| 1 | from __future__ import annotations |
| 2 | |
| 3 | import hashlib |
| 4 | import io |
| 5 | import json |
| 6 | import re |
| 7 | import sqlite3 |
| 8 | import stat |
| 9 | import tarfile |
| 10 | import tempfile |
| 11 | import uuid |
| 12 | import zipfile |
| 13 | from dataclasses import dataclass, field |
| 14 | from datetime import datetime, timezone |
| 15 | from pathlib import Path, PurePosixPath |
| 16 | from typing import Any, Iterable |
| 17 | |
| 18 | |
| 19 | MAX_FILES = 5_000 |
| 20 | MAX_FILE_BYTES = 100 * 1024 * 1024 |
| 21 | MAX_TOTAL_BYTES = 256 * 1024 * 1024 |
| 22 | TEXT_SUFFIXES = { |
| 23 | ".css", |
| 24 | ".html", |
| 25 | ".js", |
| 26 | ".json", |
| 27 | ".jsonl", |
| 28 | ".md", |
| 29 | ".py", |
| 30 | ".sh", |
| 31 | ".toml", |
| 32 | ".ts", |
| 33 | ".txt", |
| 34 | ".yaml", |
| 35 | ".yml", |
| 36 | } |
| 37 | MEMORY_NAMES = { |
| 38 | "memory.md", |
| 39 | } |
| 40 | INSTRUCTION_NAMES = { |
| 41 | "agents.md", |
| 42 | "claude.md", |
| 43 | "identity.md", |
| 44 | "soul.md", |
| 45 | "tools.md", |
| 46 | "user.md", |
| 47 | } |
| 48 | PROJECT_PATH_KEYS = ("cwd", "directory", "workspace") |
| 49 | SENSITIVE_NAMES = { |
| 50 | ".credentials.json", |
| 51 | ".env", |
| 52 | "auth.json", |
| 53 | "credentials.json", |
| 54 | "openclaw.json", |
| 55 | "settings.json", |
| 56 | } |
| 57 | SECRET_RE = re.compile( |
| 58 | r"(?i)(\b(?:api[_-]?key|access[_-]?token|refresh[_-]?token|password|passwd|secret)\b[\"']?\s*[:=]\s*[\"']?)([^\s\"']+)" |
| 59 | ) |
| 60 | BEARER_RE = re.compile(r"(?i)(Authorization\s*:\s*Bearer\s+)([^\s\"']+)") |
| 61 | PRIVATE_KEY_RE = re.compile( |
| 62 | r"-----BEGIN [^-]*PRIVATE KEY-----[\s\S]*?-----END [^-]*PRIVATE KEY-----" |
| 63 | ) |
| 64 | DATA_URL_RE = re.compile(r"data:[^;,\s]+;base64,[A-Za-z0-9+/=_-]+") |
| 65 | |
| 66 | |
| 67 | @dataclass(slots=True) |
| 68 | class Upload: |
| 69 | name: str |
| 70 | data: bytes |
| 71 | |
| 72 | |
| 73 | @dataclass(slots=True) |
| 74 | class Event: |
| 75 | kind: str |
| 76 | text: str = "" |
| 77 | timestamp: float = 0.0 |
| 78 | thoughts: list[str] = field(default_factory=list) |
| 79 | tool_name: str = "" |
| 80 | tool_args: dict[str, Any] = field(default_factory=dict) |
| 81 | tool_result: str = "" |
| 82 | |
| 83 | |
| 84 | @dataclass(slots=True) |
| 85 | class Conversation: |
| 86 | source_id: str |
| 87 | title: str |
| 88 | events: list[Event] |
| 89 | metadata: dict[str, Any] = field(default_factory=dict) |
| 90 | |
| 91 | |
| 92 | @dataclass(slots=True) |
| 93 | class Asset: |
| 94 | path: str |
| 95 | data: bytes |
| 96 | |
| 97 | |
| 98 | @dataclass(slots=True) |
| 99 | class Project: |
| 100 | source_id: str |
| 101 | title: str |
| 102 | path: str |
| 103 | conversation_ids: list[str] = field(default_factory=list) |
| 104 | |
| 105 | |
| 106 | @dataclass(slots=True) |
| 107 | class Bundle: |
| 108 | source: str |
| 109 | conversations: list[Conversation] = field(default_factory=list) |
| 110 | projects: list[Project] = field(default_factory=list) |
| 111 | memories: list[Asset] = field(default_factory=list) |
| 112 | instructions: list[Asset] = field(default_factory=list) |
| 113 | skills: dict[str, list[Asset]] = field(default_factory=dict) |
| 114 | excluded: list[str] = field(default_factory=list) |
| 115 | warnings: list[str] = field(default_factory=list) |
| 116 | redactions: int = 0 |
| 117 | |
| 118 | @property |
| 119 | def knowledge(self) -> list[Asset]: |
| 120 | return [*self.memories, *self.instructions] |
| 121 | |
| 122 | def summary(self) -> dict[str, int]: |
| 123 | return { |
| 124 | "chats": len(self.conversations), |
| 125 | "projects": len(self.projects), |
| 126 | "messages": sum(len(item.events) for item in self.conversations), |
| 127 | "memories": len(self.memories), |
| 128 | "instructions": len(self.instructions), |
| 129 | "knowledge": len(self.memories) + len(self.instructions), |
| 130 | "skills": len(self.skills), |
| 131 | "excluded": len(self.excluded), |
| 132 | "redactions": self.redactions, |
| 133 | } |
| 134 | |
| 135 | |
| 136 | def _safe_name(name: str) -> str: |
| 137 | value = str(PurePosixPath(name.replace("\\", "/"))) |
| 138 | path = PurePosixPath(value) |
| 139 | if path.is_absolute() or ".." in path.parts: |
| 140 | raise ValueError(f"Unsafe archive path: {name!r}") |
| 141 | return value.lstrip("./") |
| 142 | |
| 143 | |
| 144 | def expand_uploads(uploads: Iterable[Upload]) -> list[Upload]: |
