The 2b9791f fix mirrored only the characters the first review happened to
measure ('=' and ';'), so `custom={'AAA': 'first', 'A:B': 'x', 'ZZZ': 'never'}`
still raised from AddCustomInfo('A:B') on the live engine after subject and
AAA were applied -- the same half-write for 11 other keys. Re-measured on
AutoCAD 2026 by sweeping every character 0x20-0x7E through a scratch
document: exactly thirteen printable characters are 'Invalid key' anywhere
in a key (" * , / : ; < = > ? \ ` |); control characters, internal spaces,
a no-break space and unicode are accepted; leading/trailing whitespace
(space, tab, NBSP) is refused. `_CUSTOM_KEY_FORBIDDEN` is now that set,
the internal-NBSP refusal (wrong: measured accepted) is gone, and the
validator / tool docstrings state the measured rule instead of a closed
enumeration that was false.
A delete (`{key: None}`) is exempt from the AddCustomInfo syntax rules:
RemoveCustomByKey never validates syntax (measured: 'Key not found', never
'Invalid key'), and 2b9791f ran the check before the None branch, so a key
an ezdxf script wrote ('A=B', ' PAD ') showed in drawing_properties_get and
could never be removed. The line-break rule still applies both ways.
Also measured: custom keys are case-insensitive in AutoCAD (AddCustomInfo
over an existing case-variant is 'Duplicate key'; SetCustomByKey and
RemoveCustomByKey match either spelling and keep the stored one). The COM
routing was case-sensitive, so `{'PROJECT': ...}` over 'Project' went to
AddCustomInfo and failed mid-write -- the same class. Both engines now
match existing keys casefolded (update in place under the stored spelling,
never a second tag), and a request naming two case-variants is refused.
Tests: parametrized validator sweep over all thirteen characters, the
accepted set (control chars, NBSP, unicode, the other printables), the
delete exemption headlessly on a DXF carrying such keys and on the fake
SummaryInfo, case-variant refusal, and case-insensitive routing against a
fake that raises 'Duplicate key' / 'Key not found' like AutoCAD. Confirmed
live through the real ComBackend on a scratch document (Close(False)).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Three header-only-write / half-write classes in drawing_properties_set:
1. Measured: ezdxf emits $CUSTOMPROPERTYTAG / $CUSTOMPROPERTY only after
$LASTSAVEDBY, which does not exist before R2004, so on an R12 (AC1009) or
R2000 (AC1015) document the write reported ok, drawing_properties_get
echoed it from memory, and drawing_save_as (version kept) wrote nothing --
the class f3ff703 removed for CANNOSCALE. The headless engine now refuses
with a ValueError naming the DXF version and "save as R2004 or newer",
before custom_vars is touched; the dwgprops capability reason no longer
claims custom_properties_supported unconditionally.
2. A line break in a custom key or value passed validation, ezdxf wrote it
unescaped, and the saved DXF failed to reopen (DXFStructureError) -- the
whole drawing lost, reported as success. Refused by name in the shared
validator.
3. Measured on AutoCAD 2026: AddCustomInfo rejects keys with leading/trailing
whitespace and keys containing '=', ';' or a no-break space as 'Invalid
key' -- mid-write, after the summary and earlier keys were applied, as an
unnamed COM error -- while the headless engine wrote them. The shared
validator now mirrors those rules, so both engines refuse before writing
(internal spaces, tabs, unicode and 300-char keys stay accepted, as
AutoCAD accepts them).
Pinned by save+reopen tests on R12/R2000 (refused, file reads back empty),
R2004 (round-trips), the newline case (refused, the file still opens), and
the fake-SummaryInfo COM test (no ActiveX call reaches SummaryInfo).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Summary fields are refused headlessly with capability dwgprops before any
write; custom properties are $CUSTOMPROPERTYTAG/$CUSTOMPROPERTY pairs and
survive save + reopen. Validation is shared, by name, and never str()-coerces.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Three authored claims in SYSVAR_CATALOG were false against the engines:
- HPANG said "degrees" but ActiveX GetVariable/SetVariable (and AutoLISP
getvar) hold radians - measured on AutoCAD 2026 in a scratch document:
SetVariable(pi/4) reads back 0.7853981633974483 on both paths. The row
now says radians, like POLARANG.
- ANGBASE said "radians" with both engines True, but the DXF header stores
$ANGBASE in degrees (AutoCAD saved the 1.5708-radian seat value as 90.0)
and ezdxf read/wrote it raw, so the two engines disagreed by a factor of
57.3. EzdxfBackend now translates that header variable at its boundary
(_HEADER_DEGREE_SYSVARS): the tool speaks radians on both engines, the
file keeps degrees.
- OSMODE claimed engines.ezdxf True, but $OSMODE is an R12-only header
variable ezdxf never exports for R2000+: a headless write reported
ok: True and read back None after save + reopen. It is registry-saved
in AutoCAD, so it joins _UNSAVED_SYSVARS (refused with
capability: registry_sysvar; the facade's `osmode` key refuses and
snapshots None headlessly, like polar / polar_angle). Headless reads of
every registry-saved variable now report None instead of whatever the
ezdxf template holds in memory.
The gate that should have caught all three, tests/test_sysvar_catalog.py's
default round-trip, wrote 56 no-ops out of 66 and never saved: it now
writes a value that differs from the fresh document's for every writable
ezdxf-True entry, saves, reopens and reads back from the file. Two new
tests pin OSMODE's refusal and ANGBASE's radians-at-the-tool /
degrees-in-the-file split.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
90 authored rows (type, range/enum, default, meaning, saved-in, per-engine
support, friendly key); every ezdxf claim is checked against the real
backend and the registry-saved set is pinned to what the backend refuses.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Measured: an R12 (AC1009) document has no OBJECTS section, so ezdxf exports
neither AcDbVariableDictionary/DICTIONARYVAR nor the seeded SCALE objects, and
drawing_save_as keeps AC1009. `_set_annotation_scale` reported `ok: True,
changed 1:1 -> 1:50` and the file reloaded as 1:1 -- the header-only write
that vanishes on reload, the class this task exists to remove, and a breach
of the backend's own R12 honesty convention (viewport_set_scale / lock).
The write is now refused with a ValueError naming R12 and "save as R2000 or
newer", after parse_scale (a malformed value is still named as such) and
before either dictionary is touched, so a refused call leaves the root
dictionary as it found it and drawing_settings still applies the sibling keys
(limits / grid / ortho survive R12). The refusal is listed with the others in
the contract and server docstrings. Pinned by two tests that save, reopen and
read the R12 file back.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
An ezdxf-authored drawing has no ACAD_SCALELIST entries, and
`_ensure_scale_entry` appended only the requested scale. AutoCAD repopulates
its default list only when the dictionary is *empty*, so the saved file
opened with a one-entry scale list and `SetVariable("CANNOSCALE", "1:1")`
failed with `Error setting system variable` -- the annotation scale could
no longer be changed in AutoCAD, while the tool reported `ok: True` and the
test pinned the degraded state (`len(keys) == 1`).
An empty list is now seeded with AutoCAD's own default metric list before
the requested scale is considered: the 17 entries a fresh AutoCAD 2026
drawing carries (`Documents.Add()` saved as DXF, MEASUREMENT 1), in that
order and under AutoCAD's own keys A0 ... A9, B0 ... B6. New entries take
the next key in that sequence (`_scale_list_key`) instead of being keyed by
their name; presence is still decided by group 300, so a non-empty
AutoCAD-authored list is never re-seeded or duplicated. Verified live: the
headless file now opens in AutoCAD 2026 with the 17 keys, CANNOSCALE 1:50
current, and switching to 1:1 and 1:100 works.
