424 lines
19 KiB
Python
424 lines
19 KiB
Python
"""Kiro Agent 执行日志全量解析器 - 最终版 v2
|
||
修正 Hash 规律分析，基于实际验证结果。
|
||
"""
|
||
import json
|
||
from pathlib import Path
|
||
from datetime import datetime, timezone, timedelta
|
||
|
||
# === 路径 ===
|
||
BASE = Path(r"C:\temp\FX\globalStorage\kiro.kiroagent")
|
||
EXEC_ID_DIR = "4d8f681b2b78799de676ab36904d08c7"
|
||
ACTION_LOG_ID = "414d1636299d2b9e4ce7e17fb11f63e9"
|
||
LOG_FILE_HASH = "794adc0617d71b2231c20ac9d101c7a6"
|
||
DIFF_PARENT_HASH = "74a08cf8613c7dec4db7b264470db812"
|
||
|
||
LOG_PATH = BASE / EXEC_ID_DIR / ACTION_LOG_ID / LOG_FILE_HASH
|
||
DIFF_DIR = BASE / EXEC_ID_DIR / DIFF_PARENT_HASH
|
||
SESSION_DIR = BASE / "workspace-sessions" / "ZDpcVXNlcnNcTmVvXGRlc2t0b3BcVA__"
|
||
SESSION_JSON = SESSION_DIR / "dba26892-8e58-447f-ad82-bc9459a51753.json"
|
||
|
||
OUT_MD = Path(r"C:\temp\FX\kiro_full_session_record.md")
|
||
|
||
CST = timezone(timedelta(hours=8))
|
||
|
||
def ts(ms):
|
||
if not ms: return "N/A"
|
||
return datetime.fromtimestamp(ms / 1000, tz=CST).strftime("%Y-%m-%d %H:%M:%S")
|
||
|
||
def trunc(s, n=3000):
|
||
if not isinstance(s, str): return str(s)
|
||
return s if len(s) <= n else s[:n] + f"\n... [截断，原文共 {len(s)} 字符]"
|
||
|
||
def safe_json(obj, n=5000):
|
||
s = json.dumps(obj, ensure_ascii=False, indent=2)
|
||
return s if len(s) <= n else s[:n] + f"\n... [截断，原文共 {len(s)} 字符]"
|
||
|
||
|
||
def parse_messages(messages):
|
||
conversation = []
|
||
for i, msg in enumerate(messages):
|
||
entries = msg.get("entries", [])
|
||
parsed = []
|
||
for entry in entries:
|
||
if not isinstance(entry, dict):
|
||
continue
|
||
etype = entry.get("type", "unknown")
|
||
if etype == "text":
|
||
parsed.append({"type": "text", "text": entry.get("text", "")})
|
||
elif etype == "toolUse":
|
||
parsed.append({
|
||
"type": "toolUse", "id": entry.get("id"),
|
||
"name": entry.get("name"), "args": entry.get("args", {}),
|
||
"requestMessageId": entry.get("requestMessageId"),
|
||
})
|
||
elif etype == "toolUseResponse":
|
||
parsed.append({
|
||
"type": "toolUseResponse", "id": entry.get("id"),
|
||
"name": entry.get("name"), "args": entry.get("args", {}),
|
||
"message": entry.get("message", ""), "success": entry.get("success"),
|
||
})
|
||
elif etype == "document":
|
||
doc = entry.get("document", {})
|
||
parsed.append({
|
||
"type": "document", "doc_type": doc.get("type"),
|
||
"target": doc.get("target"), "doc_keys": list(doc.keys()),
|
||
})
|
||
else:
|
||
parsed.append({"type": etype, "raw_keys": list(entry.keys())})
|
||
conversation.append({
|
||
"index": i, "role": msg.get("role", "?"),
|
||
"messageId": msg.get("messageId", "?"),
|
||
"forcedRole": msg.get("forcedRole"), "entries": parsed,
|
||
})
|
||
return conversation
|
||
|
||
def parse_actions(actions):
|
||
timeline = []
|
||
for i, action in enumerate(actions):
|
||
entry = {