| 145 | result: list[Upload] = [] |
| 146 | total = 0 |
| 147 | |
| 148 | def add(name: str, data: bytes) -> None: |
| 149 | nonlocal total |
| 150 | clean = _safe_name(name) |
| 151 | if not clean or clean.endswith("/"): |
| 152 | return |
| 153 | if len(data) > MAX_FILE_BYTES: |
| 154 | raise ValueError(f"File exceeds the 100 MiB limit: {clean}") |
| 155 | total += len(data) |
| 156 | if total > MAX_TOTAL_BYTES: |
| 157 | raise ValueError("Expanded upload exceeds the 256 MiB limit") |
| 158 | result.append(Upload(clean, data)) |
| 159 | if len(result) > MAX_FILES: |
| 160 | raise ValueError(f"Upload contains more than {MAX_FILES} files") |
| 161 | |
| 162 | for upload in uploads: |
| 163 | lower = upload.name.lower() |
| 164 | if lower.endswith(".zip"): |
| 165 | with zipfile.ZipFile(io.BytesIO(upload.data)) as archive: |
| 166 | for member in archive.infolist(): |
| 167 | mode = member.external_attr >> 16 |
| 168 | if stat.S_ISLNK(mode): |
| 169 | raise ValueError(f"Archive symlinks are not accepted: {member.filename}") |
| 170 | if member.is_dir(): |
| 171 | continue |
| 172 | add(member.filename, archive.read(member)) |
| 173 | elif lower.endswith((".tar", ".tar.gz", ".tgz")): |
| 174 | with tarfile.open(fileobj=io.BytesIO(upload.data), mode="r:*") as archive: |
| 175 | for member in archive.getmembers(): |
| 176 | if member.issym() or member.islnk() or member.isdev(): |
| 177 | raise ValueError(f"Archive links and devices are not accepted: {member.name}") |
| 178 | if not member.isfile(): |
| 179 | continue |
| 180 | source = archive.extractfile(member) |
| 181 | if source is not None: |
| 182 | add(member.name, source.read(MAX_FILE_BYTES + 1)) |
| 183 | else: |
| 184 | add(upload.name, upload.data) |
| 185 | return result |
| 186 | |
| 187 | |
| 188 | def _text(asset: Upload | Asset) -> str: |
| 189 | return asset.data.decode("utf-8-sig", "replace") |
| 190 | |
| 191 | |
| 192 | def _redact(value: str) -> tuple[str, int]: |
| 193 | count = 0 |
| 194 | |
| 195 | def secret(match: re.Match[str]) -> str: |
| 196 | nonlocal count |
| 197 | if match.group(2).startswith("$"): |
| 198 | return match.group(0) |
| 199 | count += 1 |
| 200 | return f"{match.group(1)}[REDACTED]" |
| 201 | |
| 202 | value = SECRET_RE.sub(secret, value) |
| 203 | value, bearer_count = BEARER_RE.subn(r"\1[REDACTED]", value) |
| 204 | value, key_count = PRIVATE_KEY_RE.subn("[PRIVATE KEY REDACTED]", value) |
| 205 | value, data_count = DATA_URL_RE.subn("[EMBEDDED DATA OMITTED]", value) |
| 206 | return value, count + bearer_count + key_count + data_count |
| 207 | |
| 208 | |
| 209 | def _timestamp(value: Any, fallback: float = 0.0) -> float: |
| 210 | if isinstance(value, (int, float)): |
| 211 | number = float(value) |
| 212 | return number / 1000 if number > 10_000_000_000 else number |
| 213 | if isinstance(value, str) and value: |
| 214 | try: |
| 215 | return datetime.fromisoformat(value.replace("Z", "+00:00")).timestamp() |
| 216 | except ValueError: |
| 217 | pass |
| 218 | return fallback or datetime.now(timezone.utc).timestamp() |
| 219 | |
| 220 | |
| 221 | def _one_line(value: str, limit: int = 92) -> str: |
| 222 | text = re.sub(r"\s+", " ", value).strip() |
| 223 | return text if len(text) <= limit else text[: limit - 1].rstrip() + "…" |
| 224 | |
| 225 | |
| 226 | def _content_text(content: Any) -> str: |
| 227 | if isinstance(content, str): |
| 228 | return content |
| 229 | if isinstance(content, dict): |
| 230 | for key in ("text", "content", "user_message"): |
| 231 | if isinstance(content.get(key), str): |
| 232 | return content[key] |
| 233 | return "" |
| 234 | if isinstance(content, list): |
| 235 | parts: list[str] = [] |
| 236 | for item in content: |
| 237 | if isinstance(item, str): |
| 238 | parts.append(item) |
| 239 | elif isinstance(item, dict) and item.get("type") not in {"thinking", "reasoning"}: |
| 240 | value = item.get("text") or item.get("content") |
| 241 | if isinstance(value, str): |
| 242 | parts.append(value) |
| 243 | return "\n".join(part for part in parts if part) |
| 244 | return "" |
| 245 | |
| 246 | |
| 247 | def _tool_blocks(content: Any) -> list[dict[str, Any]]: |
| 248 | if not isinstance(content, list): |
| 249 | return [] |
| 250 | return [ |
| 251 | item |
| 252 | for item in content |
| 253 | if isinstance(item, dict) and item.get("type") in {"tool_use", "toolCall", "tool-call"} |
| 254 | ] |
| 255 | |
| 256 | |