Tests: the fresh-drawing test asserts the seeded names and keys, that
setting 1:50 appends nothing (it is a default), that 1:25 lands under B7
once, and that the list survives save + reopen with 1:1 present; a new test
pins that *reading* the annotation scale creates nothing (the read side
stays pure, AutoCAD fills an empty list itself); the AutoCAD-keyed list test
now expects the A3 key. The `drawing_settings` docstring says what each
engine does and that AutoCAD refuses the live write on a paper-space layout
with no active viewport (measured).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
`system_get_variable` / `system_set_variable` called `GetVariable` /
`SetVariable` on the object `_acad_app()` returns. Those are `AcadDocument`
members; `AcadApplication` has neither, so on a live seat every call died
with `AttributeError: AutoCAD.Application.GetVariable` before anything
reached AutoCAD -- `drawing_settings` (all keys, read and write) was dead on
the COM engine although both engines were promised. Measured on AutoCAD
2026: `hasattr(app, "GetVariable")` is False. Both methods now resolve the
document through `_acad_doc()`, whose "No drawing is open" refusal is what a
seat without a document reports. Verified live through `ComBackend`:
LIMMAX reads (277.0, 202.5); `drawing_settings({"annotation_scale": "1:50",
"polar_angle": 45, "polar": True})` applies and reads back 1:50 / 0.02 /
45.0 / True in model space.
The test fake shipped with the task had GetVariable / SetVariable on the
*Application*, so the gate asserted a call shape the real object does not
have. `_FakeApp` now carries only `ActiveDocument` + `Documents.Count` and
raises `AttributeError` for the two names, exactly like the real
Application; the sysvar store is a `_FakeDoc` that the real `_acad_doc()`
resolves. Two regression tests: the document is the target (call list), and
no open drawing raises the clear RuntimeError. The older
`tests/test_com_sprint3.py` coercion test pinned the same wrong target and
now fakes the document.
`tests/test_com_backend.py::_measure_backend` replaced the module-level
`_acad_doc` without ever restoring it, so a MagicMock document leaked into
every later test of the session -- invisible while nothing downstream
resolved the document itself, fatal once the sysvar path did. An autouse
fixture in that module restores it.
Not touched (same root cause, other owners): `app.GetVariable` /
`app.SetVariable` in `_ensure_linetype_loaded`, `drawing_info`,
`drawing_audit`, `system_run_command` and `system_run_lisp` in
backends/com_backend.py.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
The tuple/list branch put the win32com VARIANT built for SetVariable into the
returned dict. The system_set_variable tool returns that dict verbatim under
output_schema {"type": "object"}; pydantic cannot serialise a VARIANT, FastMCP
drops the structured content and the client sees "outputSchema defined but no
structured output returned" -- after AutoCAD had already applied the write, so
every raw point-variable write (LIMMIN, LIMMAX, GRIDUNIT, SNAPUNIT, INSBASE)
was a retry/undo trap on the live engine while the audit log said TOOL OK.
The VARIANT is now only the SetVariable argument; "value" carries the
coerced Python list. The fake-ActiveX test asserts the reply, and a new
Client(server.mcp) test drives the tool through the output schema so the
serialisation path is covered.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
`_ensure_scale_entry` decided "already present" with `scales.get(name)`, a
lookup by DICTIONARY key. AutoCAD saves ACAD_SCALELIST keyed A0, A1, ... with
the scale name only in the SCALE object's AcDbScale group 300, so on any
AutoCAD-authored drawing (whose default metric list already carries 1:1 ...
1:100, 2:1) the check never matched and every `annotation_scale` write
appended a second SCALE object with the same name.
Presence is now decided by iterating the list and reading each entry's
group-300 name (`_scale_entry_name`, which reads the DXFTagStorage subclass
ezdxf loads SCALE into). A genuinely new entry is still keyed by its name
(the plan's test contract); if that key is somehow taken it falls back to
the next free `A<n>` slot like AutoCAD.
Regression test seeds an A0/A1/A2-keyed list the way AutoCAD writes it and
asserts: an existing 1:50 is reused (no new key, exactly one object named
1:50), a missing 1:25 is appended exactly once and is idempotent, and the
file round-trips with a clean audit.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Headlessly, grid/snap are stored on the active VPORT, CANNOSCALE in
AcDbVariableDictionary (+ an ACAD_SCALELIST entry), and LIMMIN/LIMMAX on
the model-space layout as well as the header — measured: the header alone
is overwritten at save. Registry-saved variables (AUTOSNAP, POLARANG, …)
are refused with capability registry_sysvar instead of vanishing.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Skeleton (contracts, interfaces, file ownership, merge notes) plus the
drafted task steps for Track A hardening (H), styles (S), settings (C),
page setup and templates (P), environment (V) and the final tasks (F);
25 cross-group consistency findings reconciled before commit.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
38 tools in three sections: ISO-25/ANSI dimension, text and leader styles
with drawing_apply_standard; drawing_settings growth, a system-variable
catalogue and document properties; page setup with PDF mediabox read-back,
batch plot and five bundled templates built by the server's own tools;
multi-document on both engines, portable layer states, named views, UCS,
launch, preferences and operator prompts. Track A hardening is wave 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
scripts/smoke_pid_com.py runs the whole track against the live AutoCAD in
a new document. ComBackend.connect() is lazy by design, so the script reads
reachability from system_status()["connected"] rather than from connect()
as the plan sketched, and closes its document without saving unless --keep
is passed. On this machine (AutoCAD 2026, "25.1s (LMS Tech)") it exits 0:
5 nodes, 4 edges, 0 dangling, confidence_min 1.0, 0 critique issues,
2 XDATA chunks — identical to the headless engine on the same spec. The
first attempt exited 2 on a stale win32com gen_py cache (a package folder
holding only __pycache__); an environment fault, no backend change needed.
A/B against v1.5.1 published as ab-v1.5.1-vs-v1.6.0-dev.json: 26/29 ->
29/29, three miss -> pass, zero regressed. check_doc_numbers.py now names
the report and the two row labels in one place (AB_REPORT, AB_*_LABEL),
and README.md / benchmarks/README.md carry the new rows; the v1.5.0 and
v1.4.0 tables stay as history.
README: P&ID row, the showcase and catalogue images, the token table
re-measured (161 / 44,930 by default, 152 / 41,420 with TOOL_PACKS=core,
lean 50 / 14,007), every stale count reconciled (154 -> 166 registered,
149 -> 161 advertised, corpus 158/672 -> 162/731 by the same count, 19 ->
20 groups, 6 -> 8 resources, six-task matrix with pid_roundtrip), a
"Known limitations of the P&ID track" subsection, and the snapshot line at
this machine's 2423 collected tests (CI has not run on this tree; noted in
the changelog with the pre-existing 1369 vs 1363 drift).
CLAUDE.md: section 17, sections 5 and 7 counts, TOOL_PACKS, the two 1.6
contracts and the reader rules. Spec status -> implemented. The plan and
spec markdown were reformatted by ruff (fenced Python only) because
`ruff format --check .` was already red on them at HEAD and the gate
requires the whole tree clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
tasks_v4.py extends v3 with one task the reader can fail on its own: the
example spec is drawn through pid_from_spec's code and read back by the
graph builder, which never sees the spec. The runner, the chart and the
doc-number gate now default to v4; --matrix v3 / v2 reproduce the earlier
sets, and tests/test_benchmark_v3.py labels its full-matrix run "v3"
explicitly now that the default has moved.
The plan's adapter check expected the two unnumbered lines to come back
with no line number; draw_line builds one from the default number_format
(bare sequence when size/service/spec/insulation are empty), so the task
asserts the two authored numbers on the two process lines and sequence-
built XDATA numbers on the other two instead.
