Teknium
09e657793e
Composio-style MCP servers return un-paginated 22-47K-char payloads that
sail under the generic 100K per-result spillover threshold, bloating
context and ballooning per-turn reasoning time on long conversations.
Competitors cap harder (OpenCode/pi 50KB, Claude Code 30K, Codex ~10K
tokens). Three changes:
- mcp_* tools spill at a tighter 50K default (BudgetConfig.mcp_result_size,
config-overridable via tool_budget.mcp_result_size_chars; pinned and
per-tool overrides still win; capped by the context-scaled default).
- The persisted-output preview now teaches recovery: page the saved file
with read_file or process with execute_code instead of re-requesting the
same data from the remote API.
- Untrusted/MCP string results are scanned (bounded, first 64KB) for
provider-side elision markers ('...N more items', "has_more": true,
'saved to sandbox', data_preview) and get ONE cache-safe incompleteness
notice appended at result-construction time, before untrusted wrapping —
so the model stops treating provider-elided enumerations as complete.
- Hard 2M-char allocation cap in mcp_tool.py (text, error, and
structuredContent paths) so a pathological multi-MB server payload is
bounded before it propagates, while ordinary large results reach
spillover intact. Distilled from #56060/#56072/#56511 (issue #56059);
supersedes their 50K lossy truncation with spillover-friendly semantics.
Docs: configuration.md spillover-budget section + cli-config.yaml.example.
Co-authored-by: Stoltemberg <215755014+Stoltemberg@users.noreply.github.com>
Co-authored-by: AlexFucuson9 <295703459+AlexFucuson9@users.noreply.github.com>
Co-authored-by: Tranquil-Flow <66773372+Tranquil-Flow@users.noreply.github.com>
70 lines
2.9 KiB
Python
70 lines
2.9 KiB
Python
"""Regression tests for the MCP hard result cap (#56059).
|
|
|
|
MCP tool results had no allocation bound — a buggy or malicious MCP server
|
|
could return multi-megabyte text that floods memory and context before the
|
|
budget/spillover layer sees it. The hard cap truncates only pathological
|
|
payloads (over 2M chars by default) with a 40% head / 60% tail split;
|
|
ordinary large results pass through untouched so the 50K MCP spillover
|
|
threshold (tools/budget_config.py) can preserve them in full on disk.
|
|
|
|
Test shape adapted from PR #56511 (Tranquil-Flow); cap semantics differ —
|
|
see _MCP_HARD_RESULT_CAP_CHARS in tools/mcp_tool.py.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from tools.mcp_tool import _MCP_HARD_RESULT_CAP_CHARS, _truncate_mcp_text_result
|
|
|
|
|
|
class TestTruncateMcpTextResult:
|
|
def test_short_result_unchanged(self):
|
|
text = "x" * 100
|
|
assert _truncate_mcp_text_result(text) == text
|
|
|
|
def test_exact_limit_unchanged(self):
|
|
text = "y" * 100
|
|
assert _truncate_mcp_text_result(text, max_chars=100) == text
|
|
|
|
def test_spillover_sized_result_passes_untouched(self):
|
|
"""A 60K result (over the 50K spillover threshold) is NOT truncated
|
|
here — the budget layer must receive it intact so spillover can
|
|
preserve the full payload on disk."""
|
|
text = "z" * 60_000
|
|
assert _truncate_mcp_text_result(text) == text
|
|
|
|
def test_pathological_result_is_truncated(self):
|
|
text = "z" * (_MCP_HARD_RESULT_CAP_CHARS + 500_000)
|
|
result = _truncate_mcp_text_result(text)
|
|
assert len(result) < len(text)
|
|
assert "TRUNCATED" in result
|
|
|
|
def test_truncation_preserves_head_and_tail(self):
|
|
head_marker = "HEAD_MARKER_START"
|
|
tail_marker = "TAIL_MARKER_END"
|
|
text = head_marker + "x" * 5000 + tail_marker
|
|
result = _truncate_mcp_text_result(text, max_chars=200)
|
|
assert result.startswith(head_marker)
|
|
assert result.endswith(tail_marker)
|
|
|
|
def test_truncation_includes_omitted_count(self):
|
|
text = "a" * 5000
|
|
result = _truncate_mcp_text_result(text, max_chars=100)
|
|
assert "4,900" in result # 5000 - 100 omitted
|
|
assert "5,000" in result # total original length
|
|
|
|
def test_truncation_uses_40_60_head_tail_split(self):
|
|
text = "H" * 40 + "M" * 5000 + "T" * 60
|
|
result = _truncate_mcp_text_result(text, max_chars=100)
|
|
assert result[:40] == "H" * 40
|
|
assert result[-60:] == "T" * 60
|
|
|
|
def test_empty_result_unchanged(self):
|
|
assert _truncate_mcp_text_result("") == ""
|
|
|
|
def test_hard_cap_sits_above_spillover_threshold(self):
|
|
"""The hard cap must stay far above the MCP spillover threshold so
|
|
spillover, not lossy truncation, handles ordinary large results."""
|
|
from tools.budget_config import DEFAULT_MCP_RESULT_SIZE_CHARS
|
|
|
|
assert _MCP_HARD_RESULT_CAP_CHARS > DEFAULT_MCP_RESULT_SIZE_CHARS * 10
|