|
||
"index": i, "actionId": action.get("actionId"),
|
||
"actionType": action.get("actionType"),
|
||
"actionState": action.get("actionState"),
|
||
"emittedAt": ts(action.get("emittedAt")),
|
||
}
|
||
if action.get("subExecutionId"):
|
||
entry["subExecutionId"] = action["subExecutionId"]
|
||
if action.get("endTime"):
|
||
entry["endTime"] = ts(action["endTime"])
|
||
for k in ("intentResult", "input", "output"):
|
||
if k in action:
|
||
entry[k] = action[k]
|
||
timeline.append(entry)
|
||
return timeline
|
||
|
||
def collect_diffs():
|
||
diffs = {}
|
||
if not DIFF_DIR.exists():
|
||
return diffs
|
||
for sub in sorted(DIFF_DIR.iterdir()):
|
||
if sub.is_dir():
|
||
for f in sub.iterdir():
|
||
if f.is_file():
|
||
diffs[f"{sub.name}/{f.name}"] = {
|
||
"size": f.stat().st_size,
|
||
"content": f.read_text(encoding="utf-8", errors="replace"),
|
||
}
|
||
return diffs
|
||
|
||
def collect_ids(log, conversation, timeline):
|
||
refs = {
|
||
"executionId": log.get("executionId"),
|
||
"chatSessionId": log.get("chatSessionId"),
|
||
"actionIds": [], "subExecutionIds": [],
|
||
"toolUseIds": [], "messageIds": [],
|
||
"requestMessageIds": [], "diff_hashes": [],
|
||
}
|
||
for step in timeline:
|
||
for k, lst in [("actionId", "actionIds"), ("subExecutionId", "subExecutionIds")]:
|
||
v = step.get(k)
|
||
if v and v not in refs[lst]:
|
||
refs[lst].append(v)
|
||
for msg in conversation:
|
||
mid = msg.get("messageId")
|
||
if mid and mid not in refs["messageIds"]:
|
||
refs["messageIds"].append(mid)
|
||
for entry in msg.get("entries", []):
|
||
for k, lst in [("id", "toolUseIds"), ("requestMessageId", "requestMessageIds")]:
|
||
v = entry.get(k)
|
||
if v and v not in refs[lst]:
|
||
refs[lst].append(v)
|
||
if DIFF_DIR.exists():
|
||
refs["diff_hashes"] = [s.name for s in sorted(DIFF_DIR.iterdir()) if s.is_dir()]
|
||
return refs
|
||
|
||
|
||
def gen_md(meta, input_msg, conversation, timeline, diffs, ids, usage, session_info):
|
||
L = []
|
||
L.append("# Kiro Agent 会话全量记录\n")
|
||
L.append(f"> 生成时间: {datetime.now(CST).strftime('%Y-%m-%d %H:%M:%S')} CST\n")
|
||
|
||
# --- 1. 元数据 ---
|
||
L.append("## 1. 会话元数据\n")
|
||
L.append("| 字段 | 值 |")
|
||
L.append("|------|-----|")
|
||
for k, v in meta.items():
|
||
L.append(f"| {k} | `{v}` |")
|
||
L.append("")
|
||
if session_info:
|
||
L.append("### 会话配置\n")
|
||
L.append(f"- title: `{session_info.get('title')}`")
|
||
L.append(f"- dateCreated: `{ts(int(session_info.get('dateCreated', 0)))}`")
|
||
L.append(f"- workspaceDirectory: `{session_info.get('workspaceDirectory')}`")
|
||
L.append("")
|
||
|
||
# --- 2. 用户原始输入 ---
|
||
L.append("## 2. 用户原始输入\n")
|
||
L.append(f"```\n{trunc(input_msg, 5000)}\n```\n")
|
||
|
||
# --- 3. ID 引用关系 ---
|
||
L.append("## 3. ID/Hash 引用关系图\n")
|
||
L.append("### 3.1 目录层级 Hash（已验证）\n")
|
||
L.append("| Hash | 位置 | 验证结果 |")
|