The published A/B report still covers 26 checks; Task 21 re-runs it
against v1.5.1, which is what clears the remaining doc-number drift.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
TOOL_PACKS=all (default) | comma list of core,pid. core is everything not
claimed by another pack and can never be dropped; unknown entries are
ignored with a warning. Pack hiding composes with TOOL_PROFILE=lean (the
lean set now carries pid_symbol_insert / pid_line_draw / pid_graph, 50
tools) and ENABLE_3D. system_about reports {enabled, available, ignored};
cad_batch's unknown_tool rejection names the TOOL_PACKS gate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
render() reported symbols = len(all_specs()) without rendering anything, and
the plan's bytes > 10_000 threshold was calibrated for view_screenshot's
16 x 9 in figure, not the 20 x 22 in sheet, where a labels-only render is
85 KB. A builder that draws nothing passed the test.
render() now reports what was drawn: symbols counts the specs whose insert
left an entity with a real extent in the modelspace (blank names the rest;
a spec break is one vertical line, so either axis counts), and ink is the
share of PNG pixels that are not the background colour. The test renders a
labels-only baseline with insert_symbol monkeypatched to draw nothing and
checks that baseline reports zero symbols, is past the byte threshold, and
carries less ink than the real sheet (measured 0.0234 vs 0.0273).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two review findings on the P&ID critique focuses.
The graph's own `nearest` is computed over every port in its grid, including
the `p1..pn` ports it invents on a promoted unknown_block, so a free end near
a sheet frame was hinted to connect to the frame — a call `pid_line_draw`
refuses, since the frame carries no ACADMCP_PID payload. Every dangling row
is now re-hinted over connectable ports only (catalogue / XDATA nodes,
declared ports), so every hint is executable.
`signal_unknown` — the class the reader assigns to a signal-layer line whose
exact class it cannot tell — was skipped by the incompatibility check
because it is not in the drawable class table. Its kind is still signal, so
a signal-layer line ending on a process port is now an error as spec §11.1
requires; only an unclassified line is skipped.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
715c8ef applied the evidence rule in one direction only: an `unknown_block`
(any INSERT a plain LINE happened to touch, filed at 0.3) could not *create*
a pid_dangling_line, but it could still *suppress* one. build_graph attaches
a line end that lands on such a box and invents port `p1` at the line's own
end, so `graph["dangling"]` is empty and `_dangling`, which only read that
list, called the line connected. A process line from P-101's discharge to
the sheet frame (a CTO title block is a block reference) passed
drawing_critique and the drawing_finalize gate; the same line stopping 2 mm
short of the frame was an error. Spec §11.1 says the end resolves to nothing
in both cases — a frame, a logo or a note symbol is nothing the drawing
vouches for.
`_dangling` now reads two sources with one rule: the graph's own dangling
rows (filtered to evidenced edges, as before) plus every end of an evidenced
edge whose ref points at an `unknown_block`, at that inferred port's x/y.
The `nearest` hint is computed the way the graph computes it for a free end
(nearest port within 10·tolerance, same space), over recognised nodes' ports
only, so the invented `p1` can never be the suggested fix; `detail["touches"]`
names the block the end sits on. The other direction is unchanged: a note
leader whose only anchor is an unknown_block is not evidenced and stays out.
Tests: the reviewer's repro (landing on the frame → error; graph attaches it,
dangling=[]), the stop-short control, the finalize gate raising on it, and
the nearest-port hint naming a catalogue valve's port 1 mm away rather than
the frame's inferred port.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two ways the six focuses turned a reader guess into a gate-failing error:
- The "no P&ID content" guard was `not graph["nodes"]`, but build_graph
(include_foreign=True) files ANY INSERT as an `unknown_block` at 0.3 the
moment a plain LINE's end lands on its box. A bolt with a note leader on a
purely mechanical sheet became a P&ID node, the leader a foreign edge, and
its free end a `pid_dangling_line` error — so drawing_finalize (focus=None)
failed a sheet the seven mechanical focuses pass. Now `has_pid_content`
needs a recognised node (catalog / xdata / heuristic); an `unknown_block`
is never tagged or scored, and `_dangling` only reports an edge that is a
P&ID line by evidence (a class from XDATA or a P&ID layer, or an end on a
recognised node) — on a real P&ID the same leader is not dangling either.
- `_incompatible` read `port["kind"]` off inferred ports, which the graph
hard-codes to "process" for every foreign block. A signal line correctly
drawn to a CTO control valve's actuator was an error whose hint said to
redraw the control signal as a process line. Inferred ports are skipped;
only a declared (catalogue / XDATA) port kind is judged.
Tests: both repros as regressions (unit + the server drawing_finalize gate),
a foreign GATE_VALVE whose dangling line is still reported (the heuristic
path spec §9.3 pins is not silenced), a note leader on a real P&ID, and the
spec's own negative — all six return [] on the README gear sheet.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
instrument_index only resolved {"node"} ends of a signal line, so a
transmitter tapping the middle of a pipe (a {"junction"} end) reported
connected_to="" and an instrument on untagged equipment fell back to the
signal line's own number — which Task 13 always stamps, so the handle
fallback was unreachable and every untagged mounting was labelled with
its wire. Spec 9.5's "line numbers" clause is exactly the junction case.
_far_ends now names the far node's tag (its handle when untagged) and,
at a junction, the line number of every line the junction joins (that
line's handle when unnumbered). Two tests exercise both mountings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Closes the four verified BLOCKING findings on pid_graph.
1. scope="all" attached across layouts: the port grid, bbox candidates,
vertex grid, junctions and text lookup were pooled from every space.
The grid is now keyed by (space, cell) and every lookup filters by the
edge's space; junctions carry `space`.
2. A mirrored INSERT (extrusion -Z, as MIRROR3D / foreign DXFs store it)
got its catalogue ports on the unmirrored image at 1.0. Both engines
report `mirrored` on INSERT properties (ezdxf: extrusion z < 0; COM:
Normal z < 0; default False); transform_port takes `mirrored` and
applies x -> -x about the WCS insertion point after rotation/scale,
direction (180 - d) mod 360 — cross-checked against Insert.matrix44.
resolve_endpoint passes it through too, so draw_line starts on the
mirror image as well.
3. The text-tag fallback searched a full circle around the box top, so a
line label or note below the symbol beat the tag above it. Candidates
must now sit at or above top - height/2 (spec §9.2 step 6), nearest
wins, and line-number texts are never equipment tags.
4. A non-uniform INSERT scale re-ran the transform with the X factor and
set `inferred` (reserved for foreign-block ports). transform_port
gains `strict=False` for the reader: point ports scale per axis
exactly; a radial port keeps min(sx, sy)·r and the node drops to 0.6
with notes ["non-uniform scale"]. Node dicts gain `y_scale`,
`mirrored` and `notes` (default []).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two verified blocking findings on pid_graph.
A foreign block's inferred ports were any line end within tolerance of
its drawn box *or strictly inside it*. A sheet border or title block is
an INSERT too, and its box encloses the whole drawing, so every end that
missed a port attached to the border instead of becoming the junction
or dangling end the tool exists to report: on a border sheet the plan's
own small P&ID read back with no junction, no dangling end and its open
signal line "connected" to A3_BORDER. The test is the boundary again, as
spec 9.2 step 3 says (within tolerance of the block's bounding-box
boundary); overshoot into a body is the tolerance's business. The
nearest-insertion tie-break stays for two boundaries that share an end.
The COM engine's `geometry_bbox` carried each definition member's own
axis-aligned box through the INSERT corner by corner, which is exact at
a right-angle rotation and too large everywhere else: a 45-degree-turned
circle of radius 5 read as 7.07, a diagonal line stood upright read 7.07
wide, and the two engines disagreed on the same drawing with no marker.
Off a right angle the member's real geometry is now measured - a LINE,
POINT and straight polyline by their vertices, a CIRCLE by the
closed-form extents of the scaled-and-turned ellipse, an ARC and a
bulged polyline by their ends and the quadrant points the sweep passes
(uniform scale, reflection reversing the sense) - and cross-checked
against ezdxf's `virtual_entities` extents at 37 and 45 degrees, scaled
and mirrored. A member that cannot be measured (text, hatch, a nested
reference, an elliptical arc) is still carried by its corners and the
box then says `approximate: true`; the graph takes the tighter wall of
that box and the attribute-inclusive `bounding_box`, both of which hold
the drawn geometry.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two verified blocking findings on pid_graph.
