emrg: add unit tests for daemon token estimation and history serialization helpers

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []

tests/test_daemon.py

@@ -257,3 +257,219 @@ def test_context_section_manifesto(tmp_path):
  result = server._build_project_context_section(session)
  assert "### MANIFESTO.md" in result
  assert "Keep it simple" in result
+
+
+# ── _count_chars_for_tokens ───────────────────────────────────────
+
+
+def test_count_chars_pure_ascii():
+ """ASCII text: ~4 chars/token."""
+ server = _make_server()
+ # "helloworld" = 10 chars → 10 // 4 = 2 tokens
+ assert server._count_chars_for_tokens("helloworld") == 2
+ # 4 chars → 1 token
+ assert server._count_chars_for_tokens("abcd") == 1
+
+
+def test_count_chars_pure_cjk():
+ """CJK text: ~2 chars/token."""
+ server = _make_server()
+ # "你好世界" = 4 CJK chars → 4 // 2 = 2 tokens
+ assert server._count_chars_for_tokens("你好世界") == 2
+ # "中文" = 2 CJK chars → 1 token
+ assert server._count_chars_for_tokens("中文") == 1
+
+
+def test_count_chars_mixed():
+ """Mixed CJK + ASCII: counted separately then summed."""
+ server = _make_server()
+ # "hello世界" = 5 ASCII + 2 CJK → 5//4 + 2//2 = 1 + 1 = 2
+ assert server._count_chars_for_tokens("hello世界") == 2
+
+
+def test_count_chars_empty():
+ """Empty string → 0 tokens."""
+ server = _make_server()
+ assert server._count_chars_for_tokens("") == 0
+
+
+def test_count_chars_kana():
+ """Hiragana/Katakana counted as CJK."""
+ server = _make_server()
+ # "あいうえお" = 5 Kana → 5 // 2 = 2
+ assert server._count_chars_for_tokens("あいうえお") == 2
+
+
+# ── _estimate_tokens ──────────────────────────────────────────────
+
+
+def test_estimate_tokens_empty():
+ """Empty message list → 0 tokens."""
+ server = _make_server()
+ assert server._estimate_tokens([]) == 0
+
+
+def test_estimate_tokens_single_message():
+ """Single user message with ASCII content."""
+ server = _make_server()
+ msgs = [{"role": "user", "content": "hello world"}]
+ # 3 (overhead) + count_chars("hello world"=11) → 3 + 11//4 = 3 + 2 = 5
+ assert server._estimate_tokens(msgs) == 5
+
+
+def test_estimate_tokens_with_tool_calls():
+ """Message with embedded tool_calls adds their token cost."""
+ server = _make_server()
+ msgs = [{
+ "role": "assistant",
+ "content": "ok",
+ "tool_calls": [{"id": "1", "type": "function",
+ "function": {"name": "bash", "arguments": "ls"}}]
+ }]
+ result = server._estimate_tokens(msgs)
+ # 3 (overhead) + 2//4 (content) + count_chars(json of tool_calls)
+ assert result > 3 # at least overhead
+
+
+def test_estimate_tokens_multiple_messages():
+ """Multiple messages each add overhead."""
+ server = _make_server()
+ msgs = [
+ {"role": "user", "content": "hi"},
+ {"role": "assistant", "content": "hello"},
+ ]
+ result = server._estimate_tokens(msgs)
+ # 2 * 3 (overhead) + count_chars("hi"=2)//4 + count_chars("hello"=5)//4
+ # = 6 + 0 + 1 = 7
+ assert result == 7
+
+
+# ── _estimate_single ──────────────────────────────────────────────
+
+
+def test_estimate_single_message():
+ """Single record with content."""
+ server = _make_server()
+ record = {"role": "user", "content": "hello world"}
+ # 3 (overhead) + count_chars("hello world"=11)//4 = 3 + 2 = 5
+ assert server._estimate_single(record) == 5
+
+
+def test_estimate_single_empty_content():
+ """Record with no content → just overhead."""
+ server = _make_server()
+ record = {"role": "user"}
+ assert server._estimate_single(record) == 3
+
+
+def test_estimate_single_non_string_content():
+ """Non-string content (e.g. list) → just overhead."""
+ server = _make_server()
+ record = {"role": "user", "content": ["part1", "part2"]}
+ assert server._estimate_single(record) == 3
+
+
+# ── _records_to_text ──────────────────────────────────────────────
+
+
+def test_records_to_text_empty():
+ """Empty list → empty string."""
+ server = _make_server()
+ assert server._records_to_text([]) == ""
+
+
+def test_records_to_text_message():
+ """Message record → formatted line."""
+ server = _make_server()
+ records = [{"type": "message", "role": "user", "content": "hello",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[2026-01-15T12:00:00] user: hello" in result
+
+
+def test_records_to_text_tool_call():
+ """Tool call record with arguments."""
+ server = _make_server()
+ records = [{"type": "tool_call", "tool_name": "bash",
+ "arguments": {"command": "ls"},
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_call: bash" in result
+ assert '"command": "ls"' in result
+
+
+def test_records_to_text_tool_result():
+ """Tool result content truncated to 500 chars."""
+ server = _make_server()
+ long_content = "x" * 600
+ records = [{"type": "tool_result", "content": long_content,
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "tool_result:" in result
+ assert len(result.split("tool_result: ")[1]) == 500 # truncated
+
+
+def test_records_to_text_summary():
+ """Summary record renders with marker."""
+ server = _make_server()
+ records = [{"type": "summary", "content": "prior context here",
+ "timestamp": "2026-01-15T12:00:00.000Z"}]
+ result = server._records_to_text(records)
+ assert "[PREVIOUS SUMMARY]" in result
+ assert "prior context here" in result
+
+
+def test_records_to_text_mixed():
+ """Multiple record types in sequence."""
+ server = _make_server()
+ records = [
+ {"type": "message", "role": "user", "content": "hi",
+ "timestamp": "2026-01-15T12:00:00.000Z"},
+ {"type": "message", "role": "assistant", "content": "hey",
+ "timestamp": "2026-01-15T12:00:01.000Z"},
+ ]
+ result = server._records_to_text(records)
+ lines = result.split("\n")
+ assert len(lines) == 2
+ assert "user: hi" in lines[0]
+ assert "assistant: hey" in lines[1]
+
+
+# ── _truncate_record ──────────────────────────────────────────────
+
+
+def test_truncate_record_short():
+ """Short content passes through unchanged."""
+ server = _make_server()
+ record = {"role": "user", "content": "short"}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == "short"
+
+
+def test_truncate_record_long():
+ """Long content gets truncated with notice."""
+ server = _make_server()
+ long_content = "x" * 500
+ record = {"role": "user", "content": long_content}
+ result = server._truncate_record(record, max_tokens=100)
+ # max_chars = 100 * 2 = 200, so truncated to 200 chars + notice
+ assert len(result["content"]) == 200 + len("\n...[truncated for compact]")
+ assert "[truncated" in result["content"]
+
+
+def test_truncate_record_exact_boundary():
+ """Content exactly at max_chars boundary is NOT truncated."""
+ server = _make_server()
+ content = "x" * 200 # max_tokens=100 → max_chars=200
+ record = {"role": "user", "content": content}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["content"] == content # unchanged
+
+
+def test_truncate_record_preserves_other_fields():
+ """Only content is modified; other fields survive."""
+ server = _make_server()
+ record = {"role": "assistant", "content": "x" * 500, "tool_calls": []}
+ result = server._truncate_record(record, max_tokens=100)
+ assert result["role"] == "assistant"
+ assert result["tool_calls"] == []