||
L.append("|------|------|----------|")
|
||
L.append(f"| `ZDpcVXNlcnNcTmVvXGRlc2t0b3BcVA__` | workspace-sessions 子目录 | ✅ Base64url 编码 workspace 路径 `d:\\Users\\Neo\\desktop\\T`，`==` 填充替换为 `__` |")
|
||
L.append(f"| `dba26892-8e58-447f-ad82-bc9459a51753` | session 文件名 | ✅ UUID v4，即 chatSessionId |")
|
||
L.append(f"| `{EXEC_ID_DIR}` | 顶层目录 | ❓ 32 位 hex（MD5 格式），非 workspace 路径的 MD5，可能是 profile/account hash |")
|
||
L.append(f"| `{ACTION_LOG_ID}` | 二级目录 | ❓ 32 位 hex，非 executionId 的 MD5，可能是 execution 内部标识的 hash |")
|
||
L.append(f"| `{LOG_FILE_HASH}` | 日志文件名 | ❓ 32 位 hex，可能是固定 schema 标识或内容 hash |")
|
||
L.append(f"| `{DIFF_PARENT_HASH}` | diff 父目录 | ❓ 32 位 hex，与 execution 关联 |")
|
||
L.append(f"| `f62de366d0006e17ea00a01f6624aabf` | 执行索引文件 | ❓ 固定文件名，出现在 workspace 和 default 两处 |")
|
||
L.append(f"| diff 子目录（8位 hex） | diff 快照 | 文件内容的短 hash |")
|
||
L.append(f"| `tooluse_*` | toolUse entry id | `tooluse_` 前缀 + 随机串（如 `tooluse_fileTree`、`tooluse_DH4yvpYUCRxZsulG2G2bU1`） |")
|
||
L.append("")
|
||
|
||
L.append("### 3.2 运行时 ID\n")
|
||
L.append(f"- executionId: `{ids['executionId']}`")
|
||
L.append(f"- chatSessionId: `{ids['chatSessionId']}`")
|
||
L.append(f"- subExecutionIds: `{ids['subExecutionIds']}`")
|
||
L.append(f"\ntoolUseIds ({len(ids['toolUseIds'])} 个):\n")
|
||
for tid in ids["toolUseIds"]:
|
||
L.append(f"- `{tid}`")
|
||
L.append(f"\nmessageIds ({len(ids['messageIds'])} 个):\n")
|
||
for mid in ids["messageIds"]:
|
||
L.append(f"- `{mid}`")
|
||
L.append(f"\nrequestMessageIds ({len(ids['requestMessageIds'])} 个):\n")
|
||
for rmid in ids["requestMessageIds"]:
|
||
L.append(f"- `{rmid}`")
|
||
L.append(f"\ndiff_hashes: `{ids['diff_hashes']}`\n")
|
||
|
||
# --- 4. 对话记录 ---
|
||
L.append("## 4. 对话全量记录\n")
|
||
h = sum(1 for m in conversation if m['role']=='human')
|
||
b = sum(1 for m in conversation if m['role']=='bot')
|
||
t = sum(1 for m in conversation if m['role']=='tool')
|
||
L.append(f"共 {len(conversation)} 条消息：human={h}, bot={b}, tool={t}\n")
|
||
|
||
for msg in conversation:
|
||
emoji = {"human": "👤", "bot": "🤖", "tool": "🔧"}.get(msg["role"], "❓")
|
||
L.append(f"### Msg {msg['index']}: {emoji} {msg['role'].upper()} `{msg['messageId']}`\n")
|
||
|
||
for entry in msg["entries"]:
|
||
et = entry["type"]
|
||
if et == "text":
|
||
text = entry["text"]
|
||
if not text:
|
||
L.append("*(空)*\n")
|
||
elif len(text) > 10000 and msg["role"] == "human":
|
||
L.append(f"**[系统提示词]** ({len(text)} 字符)\n")
|
||
L.append("<details><summary>展开</summary>\n")
|
||
L.append(f"```\n{trunc(text, 20000)}\n```\n</details>\n")
|
||
else:
|
||
L.append(f"```\n{trunc(text, 5000)}\n```\n")
|
||
elif et == "toolUse":
|
||
L.append(f"**[🔧 调用]** `{entry['name']}` id=`{entry['id']}`\n")
|
||
L.append(f"```json\n{safe_json(entry['args'], 5000)}\n```\n")
|
||
elif et == "toolUseResponse":
|
||
ok = "✅" if entry.get("success") else "❌"
