added support for codex provider. fixed bug in claude code export (lines split when shouldn't be)

tests/test_claude_code.py

@@ -20,7 +20,11 @@ def _write_session(tmp_path, records, project_dir="-home-jesse-myproj", name="ab
     proj = tmp_path / project_dir
     proj.mkdir(parents=True, exist_ok=True)
     f = proj / f"{name}.jsonl"
-    f.write_text("\n".join(json.dumps(r) for r in records), encoding="utf-8")
+    # ensure_ascii=False so raw U+0085/U+2028 reach the parser — see
+    # TestExoticLineBreaks.
+    f.write_text(
+        "\n".join(json.dumps(r, ensure_ascii=False) for r in records), encoding="utf-8"
+    )
     return f
 
 
@@ -501,3 +505,29 @@ class TestMultiRoot:
         monkeypatch.setenv("CLAUDE_CONFIG_DIR", str(tmp_path / "cfg"))
         roots = resolve_roots()
         assert (tmp_path / "cfg" / "projects") in roots
+
+
+class TestExoticLineBreaks:
+    """Same U+0085 hazard as the Codex provider — see its TestExoticLineBreaks."""
+
+    # C0 controls (\x0b, \x0c) are excluded: JSON requires them escaped, so they
+    # never reach the splitter literally. These three do.
+    @pytest.mark.parametrize("sep", ["\x85", "
", "
"])
+    def test_record_with_exotic_break_survives(self, tmp_path, sep):
+        records = [
+            {
+                "type": "user",
+                "cwd": "/home/jesse/myproj",
+                "timestamp": "2026-05-01T10:00:00.000Z",
+                "message": {"role": "user", "content": f"before{sep}after"},
+            },
+        ]
+        _write_session(tmp_path, records)
+        prov = ClaudeCodeProvider(projects_dir=tmp_path)
+        prov.list_conversations()
+        conv = prov.normalize_conversation(prov.get_conversation("abc-123"))
+        texts = [
+            b["text"] for m in conv["messages"] for b in m["blocks"]
+            if b["type"] == BLOCK_TYPE_TEXT
+        ]
+        assert f"before{sep}after" in texts