-5",+ "content": [{"type": "tool_use", "id": tid, "name": name, "input": tin}]}})+ push({"type": "user", "message": {"role": "user", "content": [+ {"type": "tool_result", "tool_use_id": tid,+ "content": [{"type": "text", "text": tout}]}]}})+ path = os.path.join(d, f"{s}.jsonl")+ with open(path, "w") as f:+ f.write("\n".join(lines) + "\n")+++# ---------------------------------------------------------------- Codex++def codex_session(proj, title_msg, reply, age_min, tool=None):+ s = sid()+ t0 = NOW - timedelta(minutes=age_min)+ d = os.path.join(HOME, ".codex", "sessions", t0.strftime("%Y/%m/%d"))+ os.makedirs(d, exist_ok=True)+ stamp = t0.strftime("%Y-%m-%dT%H-%M-%S")+ lines = [+ {"timestamp": iso(age_min), "type": "session_meta", "payload": {+ "id": s, "timestamp": iso(age_min), "cwd": proj,+ "originator": "codex_cli_rs", "cli_version": "0.21.0", "source": "cli",+ "git": {"branch": "main"}}},+ {"timestamp": iso(age_min, 1), "type": "turn_context", "payload": {+ "cwd": proj, "model": "gpt-5.2-codex", "summary": "auto"}},+ {"timestamp": iso(age_min, 2), "type": "response_item", "payload": {+ "type": "message", "role": "user",+ "content": [{"type": "input_text", "text": title_msg}]}},+ ]+ if tool:+ cmd, out = tool+ lines += [+ {"timestamp": iso(age_min, 4), "type": "response_item", "payload": {+ "type": "function_call", "name": "shell",+ "arguments": json.dumps({"command": cmd}), "call_id": "call_1"}},+ {"timestamp": iso(age_min, 6), "type": "response_item", "payload": {+ "type": "function_call_output", "call_id": "call_1", "output": out}},+ ]+ lines.append({"timestamp": iso(age_min, 9), "type": "response_item", "payload": {+ "type": "message", "role": "assistant",+ "content": [{"type": "output_text", "text": reply}]}})+ with open(os.path.join(d, f"rollout-{stamp}-{s}.jsonl"), "w") as f:+ f.write("\n".join(json.dumps(l, ensure_ascii=False) for l in lines) + "\n")+++# ---------------------------------------------------------------- 其余五家++def copy_fixture_tree(rel_src, rel_dst, replacements):+ src = os.path.join(FIXTURES, rel_src)+ dst = os.path.join(HOME, rel_dst)+ os.makedirs(os.path.dirname(dst), exist_ok=True)+ with open(src) as f:+ text = f.read()+ for a, b in replacements:+ text = text.replace(a, b)+ with open(dst, "w") as f:+ f.write(text)+++def build_sqlite_agents():+ cop_dir = os.path.join(HOME, ".copilot")+ os.makedirs(cop_dir, exist_ok=True)+ c = sqlite3.connect(os.path.join(cop_dir, "session-store.db"))+ c.executescript("""+ CREATE TABLE sessions (id TEXT PRIMARY KEY, cwd TEXT, branch TEXT, summary TEXT,+ created_at TEXT, updated_at TEXT);+ CREATE TABLE turns (id INTEGER PRIMARY KEY, session_id TEXT, turn_index INTEGER,+ user_message TEXT, assistant_response TEXT, timestamp TEXT);+ """)+ ts = (NOW - timedelta(days=4)).strftime("%Y-%m-%d %H:%M:%S")+ c.execute("INSERT INTO sessions VALUES ('cop-demo-1','/Users/demo/Github/oss-metrics','main','Review PR #42 error handling',?,?)", (ts, ts))+ c.execute("INSERT INTO turns VALUES (1,'cop-demo-1',0,'Review the error handling in PR #42','The retry loop swallows the root cause; surfaced it in the log line.',?)", (ts,))+ c.commit(); c.close()++ oc_dir = os.path.join(HOME, ".local", "share", "opencode")+ os.makedirs(oc_dir, exist_ok=True)+ c = sqlite3.connect(os.path.join(oc_dir, "opencode.db"))+ c.executescript("""+ CREATE TABLE session (id TEXT PRIMARY KEY, parent_id TEXT, directory TEXT, title TEXT,+ time_created INTEGER, time_updated INTEGER, model TEXT,+ tokens_input INTEGER, tokens_output INTEGER, tokens_reasoning INTEGER, time_archived INTEGER);+ CREATE TABLE message (id TEXT PRIMARY KEY, session_id TEXT, data TEXT, time_created INTEGER);+ CREATE TABLE part (id TEXT PRIMARY KEY, session_id TEXT, message_id TEXT, data TEXT);+ """)+ t = int((NOW - timedelta(days=6)).timestamp() * 