`_point_segment_distance` put the centre of an arc larger than a
semicircle (|bulge| > 1) on the wrong side of the chord: `copysign`
discarded the sign of `radius - sagitta`, which is exactly what carries
the centre across the chord past 180 degrees. A bulge-2 arc was tested
as its own reflection, so a line ending on its true apex dangled and one
ending on empty paper made a junction. The offset now keeps that sign
and only takes the bulge's direction from the bulge.
Foreign-block port inference tested line ends against the INSERT's
`bounding_box`, which on both engines takes the ATTRIB text with it. A
TAG lettered above a valve pushed the box past the body, so a line
ending on the body's real edge on that side was *inside* the box and
reported dangling. Both engines now report `geometry_bbox` for an
INSERT - the drawn geometry only, ATTRIB/ATTDEF excluded, in WCS, omitted
when the block draws nothing (ezdxf via `virtual_entities`, COM by
carrying each definition member's box through the INSERT's origin,
scale and rotation). The graph prefers it over `bounding_box`, counts
an end as touching when it is within tolerance of the box or inside it,
and when several foreign boxes hold the end picks the nearest insertion.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
The per-vertex `bulges` list was handed out in the entity's own frame next
to `points` that had been translated to WCS. A mirrored LWPOLYLINE
(extrusion / Normal -Z, what `entity_mirror` produces) is a reflection,
which reverses every arc's sense, so the reader tested endpoints against
the arc's mirror image: a line ending on the true apex dangled, a line
ending on empty paper made a junction, and `length` (frame-invariant)
said nothing. Both engines had it — ezdxf emitted `get_points("xyb")`
raw, COM returned `GetBulge` raw after reflecting `Coordinates`.
`ocs.wcs_bulge(extrusion, bulge)` now negates the sign for a left-handed
frame (the same rule `wcs_arc_angles` applies to an ARC's endpoints) and
both `_entity_info` paths route bulges through it whenever they translate
the points. Fenced by an end-to-end mirrored-pipe graph test, the COM
`_entity_info` fake with Normal -Z, and a boundary-level OCS test.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two reader defects, both about trusting a derived number over the drawing.
Stale junctions: endpoints were resolved before a foreign line's membership
was decided, so a P&ID line ending on a foreign line that was then discarded
(it touched no node) kept `{"junction": "J1"}` while `junctions` was empty
and `stats.dangling` was one short. Node attachment now runs first for every
edge, the foreign lines that attach are admitted, and only then are junctions
and dangling ends resolved over the edges actually in the graph.
Chord length: `length` was a chord walk over the vertices, 12.5% short on a
semicircular crossing jump with no field saying so, and junction detection
tested the chord too, so a line ending on the arc dangled. Both engines now
report per-vertex `bulges` next to `points` (COM asks GetBulge only when
`Length` exceeds the chord walk, since an arc is always longer than its
chord); the reader takes the engine's bulge-aware `length` and tests
endpoints against the arc. The chord sum survives only for an engine that
reports no length, and is then labelled `length_approximate`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two review findings on pid_line_draw, both root-caused:
1. existing_pid_lines() filtered entity_list(type_filter="LWPOLYLINE") and
tested info.type == "LWPOLYLINE". A live LWPOLYLINE's ActiveX ObjectName
is AcDbPolyline (the repo's own 2026-08-06 measurement), which the COM
engine reports as POLYLINE, so on live AutoCAD every auto-built number was
seq 1, crossings always 0 and port_reuse always []. The reader now lists
without a type filter and accepts {LWPOLYLINE, POLYLINE, 2DPOLYLINE} plus
LINE — the same set the Task 14 graph reader keeps. The COM _entity_info
polyline branch matched only the DXF-flavoured "AcDbLWPolyline", so a live
LWPOLYLINE never got `points` either; it now accepts both spellings (with
the OCS normalisation), which is what makes the crossing count real there.
2. resolve_endpoint() applied only x_scale, so a mirrored INSERT
(entity_mirror writes y_scale = -1) resolved its port to a point off the
symbol and the line was drawn through the pump body. transform_port takes
an optional y_scale (default: uniform, signature unchanged) and applies the
full reflect-then-rotate transform to point, direction and radius; a
stretched symbol (|x_scale| != |y_scale|) has no circle for a radial port
and is refused with a clear ValueError instead of a silent wrong number.
Tests: mirror about both axes resolves onto the real geometry, a stretched
copy is refused, and a fake-COM AcDbPolyline probe sees the existing line
with its seq, vertices and from-ref.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Nozzle stubs started at the bounding-box edge, so on a dished head, a
hopper's sloped wall or a cone roof they floated off the vessel (default
horizontal_vessel N3/N4 by 0.8 mm, reactor N3/N4 by 0.06 mm, unbounded for
user fractions). The port now stays STUB outside the bbox and the stub is
drawn from there in to where its axis-aligned line meets the outline, via a
new geometry.scan_hits helper; jacketed reactors' stubs pass through the
jacket to the wall.
validate_vessel_params refuses height <= width for vertical capsules
(vertical_vessel, column, reactor), width <= height for horizontal_vessel
and height <= width/6 for a cone-roof tank: those folded the heads through
each other and produced geometry outside the declared bbox. Unknown keys
inside a nozzle dict are refused naming nozzles[i] and the key, like the
top-level parameters already were.
Tests: a fuzz over every accepted vessel shape x 28 nozzle positions checks
the declared bbox and that each stub's inner end lies on the outline (exact
point-to-arc/segment distance); the default-symbol repros and the refusals
are pinned. 137 of the new cases fail against the previous module.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Task 9 of the v1.6 P&ID plan: engineering/pid/symbols_equipment.py fills
the stub with three ISO 10628-2 families registered on import — rotating
(centrifugal_pump, pd_pump, compressor, fan, agitator, ejector), heat
(shell_tube_hx, plate_hx, air_cooler, fired_heater) and misc inline
fittings (filter, strainer_y, reducer with shape=concentric|eccentric,
flange_pair, spectacle_blind, orifice_plate with a signal port,
drain_vent). Every port sits on the declared bbox edge and every symbol
carries the TAG/DESC attribute pair; the catalogue harness now covers
103 specs.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two authoring slips in the valve x actuator composition:
- The diaphragm actuator's stem stopped at y=4 while the dome's flat base
sat at y=6, so the canonical control-valve symbol (PID_VALVE_*_DIAPHRAGM,
9 catalogue entries) shipped with its dome floating 2 mm above the stem.
- Actuators were placed at a fixed y=4 regardless of the body, so on the
needle body (top 5.6) the piston/motor/solenoid glyphs were drawn over
the knob and the hand bar across the needle's stem; the stem line(0,0,0,4)
also duplicated the needle's own axial line (and the butterfly disc).
Placement is now body-relative: `_body` reports `axis_top` (where its own
geometry on x=0 ends) and the stem runs from there to the glyph base at
`body_top + STEM_CLEARANCE`; the diaphragm dome's base is the stem top
(base + DIAPHRAGM_YOKE), and the FAIL attribute follows the glyph base.
The reference 8x4 bodies keep their published numbers (globe+diaphragm
signal at y=9.0, bbox top 9.0).
Tests: per-combination check that the stem starts at the body's derived
axis top, the glyph's lowest edge equals the stem top, the glyph clears
the body, and the signal port is the topmost edge; plus a catalogue-wide
no-duplicate-primitives gate. 33 combinations fail on the old geometry.