| 257 | def _events_from_messages(messages: Iterable[dict[str, Any]]) -> list[Event]: |
| 258 | events: list[Event] = [] |
| 259 | pending_tools: dict[str, Event] = {} |
| 260 | for index, message in enumerate(messages): |
| 261 | role = str(message.get("role") or message.get("type") or "").lower() |
| 262 | content = message.get("content") |
| 263 | when = _timestamp(message.get("timestamp") or message.get("created_at"), index + 1) |
| 264 | if role in {"user", "human"}: |
| 265 | text = _content_text(content) |
| 266 | if text: |
| 267 | events.append(Event("user", text=text, timestamp=when)) |
| 268 | if isinstance(content, list): |
| 269 | for block in content: |
| 270 | if not isinstance(block, dict) or block.get("type") != "tool_result": |
| 271 | continue |
| 272 | tool_id = str(block.get("tool_use_id") or "") |
| 273 | result = _content_text(block.get("content")) |
| 274 | if tool_id in pending_tools: |
| 275 | pending_tools[tool_id].tool_result = result |
| 276 | continue |
| 277 | if role in {"assistant", "agent", "ai"}: |
| 278 | text = _content_text(content) |
| 279 | if text: |
| 280 | events.append(Event("assistant", text=text, timestamp=when)) |
| 281 | for block in _tool_blocks(content): |
| 282 | tool_id = str(block.get("id") or block.get("tool_call_id") or "") |
| 283 | event = Event( |
| 284 | "tool", |
| 285 | timestamp=when, |
| 286 | tool_name=str(block.get("name") or "tool"), |
| 287 | tool_args=block.get("input") if isinstance(block.get("input"), dict) else {}, |
| 288 | ) |
| 289 | events.append(event) |
| 290 | if tool_id: |
| 291 | pending_tools[tool_id] = event |
| 292 | calls = message.get("tool_calls") |
| 293 | if isinstance(calls, str): |
| 294 | try: |
| 295 | calls = json.loads(calls) |
| 296 | except json.JSONDecodeError: |
| 297 | calls = [] |
| 298 | for call in calls or []: |
| 299 | if not isinstance(call, dict): |
| 300 | continue |
| 301 | function = call.get("function") if isinstance(call.get("function"), dict) else call |
| 302 | raw_args = function.get("arguments") or {} |
| 303 | if isinstance(raw_args, str): |
| 304 | try: |
| 305 | raw_args = json.loads(raw_args) |
| 306 | except json.JSONDecodeError: |
| 307 | raw_args = {"raw_arguments": raw_args} |
| 308 | event = Event( |
| 309 | "tool", |
| 310 | timestamp=when, |
| 311 | tool_name=str(function.get("name") or "tool"), |
| 312 | tool_args=raw_args if isinstance(raw_args, dict) else {"value": raw_args}, |
| 313 | ) |
| 314 | events.append(event) |
| 315 | tool_id = str(call.get("id") or "") |
| 316 | if tool_id: |
| 317 | pending_tools[tool_id] = event |
| 318 | continue |
| 319 | if role in {"tool", "tool_result"}: |
| 320 | tool_id = str(message.get("tool_call_id") or message.get("id") or "") |
| 321 | result = _content_text(content) |
| 322 | if tool_id in pending_tools: |
| 323 | pending_tools[tool_id].tool_result = result |
| 324 | else: |
| 325 | events.append( |
| 326 | Event( |
| 327 | "tool", |
| 328 | timestamp=when, |
| 329 | tool_name=str(message.get("tool_name") or message.get("name") or "tool"), |
| 330 | tool_result=result, |
| 331 | ) |
| 332 | ) |
| 333 | return events |
| 334 | |
| 335 | |
| 336 | def _conversation(source_id: str, title: str, events: list[Event], **metadata: Any) -> Conversation | None: |
| 337 | visible = [event for event in events if event.kind == "tool" or event.text.strip()] |
| 338 | if not visible: |
| 339 | return None |
| 340 | first_user = next((event.text for event in visible if event.kind == "user"), source_id) |
| 341 | return Conversation(source_id, _one_line(title or first_user or source_id), visible, metadata) |
| 342 | |
| 343 | |
| 344 | def _json_lines(text: str) -> list[dict[str, Any]]: |
| 345 | rows: list[dict[str, Any]] = [] |
| 346 | for line_no, line in enumerate(text.splitlines(), 1): |
| 347 | if not line.strip(): |
| 348 | continue |
| 349 | try: |
| 350 | value = json.loads(line) |
| 351 | except json.JSONDecodeError as exc: |
| 352 | raise ValueError(f"Invalid JSONL at line {line_no}: {exc.msg}") from exc |
| 353 | if isinstance(value, dict): |
| 354 | rows.append(value) |
| 355 | return rows |
| 356 | |
| 357 | |
| 358 | def _parse_codex(asset: Upload) -> list[Conversation]: |
| 359 | rows = _json_lines(_text(asset)) |
| 360 | metadata: dict[str, Any] = {} |
| 361 | events: list[Event] = [] |
| 362 | pending: dict[str, Event] = {} |
| 363 | thoughts: list[str] = [] |
| 364 | for index, row in enumerate(rows): |
