From de28f0ac17703bcc938f895fc06b50a0613726df Mon Sep 17 00:00:00 2001 From: SweetenedSuzuka <188338189+SweetenedSuzuka@users.noreply.github.com> Date: Fri, 14 Aug 2026 12:31:54 +0800 Subject: [PATCH 1/2] fix: extract text from quoted QQ JSON cards (#9565) The quoted-message parser dropped generic QQ JSON cards (e.g. mini-program shares) in both extraction paths: _extract_text_from_component_chain had no Json component branch, and _parse_onebot_segments only handled com.tencent.multimsg forward cards. Add _extract_text_from_json_card to read the top-level prompt and meta.*.title/desc/url fields (HTML-unescaped and length-capped), delegating multimsg cards to the existing handler so their behavior is unchanged. --- .../core/utils/quoted_message/chain_parser.py | 89 +++++++- tests/test_quoted_message_parser.py | 191 +++++++++++++++++- 2 files changed, 276 insertions(+), 4 deletions(-) diff --git a/astrbot/core/utils/quoted_message/chain_parser.py b/astrbot/core/utils/quoted_message/chain_parser.py index 528ce14b8b..0d5ad23235 100644 --- a/astrbot/core/utils/quoted_message/chain_parser.py +++ b/astrbot/core/utils/quoted_message/chain_parser.py @@ -1,5 +1,6 @@ from __future__ import annotations +import html import json import re from typing import Any, TypedDict @@ -10,6 +11,7 @@ File, Forward, Image, + Json, Node, Nodes, Plain, @@ -142,6 +144,10 @@ def _extract_text_from_component_chain( parts.append(f"[File:{file_name}]") elif isinstance(seg, Forward): parts.append("[Forward Message]") + elif isinstance(seg, Json): + card_text = _extract_text_from_json_card(seg.data) + if card_text: + parts.append(card_text) elif isinstance(seg, Reply): nested = _extract_text_from_reply_component( seg, @@ -261,6 +267,83 @@ def _extract_text_from_multimsg_json(raw_json: str) -> str | None: return "\n".join(texts).strip() or None +# Safety valve: keeps a single card from flooding the LLM context window. +_MAX_JSON_CARD_TEXT_LENGTH = 1000 + + +def _clean_json_card_text(value: Any) -> str | None: + if not isinstance(value, str): + return None + cleaned = html.unescape(value).strip() + return cleaned or None + + +def _extract_text_from_json_card(payload: str | dict[str, Any]) -> str | None: + """Extract readable text from a generic QQ JSON card (mini-program, music, etc.). + + Keeps the existing behavior for ``com.tencent.multimsg`` (forward) cards by + delegating to :func:`_extract_text_from_multimsg_json`; other cards are read + from their top-level ``prompt`` and ``meta`` fields. + """ + parsed = payload + if isinstance(parsed, str): + try: + parsed = json.loads(parsed) + except Exception: + return None + if not isinstance(parsed, dict): + return None + + # Unwrap the double-encoded {"data": ""} form when the outer + # dict carries no app identifier of its own. + nested = parsed.get("data") + if isinstance(nested, str) and parsed.get("app") is None: + try: + nested_parsed = json.loads(nested) + if isinstance(nested_parsed, dict): + parsed = nested_parsed + except Exception: + pass + + if parsed.get("app") == "com.tencent.multimsg": + return _extract_text_from_multimsg_json(json.dumps(parsed)) + + parts: list[str] = [] + seen: set[str] = set() + + def _append(value: str) -> None: + if value and value not in seen: + seen.add(value) + parts.append(value) + + def _append_field(label: str, value: Any) -> None: + cleaned = _clean_json_card_text(value) + if cleaned: + _append(f"{label}: {cleaned}") + + _append(_clean_json_card_text(parsed.get("prompt"))) + + meta = parsed.get("meta") + if isinstance(meta, dict): + for detail in meta.values(): + if not isinstance(detail, dict): + continue + _append_field("Title", detail.get("title")) + _append_field("Description", detail.get("desc")) + url = _clean_json_card_text( + detail.get("qqdocurl") or detail.get("jumpUrl") or detail.get("url") + ) + if url: + _append(f"URL: {url}") + + if not parts: + return None + text = "\n".join(parts).strip() + if len(text) > _MAX_JSON_CARD_TEXT_LENGTH: + text = text[:_MAX_JSON_CARD_TEXT_LENGTH].rstrip() + "..." + return text + + def _parse_onebot_segments( segments: list[Any], *, @@ -335,9 +418,9 @@ def _parse_onebot_segments( raw_json = seg_data.get("data") if isinstance(raw_json, str) and raw_json.strip(): raw_json = raw_json.replace(",", ",") - multimsg_text = _extract_text_from_multimsg_json(raw_json) - if multimsg_text: - text_parts.append(multimsg_text) + card_text = _extract_text_from_json_card(raw_json) + if card_text: + text_parts.append(card_text) return _build_parsed_payload( _join_text_parts(text_parts), diff --git a/tests/test_quoted_message_parser.py b/tests/test_quoted_message_parser.py index 37fd0c54f5..bd4e988c68 100644 --- a/tests/test_quoted_message_parser.py +++ b/tests/test_quoted_message_parser.py @@ -2,7 +2,7 @@ import pytest -from astrbot.core.message.components import Image, Plain, Reply +from astrbot.core.message.components import Image, Json, Plain, Reply from astrbot.core.utils.quoted_message_parser