feat(mcp): render voice messages with duration in chat history (#97)

## Problem

Voice messages in `_format_message_text` previously rendered as a bare
`[语音] (local_id=N, ts=T)` because msg_type=34 fell through to the generic
non-text branch with no schema-aware summarizer. LLMs reading chat history
had no way to judge whether a voice clip was worth calling `decode_voice`
on without first inspecting it.

## Fix

New helper `_format_voice_text(content)` parses the embedded
`<voicemsg voicelength="…">` and renders `[语音 N.Ns]` (duration to one
decimal, milliseconds → seconds). Type=34 dispatches through it, then
appends the existing `_id_suffix()` so the local_id annotation is
preserved end-to-end:

    [语音 3.3s] (local_id=72481, ts=1700000000)

Falls back to `[语音]` (still with `_id_suffix()`) when content is empty,
`<voicemsg>` is absent, XML parse fails, or `voicelength` is missing /
zero / non-numeric.

XML parsing routes through the existing `_parse_xml_root` so the
`_XML_UNSAFE_RE` DOCTYPE/ENTITY filter and 200KB size cap are reused —
no new XXE surface.

## Tests

12 new cases in `tests/test_voice_format.py`: happy path, subsecond,
multi-second, missing / zero / non-numeric voicelength, empty / None
content, missing `<voicemsg>` tag, malformed XML, XXE payload, and two
end-to-end cases through `_format_message_text` (with and without
voicelength) to pin the full rendered output including `_id_suffix()`.

Baseline 183 → 195 passing, 0 regressions.

## Scope

- `mcp_server.py`: adds `_format_voice_text` helper and one branch in
  `_format_message_text` (base_type == 34). No public surface change —
  this only affects formatting of messages that previously rendered as
  the bare `[语音]` fallback.
- `tests/test_voice_format.py`: new file, synthetic fixtures only (no
  real PII).

mcp_server.py

@@ -973,6 +973,21 @@ def _format_voip_message_text(content):
     return f"[通话] {status_map.get(raw_text, raw_text)}"
 
 
+def _format_voice_text(content):
+    if not content or '<voicemsg' not in content:
+        return "[语音]"
+    root = _parse_xml_root(content)
+    if root is None:
+        return "[语音]"
+    voice = root.find('.//voicemsg')
+    if voice is None:
+        return "[语音]"
+    length_ms = _parse_int(voice.get('voicelength'), 0)
+    if length_ms <= 0:
+        return "[语音]"
+    return f"[语音 {length_ms / 1000:.1f}s]"
+
+
 def _format_message_text(local_id, local_type, content, is_group, chat_username, chat_display_name, names, create_time=0):
     sender_from_content, text = _parse_message_content(content, local_type, is_group)
     base_type, _ = _split_msg_type(local_type)
@@ -985,6 +1000,8 @@ def _format_message_text(local_id, local_type, content, is_group, chat_username,
 
     if base_type == 3:
         text = f"[图片] {_id_suffix()}"
+    elif base_type == 34:
+        text = f"{_format_voice_text(text)} {_id_suffix()}"
     elif base_type == 47:
         text = "[表情]"
     elif base_type == 50:

tests/test_voice_format.py

@@ -0,0 +1,82 @@
+"""Tests for `_format_voice_text` (msg_type=34 鉴定).
+
+Voice messages previously rendered as a bare `[语音]` from the generic
+non-text branch. LLMs reading chat history had no way to (a) judge whether a
+clip was worth transcribing, or (b) call `decode_voice` without first round-
+tripping through `get_voice_messages` to look up the `local_id`.
+
+These tests pin the new behaviour: `[语音 Ns]` (duration to 1 decimal) when
+the embedded `<voicemsg voicelength="…">` is parseable, with graceful
+fallback to `[语音]` on missing / zero / malformed length.
+"""
+import unittest
+
+import mcp_server
+
+
+def _voice_xml(length_ms):
+    return (
+        f'<msg><voicemsg endflag="1" length="2048" voicelength="{length_ms}" '
+        'clientmsgid="abc" fromusername="wxid_synth_a" '
+        'cancelflag="0" voiceformat="4" forwardflag="0" /></msg>'
+    )
+
+
+class FormatVoiceTextTests(unittest.TestCase):
+    def test_renders_duration_with_one_decimal(self):
+        self.assertEqual(mcp_server._format_voice_text(_voice_xml(3300)), "[语音 3.3s]")
+
+    def test_subsecond_voice(self):
+        self.assertEqual(mcp_server._format_voice_text(_voice_xml(800)), "[语音 0.8s]")
+
+    def test_long_clip(self):
+        self.assertEqual(mcp_server._format_voice_text(_voice_xml(62000)), "[语音 62.0s]")
+
+    def test_missing_voicelength_falls_back(self):
+        xml = '<msg><voicemsg endflag="1" length="2048" /></msg>'
+        self.assertEqual(mcp_server._format_voice_text(xml), "[语音]")
+
+    def test_zero_voicelength_falls_back(self):
+        self.assertEqual(mcp_server._format_voice_text(_voice_xml(0)), "[语音]")
+
+    def test_non_numeric_voicelength_falls_back(self):
+        xml = '<msg><voicemsg voicelength="abc" /></msg>'
+        self.assertEqual(mcp_server._format_voice_text(xml), "[语音]")
+
+    def test_empty_content(self):
+        self.assertEqual(mcp_server._format_voice_text(""), "[语音]")
+        self.assertEqual(mcp_server._format_voice_text(None), "[语音]")
+
+    def test_missing_voicemsg_tag(self):
+        self.assertEqual(mcp_server._format_voice_text("<msg></msg>"), "[语音]")
+
+    def test_malformed_xml(self):
+        self.assertEqual(mcp_server._format_voice_text("<msg><voicemsg"), "[语音]")
+
+    def test_xxe_payload_rejected(self):
+        xxe = (
+            '<!DOCTYPE foo [<!ENTITY x SYSTEM "file:///etc/passwd">]>'
+            '<msg><voicemsg voicelength="1000" /></msg>'
+        )
+        self.assertEqual(mcp_server._format_voice_text(xxe), "[语音]")
+
+    def test_end_to_end_format_message_text_with_voicelength(self):
+        xml = _voice_xml(3300)
+        _, text = mcp_server._format_message_text(
+            local_id=72481, local_type=34, content=xml, is_group=False,
+            chat_username="wxid_synth_a", chat_display_name="A", names={},
+            create_time=1700000000,
+        )
+        self.assertEqual(text, "[语音 3.3s] (local_id=72481, ts=1700000000)")
+
+    def test_end_to_end_without_voicelength(self):
+        _, text = mcp_server._format_message_text(
+            local_id=99, local_type=34, content="<msg></msg>", is_group=False,
+            chat_username="wxid_synth_a", chat_display_name="A", names={},
+            create_time=0,
+        )
+        self.assertEqual(text, "[语音] (local_id=99)")
+
+
+if __name__ == "__main__":
+    unittest.main()