| 365 | payload = row.get("payload") if isinstance(row.get("payload"), dict) else {} |
| 366 | if row.get("type") == "session_meta" and not metadata: |
| 367 | metadata = dict(payload) |
| 368 | continue |
| 369 | if row.get("type") == "event_msg": |
| 370 | event_type = payload.get("type") |
| 371 | when = _timestamp(row.get("timestamp"), index + 1) |
| 372 | text = payload.get("message") |
| 373 | if not isinstance(text, str) or not text.strip(): |
| 374 | continue |
| 375 | if event_type == "user_message": |
| 376 | events.append(Event("user", text=text, timestamp=when)) |
| 377 | elif event_type == "agent_message" and payload.get("phase") == "commentary": |
| 378 | thoughts.append(text) |
| 379 | elif event_type == "agent_message" and payload.get("phase") == "final_answer": |
| 380 | events.append(Event("assistant", text=text, timestamp=when, thoughts=thoughts.copy())) |
| 381 | thoughts.clear() |
| 382 | continue |
| 383 | if row.get("type") != "response_item": |
| 384 | continue |
| 385 | item_type = payload.get("type") |
| 386 | call_id = str(payload.get("call_id") or payload.get("id") or "") |
| 387 | when = _timestamp(row.get("timestamp"), index + 1) |
| 388 | if item_type in {"function_call", "custom_tool_call"}: |
| 389 | raw = payload.get("arguments") if item_type == "function_call" else payload.get("input") |
| 390 | if isinstance(raw, str): |
| 391 | try: |
| 392 | args = json.loads(raw) |
| 393 | except json.JSONDecodeError: |
| 394 | args = {"input": raw} |
| 395 | else: |
| 396 | args = raw if isinstance(raw, dict) else {} |
| 397 | event = Event( |
| 398 | "tool", |
| 399 | timestamp=when, |
| 400 | thoughts=thoughts.copy(), |
| 401 | tool_name=str(payload.get("name") or "tool"), |
| 402 | tool_args=args, |
| 403 | ) |
| 404 | thoughts.clear() |
| 405 | events.append(event) |
| 406 | if call_id: |
| 407 | pending[call_id] = event |
| 408 | elif item_type in {"function_call_output", "custom_tool_call_output"} and call_id in pending: |
| 409 | pending[call_id].tool_result = _content_text(payload.get("output")) or str(payload.get("output") or "") |
| 410 | item = _conversation( |
| 411 | str(metadata.get("id") or metadata.get("session_id") or Path(asset.name).stem), |
| 412 | "", |
| 413 | events, |
| 414 | cwd=metadata.get("cwd"), |
| 415 | originator=metadata.get("originator"), |
| 416 | ) |
| 417 | return [item] if item else [] |
| 418 | |
| 419 | |
| 420 | def _parse_claude(asset: Upload) -> list[Conversation]: |
| 421 | rows = _json_lines(_text(asset)) |
| 422 | messages: list[dict[str, Any]] = [] |
| 423 | session_id = Path(asset.name).stem |
| 424 | cwd = "" |
| 425 | for row in rows: |
| 426 | if row.get("isSidechain") is True or row.get("type") not in {"user", "assistant"}: |
| 427 | continue |
| 428 | message = row.get("message") |
| 429 | if not isinstance(message, dict): |
| 430 | continue |
| 431 | messages.append({**message, "timestamp": row.get("timestamp")}) |
| 432 | session_id = str(row.get("sessionId") or session_id) |
| 433 | cwd = str(row.get("cwd") or cwd) |
| 434 | item = _conversation(session_id, "", _events_from_messages(messages), cwd=cwd) |
| 435 | return [item] if item else [] |
| 436 | |
| 437 | |
| 438 | def _parse_opencode(asset: Upload) -> list[Conversation]: |
| 439 | try: |
| 440 | data = json.loads(_text(asset)) |
| 441 | except json.JSONDecodeError: |
| 442 | return [] |
| 443 | if not isinstance(data, dict) or not isinstance(data.get("messages"), list): |
| 444 | return [] |
| 445 | info = data.get("info") if isinstance(data.get("info"), dict) else {} |
| 446 | events: list[Event] = [] |
| 447 | for index, wrapper in enumerate(data["messages"]): |
| 448 | if not isinstance(wrapper, dict): |
| 449 | continue |
| 450 | message = wrapper.get("info") if isinstance(wrapper.get("info"), dict) else wrapper |
| 451 | role = str(message.get("role") or "") |
| 452 | when = _timestamp((message.get("time") or {}).get("created") if isinstance(message.get("time"), dict) else None, index + 1) |
| 453 | parts = wrapper.get("parts") if isinstance(wrapper.get("parts"), list) else [] |
| 454 | text = "\n".join( |
| 455 | str(part.get("text")) |
| 456 | for part in parts |
| 457 | if isinstance(part, dict) and part.get("type") == "text" and part.get("text") |
| 458 | ) |
| 459 | if text and role in {"user", "assistant"}: |
| 460 | events.append(Event(role, text=text, timestamp=when)) |
| 461 | for part in parts: |