The plan's Step 3 snippet is synced with the fixed module.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
build_instrument drew a circle for every type, so `computer` shipped as
circle + hexagon and `plc` as circle + square + diamond (10 of the 20 bubble
blocks wrong, and the circle poked 0.67 mm outside the hexagon's flat top).
Spec 4.4: discrete = circle; dcs = circle inside a square; computer =
hexagon; plc = diamond inside a square. The outline is now chosen per type.
Location lines were always drawn from -r to +r, overshooting the hexagon's
slanted sides by 0.46 mm at y=+/-0.8 and the circle's chord by 0.06 mm.
They now end on the outline (`_bubble_half_width`): chord for the circle,
full width for the square, `r - |y| / tan 60` for the hexagon.
Tests now assert the exact primitive list per type and that every location
line stays on/inside its outline (both fail against the previous code).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
encode_values() let strings and nan/inf through float() into 1010 point
coordinates and never finiteness-checked bare 1040 reals, so
entity_set_xdata(handle, app, [["nan", 1]]) or [float("inf")] wrote a
literal "nan"/"inf" into the DXF (and every undo snapshot) and read back
as null. Numbers now go through the same rule block_specs._scalar applies:
int/float only, bool refused, math.isfinite, error naming values[i][j].
The advertised 16 KB limit was per call per app; AutoCAD's is per object
across every application (xdroom/xdsize), so three 15 KB apps on one
entity passed headless and would fail SetXData live. check_entity_budget()
now sums the entity's existing XDATA under every other app (ACAD's DSTYLE
overrides included) plus each 1001 marker with AutoCAD's resbuf sizes,
and both engines run it before registering the APPID or writing; removal
is never refused. The result's "bytes" is that same accounting.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
encode_values() accepted strings containing "\n" or "\r". ezdxf writes
1000-tag values verbatim, so the separator split the tag pair: the saved
DXF failed to reopen (DXFStructureError) and, because every undo and
transaction snapshot is also a DXF save, transaction_rollback broke after
a single such write. Both engines route through encode_values, so the
refusal (ValueError naming values[i]) now happens there, before either
backend touches the document. Only LF and CR split a DXF tag; tabs and
the other Unicode separators round-trip and stay allowed.
Tests: pure encoder refusal for \n, \r, \r\n; other control characters
still typed; ezdxf refuses-then-rollback-and-reopen round trip; COM fake
sees no SetXData / RegisteredApplications call.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
`text`, ATTDEF `prompt` and `default` were only isinstance(str)-checked, so
'A\nB' and '10^' passed validation and ezdxf's DXFAttr fixer
(`fix_one_line_text`) rewrote them on `add_text` / `add_attdef` at DEBUG
log level — the declared ATTRIB value and the one the P&ID reader gets back
differed silently, and COM (raw string) diverged from ezdxf. Same hole the
tag check closed in 4158575, left open on its three siblings.
A `_one_line(value, where)` helper now refuses '\n', '\r', any other C0
control, and a trailing '^' (ezdxf's own `is_valid_one_line_text` rule),
naming `entities[i]: 'text'` / `attdefs[i]: 'prompt'|'default'`. Tests cover
the refusals, the strings ezdxf keeps verbatim ('10^2', 'Ø20 H7', '%%C20'),
and an ezdxf round-trip proving accepted values are stored unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Three ways `engineering/pid/lines.py` could emit a silent wrong result or
an opaque crash instead of an honest refusal:
- `format_line_number` only recognised a token that was exactly `{field}`;
`P-{seq:04d}`, `{size}"-{service}` or `{bogus}x` were copied verbatim,
braces and all, into the line number. Tokens are now parsed with
`string.Formatter`: one named known field per token, a format spec is
honoured (`{seq:04d}` -> `0007`), literal decoration around the field
survives with it and drops out with it, and conversions, positional
fields, multi-field tokens, unbalanced braces and specs the value cannot
take are all refused by token name.
- `route` never checked finiteness: a NaN `end` collapsed the U candidate
to a valid two-point "straight" path ending at the stub point. Every
coordinate of start/end/waypoints and the stub are now checked and the
refusal names the offending `end.x` / `waypoints[1].y` / stub.
- A non-axis port direction (45.0, 90.5, nan) raised `KeyError` from the
AXIS lookup. `axis_direction()` now normalises with a 1e-6 deg tolerance
for engine drift and otherwise raises ValueError naming `start_dir` /
`end_dir` and how to resolve it (rotate to a multiple of 90 or pass
waypoints).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Every grammar error is prefixed with the tag's own letters, so the old
single-letter fragments (FCI/"I", PSVX/"X", ...) matched whatever the
grammar said: a mutation that dropped the letter and position from every
error left 9 of 11 invalid rows green. The fixture now quotes the error
text the spec demands ("unexpected letter I at position 3", "... after
valve V", "trailing O (open/closed) ...", "needs at least one function
letter after F"), and the test pins it to the single first error.
ABCDEF-106 was mispinned as a grammar rejection of "D": six letters
exceed the regex cap, so it never reached the grammar. It is now pinned
as a format rejection, and ABCDE-106 carries the "letter D at position 4"
case.
A fixture lint rejects any fragment that is a substring of its own tag
and requires grammar rows to name a letter, position or trailing rule,
so a future row cannot reintroduce the hole.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two ways a malformed block_define request slipped past the pre-write
validator and reached the engine:
* `ATTDEF_TAG_RE.match` with a `$` anchor accepts 'TAG\n' because `$`
matches before a trailing newline. ezdxf strips the newline on write, so
'TAG\n' + 'TAG' landed as two ATTDEFs with one tag, defeating the
duplicate-tag refusal, and the tag-keyed attribute readers then dropped
one silently. Now `fullmatch`.
* `_layer` only required a non-empty string, so a DXF-illegal name such as
'BAD<LAYER' passed and ezdxf raised DXFValueError mid-loop, after the
block existed and earlier primitives were written (with overwrite=True,
after the old definition was already wiped). The gate now refuses on the
same forbidden-character rule `sanitize_symbol_name` uses, exposed as
`security.illegal_symbol_name_chars`, with a TypeError naming
`entities[i]` and 'layer'.
Tests cover both repros and fail against the previous code.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
`add_blockref` never checks the name and `add_auto_attribs` skips autofill
when `Insert.block()` resolves to None, so a typo'd block name returned a
handle whose attribute values were silently dropped and whose dangling
INSERT `doc.audit()` later deletes. Check `name in self._doc.blocks` and
raise `ValueError` on both the attribute and no-attribute paths, matching
the COM engine's `InsertBlock` refusal for an unknown name.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Plan covers the prerequisite block_insert attribute fix (ezdxf wrapped the
INSERT in an anonymous *U block and lost its ATTRIBs), the two generic
contracts, the 43-symbol catalogue, routing, the ISA-5.1 grammar, the
graph reader, deliverables, six critique focuses, pid_from_spec, TOOL_PACKS,
resources, benchmark v4, and the live COM smoke.
Spec reconciled with the plan: 43 base symbols, FUNC/LOOP at +-2.0,
solid = DXF SOLID entity, shape/direction options, example spec uses an
actuated FCV-101 (a hand valve has no signal port), refine hints in
detail["hint"], preflight LAYER_SET_INTENT_MISMATCH, radial port matching
on the circle.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Research snapshot of the AutoCAD MCP field (20 READMEs, 2026-09-14), the
seven 1.6 tracks with measurable acceptance criteria, their order
(A -> E -> B -> G -> F -> H -> D), and the tool-pack decision.
Track A: symbol catalogue authored in code (ISO 10628-2 + ISA-5.1), two
generic contracts (block_define, entity_get/set_xdata), nine pid_* tools,
routing, the ISA-5.1 tag grammar, the pid_graph reader, six critique
focuses, TOOL_PACKS, and the test/evidence plan.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
420 -> 288 lines. The source-review leaderboard, the live-run score table and
the task-matrix heatmap all ranked this project against other people's
servers; they are gone. What is left is what this repo can prove about
itself: the correctness A/B against its own previous tags, the five benchmark
tasks that can fail, the performance lane with its regression stated, and the
limits section.