|
||
L.append(f"**[📋 结果]** `{entry['name']}` {ok} id=`{entry['id']}`\n")
|
||
if entry.get("message"):
|
||
L.append(f"```\n{trunc(entry['message'], 5000)}\n```\n")
|
||
if entry.get("args"):
|
||
L.append(f"<details><summary>完整参数</summary>\n")
|
||
L.append(f"```json\n{safe_json(entry['args'], 5000)}\n```\n</details>\n")
|
||
elif et == "document":
|
||
L.append(f"**[📄 文档]** type=`{entry.get('doc_type')}` target=`{entry.get('target')}`\n")
|
||
else:
|
||
L.append(f"**[{et}]** keys={entry.get('raw_keys')}\n")
|
||
|
||
# --- 5. Actions ---
|
||
L.append("## 5. Actions 时间线\n")
|
||
L.append(f"共 {len(timeline)} 个\n")
|
||
for step in timeline:
|
||
L.append(f"### Step {step['index']}: `{step.get('actionType','?')}` [{step.get('actionState','?')}] @ {step.get('emittedAt','?')}\n")
|
||
L.append(f"- actionId: `{step.get('actionId')}`")
|
||
if step.get("subExecutionId"):
|
||
L.append(f"- subExecutionId: `{step['subExecutionId']}`")
|
||
if step.get("endTime"):
|
||
L.append(f"- endTime: {step['endTime']}")
|
||
for k in ("intentResult", "input", "output"):
|
||
if k in step:
|
||
L.append(f"- {k}:\n```json\n{safe_json(step[k], 5000)}\n```")
|
||
L.append("")
|
||
|
||
# --- 6. Diff ---
|
||
L.append("## 6. 文件版本快照\n")
|
||
if diffs:
|
||
for path, info in diffs.items():
|
||
lang = "python" if path.endswith(".py") else "markdown" if path.endswith(".md") else ""
|
||
L.append(f"### `{path}` ({info['size']} bytes)\n")
|
||
L.append(f"```{lang}\n{trunc(info['content'], 5000)}\n```\n")
|
||
else:
|
||
L.append("*(无)*\n")
|
||
|
||
# --- 7. 资源消耗 ---
|
||
L.append("## 7. 资源消耗\n")
|
||
if usage:
|
||
L.append("| 工具 | 消耗 | 单位 |")
|
||
L.append("|------|------|------|")
|
||
total = 0
|
||
for u in usage:
|
||
tools = ", ".join(u.get("usedTools", ["-"]))
|
||
amt = u.get("usage", 0)
|
||
total += amt
|
||
L.append(f"| {tools} | {amt} | {u.get('unit','?')} |")
|
||
L.append(f"| **合计** | **{total:.4f}** | |")
|
||
L.append("")
|
||
|
||
# --- 8. 存储结构 ---
|
||
L.append("## 8. Kiro 日志存储结构\n")
|
||
L.append("```")
|
||
L.append("globalStorage/")
|
||
L.append("├── state.vscdb # VS Code 状态 SQLite")
|
||
L.append("├── state.vscdb.backup")
|
||
L.append("├── storage.json # 窗口/主题/工作区配置")
|
||
L.append("└── kiro.kiroagent/")
|
||
L.append(" ├── config.json # 上下文提供者列表")
|
||
L.append(" ├── profile.json # 用户 profile（ARN）")
|
||
L.append(" ├── .migrations/ # 迁移标记")
|
||
L.append(f" ├── <32位hex>/ # 推测: profile/account hash")
|
||
L.append(f" │ ├── f62de366... # 执行索引 JSON")
|
||
L.append(f" │ ├── <32位hex>/ # 推测: execution 相关 hash")
|
||
L.append(f" │ │ └── <32位hex> # 完整执行日志 JSON")
|
||
L.append(f" │ └── <32位hex>/ # diff 快照集合")
|
||
L.append(f" │ └── <8位hex>/ # 文件内容短 hash")
|
||
L.append(f" │ └── <filename> # 文件快照")
|
||
L.append(" ├── default/")
|
||
L.append(" │ └── f62de366... # 默认执行索引（空）")
|
||
L.append(" ├── dev_data/tokens_generated.jsonl")