| 462 | if not isinstance(part, dict) or part.get("type") != "tool": |
| 463 | continue |
| 464 | state = part.get("state") if isinstance(part.get("state"), dict) else {} |
| 465 | args = state.get("input") if isinstance(state.get("input"), dict) else {} |
| 466 | events.append( |
| 467 | Event( |
| 468 | "tool", |
| 469 | timestamp=when, |
| 470 | tool_name=str(part.get("tool") or state.get("title") or "tool"), |
| 471 | tool_args=args, |
| 472 | tool_result=_content_text(state.get("output")), |
| 473 | ) |
| 474 | ) |
| 475 | item = _conversation( |
| 476 | str(info.get("id") or Path(asset.name).stem), |
| 477 | str(info.get("title") or ""), |
| 478 | events, |
| 479 | directory=info.get("directory"), |
| 480 | project_id=info.get("projectID"), |
| 481 | ) |
| 482 | return [item] if item else [] |
| 483 | |
| 484 | |
| 485 | def _parse_hermes_jsonl(asset: Upload) -> list[Conversation]: |
| 486 | result: list[Conversation] = [] |
| 487 | for row in _json_lines(_text(asset)): |
| 488 | messages = row.get("messages") |
| 489 | if not isinstance(messages, list): |
| 490 | continue |
| 491 | item = _conversation( |
| 492 | str(row.get("id") or row.get("session_id") or uuid.uuid4()), |
| 493 | str(row.get("title") or ""), |
| 494 | _events_from_messages(message for message in messages if isinstance(message, dict)), |
| 495 | source=row.get("source"), |
| 496 | cwd=row.get("cwd"), |
| 497 | model=row.get("model"), |
| 498 | ) |
| 499 | if item: |
| 500 | result.append(item) |
| 501 | return result |
| 502 | |
| 503 | |
| 504 | def _parse_openclaw_jsonl(asset: Upload) -> list[Conversation]: |
| 505 | rows = _json_lines(_text(asset)) |
| 506 | session_id = Path(asset.name).stem |
| 507 | title = "" |
| 508 | workspace = "" |
| 509 | messages: list[dict[str, Any]] = [] |
| 510 | for row in rows: |
| 511 | if row.get("type") in {"session", "session_meta"}: |
| 512 | session_id = str(row.get("id") or row.get("sessionId") or session_id) |
| 513 | title = str(row.get("title") or title) |
| 514 | workspace = str(row.get("cwd") or row.get("workspace") or workspace) |
| 515 | continue |
| 516 | message = row.get("message") if isinstance(row.get("message"), dict) else None |
| 517 | if row.get("type") == "message" and message: |
| 518 | messages.append({**message, "timestamp": row.get("timestamp") or message.get("timestamp")}) |
| 519 | elif row.get("role"): |
| 520 | messages.append(row) |
| 521 | item = _conversation(session_id, title, _events_from_messages(messages), workspace=workspace) |
| 522 | return [item] if item else [] |
| 523 | |
| 524 | |
| 525 | def _sqlite(asset: Upload) -> tuple[sqlite3.Connection, str]: |
| 526 | handle = tempfile.NamedTemporaryFile(suffix=".sqlite", delete=False) |
| 527 | try: |
| 528 | handle.write(asset.data) |
| 529 | handle.close() |
| 530 | connection = sqlite3.connect(f"file:{handle.name}?mode=ro&immutable=1", uri=True) |
| 531 | connection.row_factory = sqlite3.Row |
| 532 | return connection, handle.name |
| 533 | except Exception: |
| 534 | Path(handle.name).unlink(missing_ok=True) |
| 535 | raise |
| 536 | |
| 537 | |
| 538 | def _close_sqlite(connection: sqlite3.Connection, path: str) -> None: |
| 539 | connection.close() |
| 540 | Path(path).unlink(missing_ok=True) |
| 541 | |
| 542 | |
| 543 | def _tables(connection: sqlite3.Connection) -> set[str]: |
| 544 | return {row[0] for row in connection.execute("SELECT name FROM sqlite_master WHERE type='table'")} |
| 545 | |
| 546 | |
| 547 | def _parse_hermes_db(asset: Upload) -> list[Conversation]: |
| 548 | connection, path = _sqlite(asset) |
| 549 | try: |
| 550 | if not {"sessions", "messages"}.issubset(_tables(connection)): |
| 551 | return [] |
| 552 | result: list[Conversation] = [] |
| 553 | for session in connection.execute("SELECT * FROM sessions ORDER BY started_at"): |
| 554 | rows = connection.execute( |
| 555 | "SELECT * FROM messages WHERE session_id = ? ORDER BY timestamp, id", (session["id"],) |
| 556 | ).fetchall() |
| 557 | messages = [dict(row) for row in rows] |
| 558 | item = _conversation( |
| 559 | str(session["id"]), |
| 560 | str(session["title"] or "") if "title" in session.keys() else "", |
| 561 | _events_from_messages(messages), |
| 562 | source=session["source"] if "source" in session.keys() else None, |
| 563 | cwd=session["cwd"] if "cwd" in session.keys() else None, |
| 564 | model=session["model"] if "model" in session.keys() else None, |
| 565 | ) |
| 566 | if item: |
| 567 | result.append(item) |