Client wiring, the environment table and the showcase tool-call sequence move
into <details> so the page scans in one screen without losing them.
Star history now charts U-Pool alongside this repo.
Every gated number still matches the tree (16/16) and the release-consistency
suite is green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every one of these was reported ok/clean to the caller while being wrong.
Silently wrong numbers
- selection_filter(min_area) read `entity.dxf.get("area", 0.0)` for SOLID,
TRACE and HATCH, none of which carry that attribute, so every hatch
measured 0.0. A 1600 mm^2 hatch was excluded by min_area=1.0 while
`filtered_by` asserted the filter had run, and COM (which reads ActiveX
.Area) returned a different selection for the same call. It now answers
from the measurement `entity_measure` uses, so the two cannot disagree.
- Every headless dimension rendered through ezdxf's bare `Standard` style:
1.0 mm text with a comma decimal marker, against ISO 3098's 2.5 mm minimum
and ISO 129's point -- while the document header itself said $DIMTXT 2.5.
`ezdxf.new(setup=True)` is not the fix; measured, it renders a 100 mm
dimension as "10000". The attributes are now stored explicitly.
- Saving writes `dimrnd = 0.0` onto the Standard style, and ezdxf applies
`xround(value, dimrnd)` whenever the attribute is present at all --
`xround(33.333, 0.0)` is 33. So on any reopened file, including one this
server saved, every dimension was rounded to a whole unit before DIMDEC
formatted it: a 12.75 mm feature dimensioned 13. AutoCAD means DIMRND 0 as
*no* rounding; the sentinel is now discarded at every load site, and a real
0.5 mm rounding is left alone.
- The DIM* header fold gave out-of-range values teeth they did not have when
they died in the header: dim_decimals=-1 returned ok and then killed the
next dimension with `ValueError: Format specifier missing precision` from
inside ezdxf, and 40 printed forty digits of float noise onto the sheet.
Ranges are now AutoCAD's own (DIMDEC 0-8, DIMZIN 0-15, lengths positive).
Security
- security.py recorded the policy as "sanitize_command and sanitize_lisp each
have exactly ONE caller ... the policy is scoped to the CHANNEL". The
premise was false. `_ensure_linetype_loaded` and `ComBackend.linetype_load`
build `_-LINETYPE _LOAD {name} {file}` and hand it to SendCommand, which
treats a newline and ';' as segment separators -- so a `linetype=` argument
carrying either injected further AutoCAD commands, and the 36-verb denylist
never saw that channel. Guarded with the DXF symbol-name rule (the
injection characters fall out of it rather than being enumerated) plus
validate_path for the file argument, which was also the one path-taking
tool reaching the filesystem unchecked. The policy comment is corrected.
Gates that could not see
- drawing_critique returned zero issues on a quarantined document: all seven
checks wrapped their backend call in a bare `except Exception`. CLAUDE.md
rule 7 makes that call the gate an agent is told to trust. Only the
quarantine now escapes; an ordinary broken check still degrades.
`_check_dimension_legibility` also read `_doc` directly from the event loop
thread, missing the gate and performing exactly the unsynchronised access
the quarantine exists to prevent. `construction_clear` answered
"deleted: 0" -- there was nothing to clear -- for the same reason.
- The iso128 focus checked layer lineweights only, so a sheet lettered at
1.0 mm critiqued clean. It now reads the effective dimension height and
separator, header-then-style, in the order the renderer resolves.
- check_doc_numbers.py derived five values it never checked, and those were
the ones that had drifted. The collected-test count (README said 1274, the
tree had 1369) and the A/B pass-rate and Fixed columns are now gated.
- benchmarks/README.md's v1.4.0 table was impossible on its face -- 21
passing plus 5 fixed is 26, not 24 -- and contradicted the report it
linked, because `ab-v1.4.0-vs-v1.5.0.json` was regenerated from a tree that
already carried the dimension fixes and so never measured the v1.5.0 tag.
Replaced with a real run: ab-v1.4.0-vs-v1.5.1.json, 21/26 -> 26/26.
Release gate
- release.yml ran one ubuntu/3.11 lane, so a Windows-only defect could reach
PyPI with no Windows job having executed. That is not hypothetical: v1.5.0
shipped `except pywintypes.com_error` raising NameError wherever pywin32 is
absent. It now runs the same three lanes as ci.yml, Windows included.
1368 passed, 1 xfailed. ruff clean. Tool inventory and all 16 gated document
numbers match the tree.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The triage over the 43 unaddressed verdicts returned 4 code and 35 doc items.
The code ones shipped in e24b46d; these are the documents.
Twenty-one applied mechanically after a validation pass confirmed each locator
still matched exactly once and each replacement was supported by the tree. Ten
were held back because a re-verify pass found them mutually exclusive -- three
patches rewriting the same benchmarks/README block, three rewriting the same
OCS paragraph -- and applying them in sequence would have corrupted the files.
Those were resolved by hand into one coherent edit each. One was rejected.
The corrections worth naming:
- **The A/B table contradicted itself in three files.** benchmarks/README.md
and CLAUDE.md said 24 checks / 21-24 / 87.5% / "fixed 3" while the suite has
26 and the published JSON says 21/26, 80.8%, five fixed. All three now agree,
and the five are split into the three new-capability `miss -> pass` and the
two repaired-defect `fail -> pass`, which is the distinction the lane exists
to make.
- **"Every number below is produced by a script"** was false twice over:
section 5's nine scores are literals a human assigned, and section 4 now
quotes hand-collected timings. The claim is narrowed to the charts, which
really are all rendered from committed data, and both exceptions are named.
- **The OCS paragraph over-quantified.** TEXT rotation was given as the single
exception; z components and HATCH boundaries are also entity-frame, and the
live backend records `2d_polyline_coordinates_not_verified_against_live_autocad`
in the same reason string. It now points at `ocs_normalized` as the live list.
- **Transactions were shown as bare parity.** The server itself grades them
differently -- `FeatureCapability(True, "undo_mark")` on COM against
`(True, "snapshot")` headless -- and COM rollback sends `_UNDO B` without
confirming it landed. The row says so.
- **"Raw commands: Opt-in"** named a switch that does not exist. Both tools are
registered unconditionally; `DANGEROUS_COMMANDS_ENABLED` only disables the
sanitizer.
- **"Wave A shipped 13 of a planned 39"** hardened an approximation: the only
source writes `~39`.
- **CAD_PROGID** said nothing beyond the connection was tested against
BricsCAD/ZWCAD/GstarCAD, implying the connection was. It was not.
Perf lane reworked against a fresh paired measurement. The old text claimed
"28% slower than v1.4.0, median of three runs" and that
`EZDXF_CALL_TIMEOUT=0` restored v1.4.0's numbers "exactly". Three problems:
`perf_suite` runs each workload once per invocation and cannot produce a
median, "exactly" was refuted by the figures in its own sentence, and the
number moved when the timeout wrapper was rebuilt. It now reports the guard's
cost -- +20.8% and +19.9%, three paired runs, median taken by hand -- says the
figures have no published artifact, and says the wrapper rework recovered about
a third of the v1.5.0 overhead rather than closing it.
perf-ezdxf.json and the perf chart regenerated against the shipped code, and
the README table rebuilt from that artifact so the two cannot drift.
1267 passed, 1 xfailed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
**COM discarded areas it had already read.** `entity_measure` fell back to
rebuilding geometry from `.Coordinates` when `.Area` *or* `.Length` was
missing, coupling two independent property reads. HATCH, REGION and 3DSOLID
expose `.Area` and neither of the others, so the fallback fired, found no
vertices, and raised "AutoCAD reported neither an area nor usable vertices" --
while holding the area. Those are the exact types the README sells the COM
backend on. Now only a missing `.Area` triggers the fallback; an absent
perimeter is reported as `None` with `perimeter_exact: False` instead of being
invented, which also removes a `round(None, 6)` TypeError waiting for the first
type that answers `.Area` and nothing else.
