Files
autocad-mcp-pro/discovery/serialize.py
T
Umut CANandClaude Opus 5 b03a7615b1 release: v1.5.0 — discovery layer, honesty debt closed, Wave A
Tool count 131 -> 154; 474 -> 1242 collected tests.

The release exists for two measured reasons. A 149-tool catalog costs a client
~40,300 tokens before it has asked for anything, and a drafter searching the way
drafters think -- FILLET, BPOLY, QSELECT, WBLOCK -- scored df=0 against the stock
index, because those words were simply not in the data.

Discovery and batching
- DISCOVERY_MODE=search replaces the catalog with search_tools + call_tool,
  ranked over an AutoCAD command and synonym corpus (discovery/aliases.py) on
  top of fastmcp's BM25 index. Measured: 40,305 -> 356 tokens advertised.
- cad_batch runs a step list in one round trip; fields= projects result-heavy
  responses. The `core` profile is gone -- discovery replaced it.

Honesty debt (Tier 0)
- OCS/ECS: `extrusion` appeared nowhere in this codebase, so mirrored and
  rotated geometry reported coordinates that were quietly wrong. backends/ocs.py
  translates at the backend boundary; every tool coordinate is WCS on both
  engines, and a tilted plane refuses rather than answering with a number that
  looks fine.
- drawing_save_as(fmt="dwg") wrote DXF bytes to a .dwg path. drawing_audit never
  repaired. analysis_measure_area shoelaced points the caller typed in.
- ezdxf calls gained a per-call timeout, so one hung call can no longer wedge a
  server whose document lock is a single asyncio.Lock.

Architecture
- The 1854-line backend ABC is 19 contract modules; @capability(key, reason=)
  makes optional methods declarative and a gate fails on undeclared holes.
- 8 layout/viewport tools, CHSPACE, CAD_PROGID escape hatch.
- Wave A: selection filters, boundary tracing, hatch depth, annotation objects,
  measurement -- 13 tools of a planned 39. The REGION and 2D-boolean family was
  cut on evidence: add_region() yields zero ACIS bytes and the greiner_hormann
  substitute loses 28.2% of the area on a square with one semicircular edge.

Verified against a live AutoCAD 2026, which found four COM defects no amount of
reading would have caught -- AcadPViewport has no ViewCenter, and the line that
set it had been swallowed by a bare except since v1.4.

Benchmark v3 (15 tasks) and what it caught
- v2 scored this server 10/10, which carried no information: every task in it
  exercised something the server was built around. Five tasks were added because
  they can fail, and three did while being written -- a hatch could not be
  measured though the refusal listed HATCH as measurable, the boundary tools and
  the measure tool disagreed about the same shape (139.2177 vs 139.2699), and a
  circle stored as two bulged vertices measured 0.0.
- Competitor reports are NOT re-scored. They are pinned v1.4 runs against v2 and
  were never asked these five questions; the chart shows those rows `not run`.
- Correctness A/B: 21/24 -> 24/24, zero regressions.

Performance, stated plainly
The published v1.4 perf report was recorded on CPython 3.14 and the new one on
3.11.15, so the apparent 2-6x speedup is the interpreter, not this release.
Re-measured on one interpreter, v1.5.0 is 28% slower creating entities and 22%
slower on the 10k roundtrip. Fully attributed: EZDXF_CALL_TIMEOUT=0 restores
v1.4.0's exact numbers. That is the deadlock fix being paid for, the knob is
documented, and winning it back is on the 1.6 roadmap.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-06 09:45:20 +03:00

115 lines
3.8 KiB
Python

"""Compact one-line cards for tool-search results.
Passed to a search transform as ``search_result_serializer``. FastMCP's default
(``serialize_tools_for_output_json``) re-emits each hit's full ``list_tools``
entry, JSON Schema and all — for this server that is roughly 1.5 kB per hit,
and a five-hit answer costs more than the crowded catalog the search was meant
to replace.
One line per hit::
entity_fillet (write) | Round a corner between two entities. | req: handle_a, handle_b, radius | acad: FILLET, F
``req:`` lists *required* parameters only. Optional parameters and types are
deliberately absent — the honest cost of this format is that a caller must
either already know the optional arguments or read the error from a failed
call. That trade is worth it at five hits per search; it would not be at one.
Scope note: the compact payload is a real saving but it is **not** what makes
this discovery layer better than stock. Stock BM25 with this same serializer
gets the same size win. The differentiators are the alias corpus (which lets a
query find anything at all) and the risk filter — not bytes.
"""
from __future__ import annotations
from collections.abc import Sequence
from fastmcp.tools.base import Tool
from discovery.aliases import aliases_for
__all__ = ["MAX_SUMMARY_CHARS", "serialize_search_results"]
# Long enough for every first docstring line in the live catalog (measured max
# 96 characters), short enough that a rambling one cannot dominate the payload.
MAX_SUMMARY_CHARS = 110
# Beyond three, AutoCAD command names stop disambiguating and start padding.
MAX_ACAD_COMMANDS = 3
EMPTY_RESULT = "No tools matched the query."
def _risk(tool: Tool) -> str:
"""Same derivation as ``discovery.transform.tool_risk``, inlined.
Imported the other way round would make ``discovery.serialize`` depend on
the transform, and the transform already depends on this module for its
default serializer.
"""
annotations = tool.annotations
if annotations is not None:
if annotations.destructiveHint:
return "destructive"
if annotations.readOnlyHint:
return "read"
return "write"
def _truncate(text: str) -> str:
if len(text) <= MAX_SUMMARY_CHARS:
return text
clipped = text[:MAX_SUMMARY_CHARS].rsplit(" ", 1)[0].rstrip(" ,;:.")
return f"{clipped}..."
def _summary(tool: Tool) -> str:
"""One line describing the tool.
Prefers the hand-written ``@cad_tool`` summary in the MCP ``meta`` channel;
falls back to the first line of the docstring, which for this server is
already written as a standalone sentence.
"""
cad = (tool.meta or {}).get("cad")
if isinstance(cad, dict) and cad.get("summary"):
return _truncate(str(cad["summary"]).strip())
for line in (tool.description or "").strip().splitlines():
if line.strip():
return _truncate(line.strip())
return "(no description)"
def _required(tool: Tool) -> str:
schema = tool.parameters or {}
required = schema.get("required")
if not isinstance(required, list) or not required:
return "(none)"
return ", ".join(str(name) for name in required)
def _acad(tool: Tool) -> str:
record = aliases_for(tool.name)
if record is None or not record.acad:
return ""
return ", ".join(record.acad[:MAX_ACAD_COMMANDS])
def _card(tool: Tool) -> str:
parts = [
f"{tool.name} ({_risk(tool)})",
_summary(tool),
f"req: {_required(tool)}",
]
acad = _acad(tool)
if acad:
parts.append(f"acad: {acad}")
return " | ".join(parts)
def serialize_search_results(tools: Sequence[Tool]) -> str:
"""Render search hits as one compact card per line."""
if not tools:
return EMPTY_RESULT
return "\n".join(_card(tool) for tool in tools)