1000)+ c.execute("INSERT INTO session VALUES ('oc-demo-1',NULL,'/Users/demo/Github/blog-engine','Trace the N+1 query in the feed',?,?,'{\"providerID\":\"anthropic\",\"id\":\"claude-sonnet-5\"}',800,400,120,NULL)", (t, t + 60000))+ c.execute("INSERT INTO message VALUES ('m1','oc-demo-1','{\"id\":\"m1\",\"role\":\"user\",\"time\":{\"created\":%d}}',?)" % (t + 1000), (t + 1000,))+ c.execute("""INSERT INTO part VALUES ('p1','oc-demo-1','m1','{"type":"text","text":"Trace the N+1 query in the feed endpoint"}')""")+ c.commit(); c.close()+++# ---------------------------------------------------------------- 数据集++def main():+ if os.path.exists(HOME):+ shutil.rmtree(HOME)+ os.makedirs(HOME)++ g = "/Users/demo/Github"++ # hero 会话(最新,详情页展示用):多轮 + thinking + 工具簇 + markdown+ claude_session(f"{g}/acme-web", "Add full-text search to the sessions list — needs to handle CJK and code substrings like useEffect(", [+ ("thinking", "FTS5 的默认 tokenizer 对 CJK 无能为力,trigram 能同时覆盖中文和代码子串。先看现有 schema 再决定迁移方案。",+ "Good constraint — plain FTS5 tokenizers won't handle CJK or `useEffect(` substrings. I'll use a **trigram tokenizer**. Let me check the current schema first."),+ ("tool", "Read", {"file_path": f"{g}/acme-web/src/db.rs"},+ "CREATE TABLE messages (id INTEGER PRIMARY KEY, session_key TEXT, seq INTEGER, text TEXT);"),+ ("tool", "Grep", {"pattern": "messages_fts", "path": "src/"},+ "no matches"),+ ("assistant", "No FTS table yet. Here's the migration:\n\n```sql\nCREATE VIRTUAL TABLE messages_fts USING fts5(\n text,\n content = 'messages',\n content_rowid = 'id',\n tokenize = 'trigram'\n);\n```\n\nTrigram indexes every 3-char window, so `二维码` and `useEffect(` both match without language-specific tokenizers. Queries under 3 chars fall back to `LIKE` — I'll flag those as degraded in the UI."),+ ("user", "中文短词比如两个字的怎么办?"),+ ("assistant", "两个字(两码点)低于 trigram 下限,走 `LIKE '%..%'` 降级路径——在 546 万行的表上实测 ~40ms,可接受。UI 上会提示「短查询已降级,长关键词更快」。我把这条也写进测试。"),+ ("tool", "Edit", {"file_path": f"{g}/acme-web/src/db.rs", "old_string": "// TODO fts", "new_string": "CREATE VIRTUAL TABLE messages_fts USING fts5(...)"},+ "ok"),+ ("assistant", "Done — migration applied, 16 tests green, search returns in under a millisecond on 280k rows."),+ ], 25)++ claude_session(f"{g}/acme-web", "Fix flaky WebSocket reconnect after wake from sleep", [+ ("tool", "Grep", {"pattern": "reconnect", "path": "src/net/"}, "src/net/ws.ts:88: setTimeout(reconnect, backoff)"),+ ("assistant", "The backoff timer survives sleep, so on wake it fires with a stale socket. Switching to an event-driven reconnect on `online` + visibility change."),+ ], 180)+ claude_session(f"{g}/acme-web", "Dark mode: design tokens for the settings page", [+ ("assistant", "Mapped all 14 hardcoded grays to semantic tokens; both palettes now pass 4.5:1 contrast."),+ ], 60 * 26)+ claude_session(f"{g}/rusty-search", "Explain this borrow checker error in indexer.rs", [+ ("assistant", "You're holding an immutable borrow of `self.index` across the `&mut self` call in the loop — clone the key list first, or restructure with `split_at_mut`."),+ ], 60 * 5)+ claude_session(f"{g}/rusty-search", "Benchmark FTS5 trigram vs LIKE fallback", [+ ("tool", "Bash", {"command": "cargo bench --bench search"}, "trigram: 0.9ms like: 41ms (280k rows)"),+ ("assistant", "Trigram wins by ~45x on realistic data; keeping LIKE only as the short-query fallback."),+ ], 60 * 49),+ claude_session(f"{g}/rusty-search", "重构索引重建的进度上报", [+ ("assistant", "把进度事件改为 Drop guard 兜底,任何提前返回路径都保证发出终态——UI 不会再卡在加载态。"),+ ], 60 * 96)+ claude_session(f"{g}/blog-engine", "Migrate CI from CircleCI to GitHub Actions", [+ ("tool", "Write", {"file_path": ".github/workflows/ci.yml", "content": "name: CI"}, "ok"),+ ("assistant", "Workflow ported; cache hit brings the build from 9m to 80s."),+ ], 