**A plain square hatch reported itself self-intersecting.** `ezdxf.path
.flattening()` repeats the start point, and `is_self_intersecting` excluded
adjacent edges *by index*, so with the duplicate present the first and last
real edges were cross-tested and registered as crossing. Measured: a 40x40
SOLID hatch came back `area: 1600.0, self_intersecting: True`. A closed loop
now answers the same whether or not it carries its closing vertex, and a real
bowtie is still caught (`area: 0.0, self_intersecting: True`).
**The bowtie warning never fired on the live backend.** `self_intersecting` was
only computed inside the COM fallback, so an LWPOLYLINE bowtie -- which has
both `.Area` and `.Length` and never reaches it -- returned `area: 0.0,
self_intersecting: null`. The confident zero this field exists to flag, going
out unflagged. It is now computed on the normal path too.
Also corrects a docstring written earlier today in `contracts/analysis.py`
claiming the live backend "asks AutoCAD for `.Area`, which has already applied
the [hatch island] style". Whether ActiveX's hatch area applies the style is
not verified against a live seat, and the method could not measure a hatch at
all. It now says so.
The ActiveX member profiles behind the COM tests are assumptions, not facts,
and the tests say so; a live seat settles them. `.Circumference` for CIRCLE and
`.Perimeter` for REGION are deliberately NOT added here -- writing more members
from the reference is what produced four wrong beliefs in v1.4/v1.5.
1267 passed, 1 xfailed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three defects found after 1.5.0 reached PyPI, two of them by the adversarial
pass over the rewritten README.
The HTTP bind guard failed open on the launch path it was written for.
`_guarded_run_async` read the bind address from the `host` kwarg and treated an
absent one as loopback. It is not: `fastmcp run` forwards that kwarg only when
`--host` was passed, and fastmcp otherwise resolves the host from
`fastmcp.settings.host`, which is environment-backed. So
FASTMCP_HOST=0.0.0.0 fastmcp run server.py:mcp --transport http
bound every interface -- 149 tools including file open/save and AutoCAD command
execution, no token -- while the guard inspected the literal "127.0.0.1" and
waved it through. v1.4.0's changelog claimed this guard "fires on every launch
path"; it fired, and looked at the wrong host. It now resolves the effective
host, and warns rather than staying silent if it cannot.
`[com]` and `[full]` could not be installed off Windows. Both listed
`pywin32>=306` with no environment marker, and pywin32 publishes Windows wheels
and no sdist, so `pip install autocad-mcp-pro[full]` died at resolution on
Linux and macOS -- and `[full]` is the extra the README recommended to
contributors. CI never caught it because every Linux lane installs `[pdf]` or
syncs from uv.lock, whose resolution markers hide the gap. Both entries carry
`; sys_platform == "win32"` now, with a release-consistency test to keep them.
`drawing_settings({"dimscale": ...})` reported success and changed nothing
headlessly. It writes `$DIMSCALE`, which is what AutoCAD reads when it draws a
dimension, so the call worked on COM; ezdxf renders from the dimstyle table
entry and never consults the header. One call, one reported success, two
different drawings. The header value now folds into the per-dimension override
on all five dimension tools -- including `dimension_aligned` and
`dimension_angular`, which do not share the tolerance-override path and the
obvious fix would have missed. DIMSCALE 1.0 writes nothing.
README: every engine-conditional family in the feature table now says which
engine (boundary tracing, WIPEOUT/REVCLOUD/MTEXT masks, CHSPACE, typed hatch
edge paths, headless undo), and the checkout install line stopped recommending
an extra that could not resolve on the reader's platform.
1257 passed, 1 xfailed. 1.5.0 is to be yanked: its diameter and radius callouts
put wrong numbers on drawings at default settings.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`_c131_cpu.py`, `_c131_micro.py` and `_c131_probe.py` were throwaway probes
written by a subagent while checking a README claim, and a7c8a7a swept them
in with `git add -A`. They are not part of the project and nothing imports
them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`view_screenshot` renders at 100 DPI with a tight bounding box, so the
showcase PNG is intrinsically 698 x 500. The README displayed it at
width="820" -- a ~17% browser upscale of a raster, on a page whose subject
is drafting precision, and next to four charts that are resolution-
independent SVG. Native width instead; sharp beats big.
The rest of the hero caption was re-audited against the DXF it describes
and every item holds: 7 engineering layers, involute gear outline, DIN 6885
keyed bore, hatched section A-A, and two ISO 129 dimensions now measuring
144.000 and 40.000 with the H7 callout -- the two that were wrong before the
diameter fix in 8f405bd.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`add_diameter_dim` / `add_radius_dim` measure to the point they are handed;
`leader_length` is a text placement. The ezdxf backend passed
`centre + (radius + leader)` as that point, so the leader became part of the
measurement: a true 40 mm bore was dimensioned 60 at the default
`leader_length=10`, and 108 at 34. Radius callouts were `leader_length` too
large the same way.
This is the worst class of defect for this project -- a wrong number, at
default settings, on the one part of a drawing that exists to carry numbers --
and it shipped in v1.4.0 and earlier. The size now comes from `radius` and
`location` only places the text.
The live COM backend was never affected: ActiveX's AddDimDiametric takes the
two chord points and the leader length as separate arguments. The two engines
had been disagreeing by 2 x leader_length on the same call.
Found while building the drawing for the new README, whose own bore callout was
wrong. Two checks added to correctness_suite.py; both report fail -> pass
against v1.4.0, not miss -> pass, which is what separates a repaired defect
from new capability in that lane. A/B is now 21/26 -> 26/26, five fixed, zero
regressed.
README rewritten from scratch
- Hero is a real ISO A3 sheet the server drew: m6 z24 involute spur gear, DIN
6885 keyed bore, section A-A, ISO 129 dimensions, ISO 286 H7 fit, ISO 7200
title block, 56 entities, drawing_critique clean. New
scripts/render_readme_showcase.py rebuilds it, and the tool-call sequence
that produces it is printed in the page.
- Star history chart, badge row, capability matrix, five evidence lanes, and a
"what this release is bad at" section carrying the 28% throughput regression.
- Every image and repo link is now an absolute URL: README.md is the PyPI long
description, where relative paths 404.
Corrections this work forced, all measured
- The df=0 example was wrong. FILLET is already in entity_fillet's own text;
the commands that genuinely scored zero are WBLOCK, PEDIT, OVERKILL,
LAYTRANS and MATCHPROP.
- 40,305 tokens is the tools/list wire payload; ~25,143 is the part forwarded
to the model. The old wording said "of schema", which is neither.
- `--tokenizer anthropic` needs a package and credentials this repo does not
lock, and refuses rather than falling back. Said so.
- Discovery ranks over the 149-tool advertised catalog, not the 154 registered.
- Five middlewares, not four; the fifth is what makes refusals machine-readable.
- Five of the 27 capability keys are environment-dependent, so the matrix
carries footnotes instead of bare ticks.
- "Every coordinate is WCS" has one deliberate exception: TEXT rotation.
- render_live_chart's axis still said "fixed 10-task matrix" while our bar came
from v3's fifteen. Each bar now carries its own matrix.
- .env.example said MAX_UNDO_STACK=50 (code: 5) and omitted four real
variables, two of them the remote-HTTP controls.
- server.json described an empty ALLOWED_PATHS as blocking access. It disables
the allowlist.
- server.py called the lean profile ~46 tools; it is 47.
- backends/capability.py cited a test file that does not exist.