import ( extract_quoted_message_images, extract_quoted_message_text, @@ -514,3 +514,192 @@ async def test_extract_quoted_message_nested_forward_id_is_resolved(): images = await extract_quoted_message_images(event) assert images == [nested_image] + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_in_reply_chain(): + reply = Reply( + id="500", + chain=[ + Json( + data={ + "app": "com.tencent.miniapp_01", + "prompt": "[QQ小程序]无穷小亮正式加入《洛克王国：世界》精灵调查团！", + "meta": { + "detail_1": { + "title": "哔哩哔哩", + "desc": "无穷小亮正式加入《洛克王国：世界》精灵调查团！", + "qqdocurl": "https://b23.tv/2PFeKuq", + } + }, + } + ) + ], + message_str="", + ) + event = SimpleNamespace( + message_obj=SimpleNamespace(message=[reply]), + bot=SimpleNamespace(api=_FailIfCalledAPI()), + get_group_id=lambda: "", + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert "无穷小亮正式加入《洛克王国：世界》精灵调查团！" in text + assert "Title: 哔哩哔哩" in text + assert "URL: https://b23.tv/2PFeKuq" in text + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_via_get_msg_fallback(): + reply = Reply(id="501", chain=None, message_str="") + event = _make_event( + reply, + responses={ + ("get_msg", "501"): { + "data": { + "message": [ + { + "type": "json", + "data": { + "data": ( + '{"app":"com.tencent.miniapp_01",' + '"prompt":"[QQ小程序]卡片标题",' + '"meta":{"detail_1":{"title":"某App",' + '"desc":"卡片描述","jumpUrl":"https://example.com/x"}}}' + ) + }, + } + ] + } + } + }, + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert "卡片标题" in text + assert "Title: 某App" in text + assert "卡片描述" in text + assert "https://example.com/x" in text + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_multimsg_still_extracts_news(): + reply = Reply(id="502", chain=None, message_str="") + event = _make_event( + reply, + responses={ + ("get_msg", "502"): { + "data": { + "message": [ + { + "type": "json", + "data": { + "data": ( + '{"app":"com.tencent.multimsg",' + '"config":{"forward":1},' + '"meta":{"detail":{"news":[' + '{"text":"Alice: hello"},{"text":"Bob: world"}]}}}' + ) + }, + } + ] + } + } + }, + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert "Alice: hello" in text + assert "Bob: world" in text + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_html_entities_unescaped(): + reply = Reply(id="503", chain=None, message_str="") + event = _make_event( + reply, + responses={ + ("get_msg", "503"): { + "data": { + "message": [ + { + "type": "json", + "data": { + "data": ( + '{"app":"com.tencent.miniapp_01",' + '"prompt":"[QQ小程序]实体测试",' + '"meta":{"detail_1":{"title":"T&A"}}}' + ) + }, + } + ] + } + } + }, + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert "[QQ小程序]实体测试" in text + assert "Title: T&A" in text + assert "[" not in text + assert "&" not in text + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_double_encoded_wrapper(): + reply = Reply( + id="505", + chain=[ + Json( + data={ + "data": ( + '{"app":"com.tencent.miniapp_01",' + '"prompt":"包裹卡片","meta":{"detail_1":{"desc":"内层描述"}}}' + ) + } + ) + ], + message_str="", + ) + event = SimpleNamespace( + message_obj=SimpleNamespace(message=[reply]), + bot=SimpleNamespace(api=_FailIfCalledAPI()), + get_group_id=lambda: "", + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert "包裹卡片" in text + assert "内层描述" in text + + +@pytest.mark.asyncio +async def test_extract_quoted_message_text_json_card_length_capped(): + long_desc = "字" * 5000 + reply = Reply( + id="504", + chain=[ + Json( + data={ + "app": "com.tencent.miniapp_01", + "meta": {"detail_1": {"desc": long_desc}}, + } + ) + ], + message_str="", + ) + event = SimpleNamespace( + message_obj=SimpleNamespace(message=[reply]), + bot=SimpleNamespace(api=_FailIfCalledAPI()), + get_group_id=lambda: "", + ) + + text = await extract_quoted_message_text(event) + assert text is not None + assert text.endswith("...") + assert len(text) < 1500 + assert long_desc not in text From 9ccf9455b64f24d74dbf141ab0e01576d69d7213 Mon Sep 17 00:00:00 2001 From: SweetenedSuzuka <188338189+SweetenedSuzuka@users.noreply.github.com> Date: Fri, 14 Aug 2026 14:19:58 +0800 Subject: [PATCH 2/2] refactor: guard prompt before passing to _append in JSON card extraction Only call _append with a truthy prompt to match the str-typed signature (_clean_json_card_text may return None). No behavior change. --- astrbot/core/utils/quoted_message/chain_parser.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/astrbot/core/utils/quoted_message/chain_parser.py b/astrbot/core/utils/quoted_message/chain_parser.py index 0d5ad23235..42558b2e9c 100644 --- a/astrbot/core/utils/quoted_message/chain_parser.py +++ b/astrbot/core/utils/quoted_message/chain_parser.py @@ -321,7 +321,9 @@ def _append_field(label: str, value: Any) -> None: if cleaned: _append(f"{label}: {cleaned}") - _append(_clean_json_card_text(parsed.get("prompt"))) + prompt = _clean_json_card_text(parsed.get("prompt")) + if prompt: + _append(prompt) meta = parsed.get("meta") if isinstance(meta, dict):