60 * 72)+ claude_session(f"{g}/blog-engine", "Write integration tests for the RSS feed", [+ ("assistant", "Added 6 cases incl. CDATA titles and pubDate timezones — found one real bug in the encoder."),+ ], 60 * 24 * 6)+ claude_session(f"{g}/dotfiles", "Set up tmux + zsh keybindings", [+ ("assistant", "Prefix moved to C-a, vi-mode copy bindings, and fzf history on C-r."),+ ], 60 * 24 * 8)+ claude_session(f"{g}/oss-metrics", "Debug memory leak in the image pipeline", [+ ("tool", "Bash", {"command": "leaks --atExit -- ./target/debug/pipeline"}, "3 leaks for 48KB: decoder ring buffer"),+ ("assistant", "The decoder ring buffer is never drained on the error path — fixed with a Drop impl."),+ ], 60 * 24 * 10)+ claude_session(f"{g}/oss-metrics", "Paginate the contributors API", [+ ("assistant", "Cursor-based pagination with a stable sort key; off-by-one on the last page fixed."),+ ], 60 * 24 * 12)+ claude_session(f"{g}/oss-metrics", "优化首屏加载速度", [+ ("assistant", "关键路径上的三个串行请求并行化,LCP 从 2.8s 降到 1.1s;字体子集化又省了 180KB。"),+ ], 60 * 24 * 15)++ codex_session(f"{g}/acme-web", "Refactor the plugin loader to async", "Loader is fully async now; startup no longer blocks on plugin IO.", 60 * 7,+ tool=(["rg", "load_plugin", "src"], "src/plugins.rs:41: pub fn load_plugin(...)"))+ codex_session(f"{g}/acme-web", "Fix off-by-one in pagination", "The cursor skipped the boundary row; inclusive compare fixes it.", 60 * 36)+ codex_session(f"{g}/rusty-search", "Add a benchmark harness for the scanner", "Criterion harness added with three fixed-load scenarios.", 60 * 48,+ tool=(["cargo", "bench"], "scan_cold: 4.1s scan_warm: 0.09s"))+ codex_session(f"{g}/blog-engine", "Upgrade to Tailwind v4", "Migrated config to CSS-first; purge list no longer needed.", 60 * 24 * 5)+ codex_session(f"{g}/oss-metrics", "Profile the slow cold start", "Cold start was dominated by sync font loading; now deferred.", 60 * 24 * 9)++ # cursor / kiro / gemini:复制 fixture 骨架,替换成演示文案+ copy_fixture_tree(+ "cursor/projects/wakefx-cursor-proj/agent-transcripts/33333333-aaaa-bbbb-cccc-000000000003/33333333-aaaa-bbbb-cccc-000000000003.jsonl",+ ".cursor/projects/demo-acme-web/agent-transcripts/33333333-aaaa-bbbb-cccc-000000000003/33333333-aaaa-bbbb-cccc-000000000003.jsonl",+ [("wakefx", "acme-web"),+ ("把二维码扫描组件抽出来,注意 useEffect() 的清理", "Extract the QR scanner into a reusable hook"),+ ("先看现有组件结构,再抽公共 hook。", "Looking at the component structure first, then extracting a shared hook.")])+ copy_fixture_tree(+ "kiro/sessions/cli/44444444-aaaa-bbbb-cccc-000000000004.jsonl",+ ".kiro/sessions/cli/44444444-aaaa-bbbb-cccc-000000000004.jsonl",+ [("用 Kiro 重构二维码扫描,注意 useEffect() 清理", "Fix the CLI arg parser on quoted paths"),+ ("已拆分扫描组件并补上清理逻辑。", "Quoted paths now round-trip through the parser; added a regression test.")])+ copy_fixture_tree(+ "kiro/sessions/cli/44444444-aaaa-bbbb-cccc-000000000004.json",+ ".kiro/sessions/cli/44444444-aaaa-bbbb-cccc-000000000004.json",+ [("/Users/tester/Github/wakefx", f"{g}/dotfiles"),+ ("Kiro QR session", "Fix the CLI arg parser on quoted paths")])+ copy_fixture_tree(+ "gemini/tmp/wakefx-gem/chats/session-2026-08-04T12-00-00.jsonl",+ ".gemini/tmp/demo-gem/chats/session-2026-08-04T12-00-00.jsonl",+ [("wakefx-gem", "demo-gem"),+ ("Gemini 帮我调试二维码解码,顺带看 useEffect()", "Why does the sitemap generator skip drafts?"),+ ("可以,先在解码回调里打日志。", "Drafts carry a null publish date — the generator filters on it; I added a flag to include them.")])+ os.makedirs(os.path.join(HOME, ".gemini"), exist_ok=True)+ with open(os.path.join(HOME, ".gemini", "projects.json"), "w") as f:+ json.dump({"projects": {f"{g}/blog-engine": "demo-gem"}}, f)++ build_sqlite_agents()+ print(f"demo home ready: {HOME}")+ print(f"run: HOME={HOME} {REPO}/dist/Wake.app/Contents/MacOS/Wake")+++if __name__ == "__main__":+ main()