1249 passed, 1 xfailed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Python evaluates an `except` clause on the way out of every failed call,
not only the ones it is meant to catch. `pywintypes` is bound only when
`sys.platform == "win32"` and the import succeeded, so on any host
without pywin32 -- all of Linux, and Windows installed without the
`[com]` extra -- an ordinary exception leaving `ComBackend._run` hit that
clause and became `NameError: name 'pywintypes' is not defined`, burying
the real error.
Every Linux CI leg failed on exactly this, in
test_ordinary_errors_are_unchanged. The timeout tests survived only
because they matched an earlier `except`; nothing else in the suite sends
a non-timeout exception through `_run`.
Fixed with a module-level `_COM_ERROR` tuple: `(pywintypes.com_error,)`
where the import worked, `()` where it did not. An empty tuple catches
nothing, which is correct when no com_error type exists to be raised -- a
placeholder Exception subclass would be worse than the bug, silently
swallowing unrelated errors.
Reproduced on Windows before fixing by deleting the module attribute,
which resolves the name the same way the absent import does; that is now
a test, alongside one pinning `_COM_ERROR` as a tuple. Verified through
the locked environment CI uses: 1244 passed, 1 xfailed.
README test-count snapshot 1243 -> 1245.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The v1.5.0 lock was generated before the version bump, so its own
[[package]] entry said autocad-mcp-pro 1.4.0 while pyproject said 1.5.0.
No dependency differed -- the transitive graph is byte-identical -- but
`uv sync --locked` will not re-resolve, so lint, test-linux 3.11,
test-linux 3.12 and test-windows all failed at install before running a
single test.
`uv lock` fixes it in one line. The gate that should have caught it did
not exist: test_release_consistency chains pyproject -> version.py ->
CHANGELOG -> README -> server.json, and the lockfile was the one release
artifact carrying a version string nothing checked. Added
test_uv_lock_records_the_pyproject_version, verified RED against the
broken lock and green against the fixed one.
Reproduced the CI lanes locally into a throwaway environment before
pushing this time: `uv sync --locked --only-group dev`, then
`--locked --extra pdf`, then ruff and pytest under `uv run --no-sync`.
1242 passed, 1 xfailed.
README test-count snapshot 1242 -> 1243 for the new gate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
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>
Leaderboard restyled to the LLM-benchmark look: dark background, rank-numbered labels, per-rank color palette, 'NN / 100 - grade' annotations. The live-run, task-matrix and performance charts share the same dark theme for a consistent README identity. Charts regenerated deterministically from their data files.
Co-authored-by: Cursor <cursoragent@cursor.com>
source_review.json extended with three projects reviewed 2026-07-23 (LokmenoWer/best-cad-mcp B/74, daobataotie/CAD-MCP B/35, zh19980811/Easy-MCP-AutoCad C/30) and our entry updated to the released v1.4.0 (95: CI matrix, PyPI packaging, A/B gate, paper space, ISO 286, opt-in solids). Chart regenerated via render_chart; README benchmark section now opens with the leaderboard + 9-row table, source-review lane merged into it.
Co-authored-by: Cursor <cursoragent@cursor.com>
Ran benchmarks/compare_versions.py against the v1.3.0 tag: 21/21 on both versions, 0 fixed / 0 regressed - the v1.4.0 feature+infra work introduced no correctness regressions. Main README gains lane 4 (version A/B, with the historical v1.0.0 -> v1.1.0 contrast); benchmarks/README records the release-gate result and drops the stale 'performance coming' stub.
Co-authored-by: Cursor <cursoragent@cursor.com>
- TOOL_PROFILE=lean|core|full applied in the lifespan via server-level enable/disable; system_about reports the active profile; opt-in solid_* tools are hidden while ENABLE_3D=false
- layouts: layout_list/create/set_current + viewport_create (scaled model viewports) on both backends; drawing_export_pdf(layout=...); ezdxf uses TILEMODE + active-layout binding, COM switches ActiveLayout and restores it
- ISO 286: engineering/fits.py authored IT4-11 + fundamental-deviation tables (1-500 mm; shafts d..p/h/js, holes D..H/JS) with clear out-of-scope errors; dimension_linear/radius/diameter gained fit= resolved from the measured nominal
- tolerances: DIMTM now signed in deviation/limit modes so double-positive fits (p6 +0.035/+0.022) render correctly; legacy positive-magnitude contract preserved
- solids: solid_box/cylinder/extrude/revolve/boolean native on COM (AddBox/AddCylinder/AddRegion+AddExtrudedSolid/AddRevolvedSolid/Boolean); ezdxf reports the honest solid_3d capability boundary
- capability maps: paper_space native, new viewport_render feature, ENABLE_3D-gated solid_3d
- 60 new tests (profiles, layouts, viewports, layout PDF, fits, gating); version pins in tests now read version.__version__
Co-authored-by: Cursor <cursoragent@cursor.com>
ruff format was never run repo-wide; CI now enforces 'ruff format --check', so normalize every module once. No behavior changes.
Co-authored-by: Cursor <cursoragent@cursor.com>
dimension_linear always failed with '<unknown>.AddDimLinear' on the COM
backend — pywin32's dynamic dispatch error for a method that does not
exist: the AutoCAD ActiveX API has no AddDimLinear. AddDimRotated takes
the same arguments (XLine1Point, XLine2Point, DimLineLocation,
RotationAngle), so the call site only changes the method name.
Adds a mocked regression test asserting AddDimRotated is called with
the exact point/rotation arguments and AddDimLinear never is.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
pip install -e . failed with 'Unable to determine which files to ship
inside the wheel' — the flat layout has no directory matching the
project name (autocad_mcp_pro). Ship server.py plus its root-level
imports (config.py, security.py) and the backends/ and engineering/
packages explicitly.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adds production-grade engineering features and tools: 2D GD&T (feature control frames + datum features) via engineering/gdt.py, ISO‑129 dimension tolerance support (engineering/tolerances.py) and an ezdxf/COM plumbing for toleranced dims, a scalar drawing score + invalidity ratio (engineering/scoring.py) wired into drawing_finalize, in-place entity editors (entity_edit_text/entity_edit_geometry), and a friendly drawing_settings facade. Includes server tool wiring, COM/ezdxf backend implementations, datum-consistency critique, many unit tests, README/CLAUDE/CHANGELOG updates, and a version bump to 1.2.0. Also corrects COM backend capability label (entities_2d).
New COM-only Entity Query tool that reads AutoCAD's implied ("pickfirst")
selection set and returns the selected entity handles + EntityInfo, so the AI
can scope work (e.g. dimension_auto) to exactly what the user picked in the
viewport instead of the whole drawing. Resolves#1 — the layer-juggling
workaround is no longer needed.
- backends/base.py: abstract selection_get() returning a status dict
- backends/com_backend.py: reads ActiveDocument.PickfirstSelectionSet, surfaces
the PICKFIRST sysvar, handles the empty-selection case
- backends/ezdxf_backend.py: headless stub (ok=False, empty handles, same shape)
- server.py: @mcp.tool registration (readOnlyHint), SECTION 5 3 -> 4 tools
- README + CHANGELOG + pyproject: 110 -> 111 tools, v1.1.0 -> v1.1.1
Verified live on AutoCAD 2027 (COM): selection_get correctly returned the
user's grip selection across two distinct picks; 318 tests pass, ruff-clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- benchmarks/correctness_suite.py: 21 deterministic headless (ezdxf) checks
across core ops, dimensions, arrays, geometry, query/parity, render, the
quality gate, and engineering.
- benchmarks/compare_versions.py: one-command A/B runner — checks out an older
ref into a throwaway worktree and runs the same suite against both, each check
in its own subprocess so a crash is a miss, not a run-killer.
- benchmarks/README.md: methodology, results table, honesty caveats.
- README: Benchmarks section now reports the measured result —
v1.0.0 8/21 (38.1%) -> v1.1.0 21/21 (100%); 13 fixed, 0 regressions.