|
||
L.append(" ├── index/")
|
||
L.append(" │ ├── docs.sqlite")
|
||
L.append(" │ └── globalContext.json")
|
||
L.append(" └── workspace-sessions/")
|
||
L.append(" └── <base64url(workspace_path)>/ # == 替换为 __")
|
||
L.append(" ├── sessions.json # 会话列表")
|
||
L.append(" └── <chatSessionId>.json # 会话状态")
|
||
L.append("```\n")
|
||
|
||
L.append("### Hash 命名规律总结\n")
|
||
L.append("| 类型 | 格式 | 规律 |")
|
||
L.append("|------|------|------|")
|
||
L.append("| workspace-sessions 子目录 | Base64url | workspace 绝对路径的 base64url 编码，`==` 填充替换为 `__` |")
|
||
L.append("| session 文件名 | UUID v4 | 即 chatSessionId |")
|
||
L.append("| 顶层 32 位 hex 目录 | MD5 格式 | 非 workspace 路径 MD5，推测为 profile ARN 或 account ID 的 hash |")
|
||
L.append("| 二级 32 位 hex 目录 | MD5 格式 | 非 executionId 的 MD5，推测为内部标识 hash |")
|
||
L.append("| 日志文件名 | 32 位 hex | 可能是固定 schema 标识 |")
|
||
L.append("| 执行索引文件名 | 32 位 hex | 固定值 `f62de366...`，出现在多处 |")
|
||
L.append("| diff 子目录 | 8 位 hex | 文件内容短 hash |")
|
||
L.append("| toolUse ID | `tooluse_` + 随机串 | 特殊的有 `tooluse_fileTree`（内置工具） |")
|
||
L.append("")
|
||
|
||
# --- 9. 数据 Schema ---
|
||
L.append("## 9. 执行日志 JSON Schema\n")
|
||
L.append("```")
|
||
L.append("{")
|
||
L.append(' "executionId": "UUID",')
|
||
L.append(' "chatSessionId": "UUID",')
|
||
L.append(' "workflowType": "chat-agent",')
|
||
L.append(' "autonomyMode": "Autopilot",')
|
||
L.append(' "status": "succeed",')
|
||
L.append(' "startTime": <毫秒时间戳>,')
|
||
L.append(' "endTime": <毫秒时间戳>,')
|
||
L.append(' "contextUsagePercentage": <float>,')
|
||
L.append(' "input": {')
|
||
L.append(' "data": {')
|
||
L.append(' "messages": [{ "role": "user", "content": [{ "text": "..." }] }],')
|
||
L.append(' "chatSessionId": "UUID"')
|
||
L.append(' },')
|
||
L.append(' "documents": []')
|
||
L.append(' },')
|
||
L.append(' "actions": [')
|
||
L.append(' {')
|
||
L.append(' "type": "AgentExecutionAction",')
|
||
L.append(' "executionId": "UUID",')
|
||
L.append(' "actionId": "UUID",')
|
||
L.append(' "actionType": "intent|tool|generation|subAgent",')
|
||
L.append(' "actionState": "start|succeed|failed",')
|
||
L.append(' "chatSessionId": "UUID",')
|
||
L.append(' "emittedAt": <毫秒时间戳>,')
|
||
L.append(' "endTime?": <毫秒时间戳>,')
|
||
L.append(' "subExecutionId?": "UUID",')
|
||
L.append(' "intentResult?": { "classification": "do|..." },')
|
||
L.append(' "input?": { "toolName": "...", ... },')
|
||
L.append(' "output?": { "text": "...", ... }')
|
||
L.append(' }')
|
||
L.append(' ],')
|
||
L.append(' "context": {')
|
||
L.append(' "messages": [')
|
||
L.append(' {')
|
||
L.append(' "role": "human|bot|tool",')
|
||
L.append(' "messageId": "UUID",')
|
||
L.append(' "forcedRole": null,')
|
||
L.append(' "entries": [')
|
||
L.append(' { "type": "text", "text": "..." },')
|
||
L.append(' { "type": "toolUse", "id": "tooluse_*", "name": "...", "args": {...}, "requestMessageId": "UUID" },')