| 568 | return result |
| 569 | finally: |
| 570 | _close_sqlite(connection, path) |
| 571 | |
| 572 | |
| 573 | def _parse_openclaw_db(asset: Upload) -> list[Conversation]: |
| 574 | connection, path = _sqlite(asset) |
| 575 | try: |
| 576 | if not {"session_windows", "transcript_events"}.issubset(_tables(connection)): |
| 577 | return [] |
| 578 | result: list[Conversation] = [] |
| 579 | for window in connection.execute("SELECT * FROM session_windows ORDER BY created_at"): |
| 580 | messages: list[dict[str, Any]] = [] |
| 581 | for row in connection.execute( |
| 582 | "SELECT event_json, created_at FROM transcript_events WHERE session_id = ? ORDER BY seq", |
| 583 | (window["session_id"],), |
| 584 | ): |
| 585 | try: |
| 586 | event = json.loads(row["event_json"]) |
| 587 | except (TypeError, json.JSONDecodeError): |
| 588 | continue |
| 589 | message = event.get("message") if isinstance(event.get("message"), dict) else None |
| 590 | if event.get("type") == "message" and message: |
| 591 | messages.append({**message, "timestamp": event.get("timestamp") or row["created_at"]}) |
| 592 | elif event.get("role"): |
| 593 | messages.append({**event, "timestamp": event.get("timestamp") or row["created_at"]}) |
| 594 | item = _conversation( |
| 595 | str(window["session_id"]), |
| 596 | str(window["display_name"] or "") if "display_name" in window.keys() else "", |
| 597 | _events_from_messages(messages), |
| 598 | session_key=window["session_key"], |
| 599 | channel=window["channel"] if "channel" in window.keys() else None, |
| 600 | model=window["model"] if "model" in window.keys() else None, |
| 601 | ) |
| 602 | if item: |
| 603 | result.append(item) |
| 604 | return result |
| 605 | finally: |
| 606 | _close_sqlite(connection, path) |
| 607 | |
| 608 | |
| 609 | def _is_sqlite(asset: Upload) -> bool: |
| 610 | return asset.data.startswith(b"SQLite format 3\x00") |
| 611 | |
| 612 | |
| 613 | def _discover_assets(files: list[Upload], bundle: Bundle) -> None: |
| 614 | skill_roots: dict[str, str] = {} |
| 615 | for item in files: |
| 616 | path = PurePosixPath(item.name) |
| 617 | if path.name.lower() == "skill.md": |
| 618 | key = str(path.parent) |
| 619 | skill_roots[key] = re.sub(r"[^a-z0-9_-]+", "_", path.parent.name.lower()).strip("_") or "skill" |
| 620 | for item in files: |
| 621 | path = PurePosixPath(item.name) |
| 622 | suffix = path.suffix.lower() |
| 623 | sensitive = path.name.lower() in SENSITIVE_NAMES or suffix in {".key", ".pem"} |
| 624 | for root, slug in skill_roots.items(): |
| 625 | root_path = PurePosixPath(root) |
| 626 | try: |
| 627 | relative = path.relative_to(root_path) |
| 628 | except ValueError: |
| 629 | continue |
| 630 | if sensitive: |
| 631 | bundle.excluded.append(f"{item.name}: credentials are never imported") |
| 632 | else: |
| 633 | bundle.skills.setdefault(slug, []).append(Asset(str(relative), item.data)) |
| 634 | break |
| 635 | else: |
| 636 | lower_parts = {part.lower() for part in path.parts} |
| 637 | if suffix == ".md" and path.name.lower() in INSTRUCTION_NAMES: |
| 638 | bundle.instructions.append(Asset(item.name, item.data)) |
| 639 | elif suffix == ".md" and ( |
| 640 | path.name.lower() in MEMORY_NAMES or "memory" in lower_parts or "memories" in lower_parts |
| 641 | ): |
| 642 | bundle.memories.append(Asset(item.name, item.data)) |
| 643 | elif sensitive: |
| 644 | bundle.excluded.append(f"{item.name}: settings or credentials are never imported") |
| 645 | |
| 646 | |
| 647 | def _discover_projects(bundle: Bundle) -> None: |
| 648 | by_path: dict[str, Project] = {} |
| 649 | for conversation in bundle.conversations: |
| 650 | project_path = next( |
| 651 | ( |
| 652 | str(conversation.metadata.get(key) or "").strip() |
| 653 | for key in PROJECT_PATH_KEYS |
| 654 | if str(conversation.metadata.get(key) or "").strip() |
| 655 | ), |
| 656 | "", |
| 657 | ) |
| 658 | if not project_path or project_path in {".", "/", "\\"}: |
| 659 | continue |
| 660 | normalized = project_path.replace("\\", "/").rstrip("/") |
| 661 | title = PurePosixPath(normalized).name or "Imported project" |
| 662 | source_id = hashlib.sha256(f"{bundle.source}:{project_path}".encode()).hexdigest()[:12] |
| 663 | project = by_path.setdefault(source_id, Project(source_id, title, project_path)) |
| 664 | project.conversation_ids.append(conversation.source_id) |
| 665 | bundle.projects = list(by_path.values()) |
| 666 | |
| 667 | |