|
||
L.append(' { "type": "toolUseResponse", "id": "tooluse_*", "name": "...", "args": {...}, "message": "...", "success": bool },')
|
||
L.append(' { "type": "document", "document": { "type": "...", "target": "...", ... } }')
|
||
L.append(' ]')
|
||
L.append(' }')
|
||
L.append(' ]')
|
||
L.append(' },')
|
||
L.append(' "usageSummary": [')
|
||
L.append(' { "usedTools?": ["toolName"], "unit": "credit", "unitPlural": "credits", "usage": <float> }')
|
||
L.append(' ],')
|
||
L.append(' "result": { "status": "success", "executionId": "UUID", "result": "..." }')
|
||
L.append("}")
|
||
L.append("```\n")
|
||
|
||
return "\n".join(L)
|
||
|
||
|
||
|
||
def main():
|
||
print("=== Kiro Agent 日志全量解析 v2 ===\n")
|
||
with open(LOG_PATH, "r", encoding="utf-8") as f:
|
||
log = json.load(f)
|
||
print(f"日志: {LOG_PATH.stat().st_size / 1024:.1f} KB")
|
||
|
||
meta = {
|
||
"executionId": log.get("executionId"),
|
||
"chatSessionId": log.get("chatSessionId"),
|
||
"workflowType": log.get("workflowType"),
|
||
"autonomyMode": log.get("autonomyMode"),
|
||
"status": log.get("status"),
|
||
"startTime": ts(log.get("startTime")),
|
||
"endTime": ts(log.get("endTime")),
|
||
"duration": f"{(log.get('endTime',0) - log.get('startTime',0)) / 1000:.1f}s",
|
||
"contextUsage": f"{log.get('contextUsagePercentage',0):.2f}%",
|
||
}
|
||
|
||
# 用户输入
|
||
input_text = ""
|
||
for msg in log.get("input",{}).get("data",{}).get("messages",[]):
|
||
for entry in msg.get("content", msg.get("entries", [])):
|
||
if isinstance(entry, dict) and entry.get("text"):
|
||
input_text += entry["text"] + "\n"
|
||
|
||
conversation = parse_messages(log.get("context",{}).get("messages",[]))
|
||
timeline = parse_actions(log.get("actions",[]))
|
||
diffs = collect_diffs()
|
||
ids = collect_ids(log, conversation, timeline)
|
||
|
||
session_info = None
|
||
if SESSION_JSON.exists():
|
||
with open(SESSION_JSON, "r", encoding="utf-8") as f:
|
||
session_info = json.load(f)
|
||
|
||
usage = log.get("usageSummary", [])
|
||
|
||
md = gen_md(meta, input_text, conversation, timeline, diffs, ids, usage, session_info)
|
||
with open(OUT_MD, "w", encoding="utf-8") as f:
|
||
f.write(md)
|
||
|
||
lines = len(md.splitlines())
|
||
size_kb = OUT_MD.stat().st_size / 1024
|
||
print(f"输出: {OUT_MD} ({size_kb:.1f} KB, {lines} 行)")
|
||
print(f"消息: human={sum(1 for m in conversation if m['role']=='human')}, "
|
||
f"bot={sum(1 for m in conversation if m['role']=='bot')}, "
|
||
f"tool={sum(1 for m in conversation if m['role']=='tool')}")
|
||
print(f"工具调用: {sum(1 for m in conversation for e in m['entries'] if e['type']=='toolUse')}")
|
||
print(f"Actions: {len(timeline)}, Diffs: {len(diffs)}")
|
||
print(f"toolUseIds: {len(ids['toolUseIds'])}, messageIds: {len(ids['messageIds'])}")
|
||
|
||
if __name__ == "__main__":
|
||
main()
|