| 668 | def parse_bundle(source: str, uploads: Iterable[Upload]) -> Bundle: |
| 669 | if source not in {"openclaw", "hermes", "opencode", "claude", "codex"}: |
| 670 | raise ValueError("Choose one of the five supported source harnesses") |
| 671 | files = expand_uploads(uploads) |
| 672 | bundle = Bundle(source) |
| 673 | _discover_assets(files, bundle) |
| 674 | seen: set[str] = set() |
| 675 | for asset in files: |
| 676 | lower = asset.name.lower() |
| 677 | conversations: list[Conversation] = [] |
| 678 | try: |
| 679 | if _is_sqlite(asset): |
| 680 | conversations = _parse_hermes_db(asset) if source == "hermes" else _parse_openclaw_db(asset) if source == "openclaw" else [] |
| 681 | elif source == "codex" and lower.endswith(".jsonl"): |
| 682 | conversations = _parse_codex(asset) |
| 683 | elif source == "claude" and lower.endswith(".jsonl") and "/history.jsonl" not in f"/{lower}": |
| 684 | conversations = _parse_claude(asset) |
| 685 | elif source == "opencode" and lower.endswith(".json"): |
| 686 | conversations = _parse_opencode(asset) |
| 687 | elif source == "hermes" and lower.endswith(".jsonl"): |
| 688 | conversations = _parse_hermes_jsonl(asset) |
| 689 | elif source == "openclaw" and lower.endswith(".jsonl"): |
| 690 | conversations = _parse_openclaw_jsonl(asset) |
| 691 | except (ValueError, sqlite3.DatabaseError) as exc: |
| 692 | bundle.warnings.append(f"{asset.name}: {exc}") |
| 693 | continue |
| 694 | for conversation in conversations: |
| 695 | identity = f"{source}:{conversation.source_id}" |
| 696 | if identity not in seen: |
| 697 | seen.add(identity) |
| 698 | bundle.conversations.append(conversation) |
| 699 | |
| 700 | _discover_projects(bundle) |
| 701 | |
| 702 | for conversation in bundle.conversations: |
| 703 | for event in conversation.events: |
| 704 | event.text, count = _redact(event.text) |
| 705 | bundle.redactions += count |
| 706 | event.tool_result, count = _redact(event.tool_result) |
| 707 | bundle.redactions += count |
| 708 | raw_args, count = _redact(json.dumps(event.tool_args, ensure_ascii=False)) |
| 709 | bundle.redactions += count |
| 710 | try: |
| 711 | event.tool_args = json.loads(raw_args) |
| 712 | except json.JSONDecodeError: |
| 713 | event.tool_args = {"redacted": raw_args} |
| 714 | for asset in bundle.knowledge: |
| 715 | text, count = _redact(_text(asset)) |
| 716 | bundle.redactions += count |
| 717 | asset.data = text.encode() |
| 718 | for assets in bundle.skills.values(): |
| 719 | for asset in assets: |
| 720 | if PurePosixPath(asset.path).suffix.lower() in TEXT_SUFFIXES: |
| 721 | text, count = _redact(_text(asset)) |
| 722 | bundle.redactions += count |
| 723 | asset.data = text.encode() |
| 724 | if not bundle.conversations and not bundle.knowledge and not bundle.skills: |
| 725 | bundle.warnings.append("No supported chats, memories, instructions, or skills were found in this upload") |
| 726 | return bundle |
| 727 | |
| 728 | |
| 729 | def _iso(timestamp: float) -> str: |
| 730 | return datetime.fromtimestamp(timestamp, timezone.utc).isoformat().replace("+00:00", "Z") |
| 731 | |
| 732 | |
| 733 | def build_a0_chat(conversation: Conversation, source: str) -> dict[str, Any]: |
| 734 | events = conversation.events |
| 735 | if not events: |
| 736 | raise ValueError("Conversation contains no importable events") |
| 737 | messages: list[dict[str, Any]] = [] |
| 738 | logs: list[dict[str, Any]] = [] |
| 739 | sequence = 0 |
| 740 | last_timestamp = 0.0 |
| 741 | |
| 742 | def message(ai: bool, content: Any, when: float, kind: str) -> dict[str, Any]: |
| 743 | nonlocal sequence |
| 744 | sequence += 1 |
| 745 | item = { |
| 746 | "_cls": "Message", |
| 747 | "id": str(uuid.uuid4()), |
| 748 | "ai": ai, |
| 749 | "content": content, |
| 750 | "metadata": {"imported_from": source, "source_event": kind, "source_timestamp": _iso(when)}, |
| 751 | "sequence": sequence, |
| 752 | "summary": "", |
| 753 | "tokens": 0, |
| 754 | } |
| 755 | messages.append(item) |
| 756 | return item |
| 757 | |
| 758 | def log(kind: str, item: dict[str, Any], heading: str, content: str, kvps: dict[str, Any], when: float) -> None: |
| 759 | nonlocal last_timestamp |
| 760 | last_timestamp = max(when, last_timestamp + 0.000001) |
| 761 | logs.append( |
| 762 | { |
| 763 | "no": len(logs), |
| 764 | "id": item["id"], |
| 765 | "type": kind, |
| 766 | "heading": heading, |
| 767 | "content": content, |
| 768 | "kvps": kvps, |
| 769 | "timestamp": last_timestamp, |
| 770 | "agentno": 0, |
| 771 | } |
| 772 | ) |
| 773 | |
| 774 | for index, event in enumerate(events): |
| 775 | when = event.timestamp or index + 1.0 |
| 776 | if event.kind == "user": |
| 777 | item = message(False, {"user_message": event.text}, when, "user") |
| 778 | log("user", item, "", event.text, {"attachments": []}, when) |
| 779 | elif event.kind == "assistant": |
| 780 | envelope = { |
| 781 | "thoughts": event.thoughts or [f"Imported public response from {source}."], |
| 782 | "headline": _one_line(event.text), |
| 783 | "tool_name": "response", |
| 784 | "tool_args": {"text": event.text}, |
| 785 | } |
| 786 | raw = json.dumps(envelope, ensure_ascii=False) |
| 787 | item = message(True, raw, when, "assistant") |
| 788 | log("agent", item, envelope["headline"], raw, envelope, when) |
| 789 | log("response", item, "icon://chat Responding", event.text, {"finished": True}, when + 0.000001) |
| 790 | elif event.kind == "tool": |
| 791 | name = event.tool_name or "tool" |
| 792 | envelope = { |
| 793 | "thoughts": event.thoughts or [f"Imported historical {source} tool activity."], |
| 794 | "headline": f"Using imported {name} record", |
| 795 | "tool_name": name, |
| 796 | "tool_args": event.tool_args, |
| 797 | } |
| 798 | raw = json.dumps(envelope, ensure_ascii=False) |
| 799 | call = message(True, raw, when, "tool_call") |
| 800 | log("agent", call, envelope["headline"], raw, envelope, when) |
| 801 | result_text = event.tool_result or "No retained tool result was available." |
| 802 | result = message( |
| 803 | False, |
| 804 | {"tool_name": name, "tool_result": result_text, "file": ""}, |
| 805 | when + 0.000001, |
| 806 | "tool_result", |
| 807 | ) |
| 808 | log( |
| 809 | "tool", |
| 810 | result, |
| 811 | f"icon://construction Using tool '{name}'", |
| 812 | result_text, |
| 813 | {**event.tool_args, "_tool_name": name}, |
| 814 | when + 0.000001, |
| 815 | ) |
| 816 | |
| 817 | history = { |
| 818 | "_cls": "History", |
| 819 | "counter": sequence, |
| 820 | "bulks": [], |
| 821 | "topics": [], |
| 822 | "current": {"_cls": "Topic", "summary": "", "messages": messages}, |
| 823 | } |
| 824 | first = min((event.timestamp for event in events if event.timestamp), default=1.0) |
| 825 | last = max((event.timestamp for event in events if event.timestamp), default=first) |
| 826 | chat_id = hashlib.sha256(f"{source}:{conversation.source_id}".encode()).hexdigest()[:8] |
| 827 | return { |
| 828 | "id": chat_id, |
| 829 | "name": conversation.title, |
| 830 | "created_at": _iso(first), |
| 831 | "type": "user", |
| 832 | "last_message": _iso(last), |
| 833 | "agents": [ |
| 834 | { |
| 835 | "number": 0, |
| 836 | "agent_profile": "default", |
| 837 | "data": {}, |
| 838 | "history": json.dumps(history, ensure_ascii=False), |
| 839 | } |
| 840 | ], |
| 841 | "streaming_agent": 0, |
| 842 | "agent_profile": "default", |
| 843 | "log": {"guid": str(uuid.uuid4()), "logs": logs, "progress": "", "progress_no": 0}, |
| 844 | "data": { |
| 845 | "_migrate_agents": { |
| 846 | "format_version": 1, |
| 847 | "source": source, |
| 848 | "source_id": conversation.source_id, |
| 849 | "source_metadata": conversation.metadata, |
| 850 | "hidden_reasoning_included": False, |
| 851 | "historical_tools_are_replayable": False, |
| 852 | } |
| 853 | }, |
| 854 | "output_data": {}, |
| 855 | } |
| 856 | |
| 857 | |
| 858 | def preview(bundle: Bundle) -> dict[str, Any]: |
| 859 | return { |
| 860 | "ok": True, |
| 861 | "source": bundle.source, |
| 862 | "summary": bundle.summary(), |
| 863 | "chats": [ |
| 864 | { |
| 865 | "id": item.source_id, |
| 866 | "title": item.title, |
| 867 | "messages": len(item.events), |
| 868 | "metadata": item.metadata, |
| 869 | } |
| 870 | for item in bundle.conversations[:200] |
| 871 | ], |
| 872 | "projects": [ |
| 873 | { |
| 874 | "id": item.source_id, |
| 875 | "title": item.title, |
| 876 | "path": item.path, |
| 877 | "chats": len(item.conversation_ids), |
| 878 | } |
| 879 | for item in bundle.projects[:200] |
| 880 | ], |
| 881 | "memories": [item.path for item in bundle.memories[:200]], |
| 882 | "instructions": [item.path for item in bundle.instructions[:200]], |
| 883 | "knowledge": [item.path for item in bundle.knowledge[:200]], |
| 884 | "skills": sorted(bundle.skills)[:200], |
| 885 | "excluded": bundle.excluded[:200], |
| 886 | "warnings": bundle.warnings[:200], |
| 887 | } |