adapt for QCSchema v2 and Python v3.14 by loriab · Pull Request #110 · psi-rking/optking

loriab · 2026-02-28T08:24:12Z

Goals

Sorry for the monster PR. Basic goals are as it says on the tin:

let optking work with QCSchema v2. It's set up to emit v1 if fed v1 and emit v2 if fed v2, no cross like QCEngine or Psi4 itself allows.
let optking work with Python 3.14. Pydantic v1 API itself doesn't work with py314+, so using QCSchema v2 is the only way.
let optking continue working in old environments. Unlike QCManyBody, whose adaptation requires qcel v0.50.0rc3, optking should mostly keep working. This isn't guaranteed, but I do have this branch installed in the below, which isn't cutting edge:

numpy                     1.26.4          py311h64a7726_0    conda-forge
optking                   0.4.2+0.g6200e33.dirty           dev_0    <develop>
pint                      0.24.1             pyhd8ed1ab_1    conda-forge
psi4                      1.9.1           py311he3e7f2e_3    conda-forge
pydantic                  2.8.2              pyhd8ed1ab_0    conda-forge
python                    3.11.9          hb806964_0_cpython    conda-forge
qcelemental               0.28.0             pyhd8ed1ab_0    conda-forge
qcengine                  0.34.0             pyhd8ed1ab_0    conda-forge

Notes

Known Problems

with my old py311 env with no other modernized packages (conda list above), there's an error with sf4_quasilinear that doesn't really look related to these changes. I'm ignoring for now.
the test_atom is failing when running through v2 because you're setting an error and computing energy rather than setting a gradient. this'd be good to fix medium-term

Next Steps

QCElemental 0.50.0rc3 and QCManyBody 0.6.1 needed for fullest performance are released.
QCEngine 0.50.0rc2 is nearly there.
optking, then, is my last piece needed for Psi4 with QCSchema v2 or Py 3.14 . It'd be nice to get a v0.5 release of this in the next couple weeks. I can work on Psi4 with internal optking builds for a while. How does that timeframe sound?

Added After Above

Thank goodness you had done all that options prep for Pydantic v2. I needed to convert all the references in psi4's read_options to your v2 file since Psi4 is entirely v2, and the documentation tools (like Pyd itself) can't face both at once.

…ma v2-adapted psi4

…sed QCEngine and new v0.5x QCSchema v2-based QCEngine

…t_through_json excluded)

codecov · 2026-02-28T08:45:25Z

Codecov Report

❌ Patch coverage is 62.79070% with 80 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.12%. Comparing base (6200e33) to head (555d1d0).
⚠️ Report is 23 commits behind head on master.

Files with missing lines	Patch %	Lines
optking/optwrapper.py	55.95%	37 Missing ⚠️
optking/compute_wrappers.py	52.94%	24 Missing ⚠️
optking/opt_helper.py	71.79%	11 Missing ⚠️
optking/molsys.py	42.85%	4 Missing ⚠️
optking/optimize.py	50.00%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #110      +/-   ##
==========================================
- Coverage   82.74%   82.12%   -0.62%     
==========================================
  Files          36       36              
  Lines        7950     8068     +118     
==========================================
+ Hits         6578     6626      +48     
- Misses       1372     1442      +70

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

psi-rking

Assuming you can get this to run with v1 and v2 QCSchema, looks good to me!

On the output, it would be good to double-check that this is the best source of output data, but I don't think this need hold up your pull request here. Any tweaks there should not involve changes to format, right?

psi-rking · 2026-03-18T21:18:58Z

+    from qcelemental.models.v2 import Molecule, OptimizationInput, Model, AtomicSpecification
+
+    # WARNING. The user MUST set fix_com and fix_orientation to True.
+    # optimization will almost certainly fail otherwise


Why must orientation be fixed? Because qcengine doesn't guarantee gradients come back with the same COM/orientation? Or because qcengine has its own orientation? Just wondering.

I was just copying the warning from the v1 snippet. But yes, it's b/c input Cart w/o fix_* (or w/False) will return gradient in whatever the QC program's natural orientation is. fix_*=True demands results in input Cart orientation. At least one of your routes imposes the fix_*=True, but I didn't change or enforce that more or universally.

psi-rking · 2026-03-18T21:32:29Z


    result = optking.optimize_psi4("ccsd")
-    print(result["trajectory"][-1].keys())
+    trajkey = "trajectory" if _schver == 1 else "trajectory_results"


That change was getting a little annoying, I bet?

psi-rking · 2026-03-18T22:05:54Z

+        if spec_schema_name == "qcschema_many_body_specification":
+            model = "(proc_spec_in_options)"
+            options = opt_input["specification"]["specification"]
+            if "model" in (qc := options["specification"]):  # single spec


I think that I understand what this is doing; pretty cool!

AlexHeide · 2026-04-16T20:14:43Z

When I was last doing docs updates, the docstrings in v2/optparams.py didn't get all the updates that v1 received since only v1 was visible online. v2 should now render correctly. I've been having issues with the sphinx substitutions with autodoc_pydantic. I'm not sure if you've run into this as well Lori but I've resorted to using explicit :py:attr:name ` for now.

AlexHeide · 2026-04-16T20:16:20Z

I'm going to approve and will come back later to check SF4 quasilinear. That test breaking is not surprising - just a reminder to get molsym integrated.

AlexHeide

Thank you for all the code updates and keeping optking up to date with QC*

…ror)

AlexHeide · 2026-04-17T14:45:16Z

LGTM. Now that I've added some things I can't review though 😄

psi-rking · 2026-04-17T16:37:17Z

LGTM. Are we ready to merge? On the SF4 quasilinear case, optking has been unreliable in the past. I presume the problem is with linear bends, but I haven't looked at it in awhile. I'm interested in a keyword that would turn all derivative B matrices over to numerical determination. Then we only need handle discontinuities in the internal coordinate itself. For the purposes of optimization only, there never was a good reason to use analytic derivative expressions that introduce additional singularities.

…

On Fri, Apr 17, 2026 at 9:45 AM AlexHeide ***@***.***> wrote: *AlexHeide* left a comment (psi-rking/optking#110) <#110?email_source=notifications&email_token=AA4C4TFCZ6H3ZLY3H23HNJ34WI7RFA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMRWHEYDEMBRGQ3KM4TFMFZW63VQOJSXM2LFO5PXEZLROVSXG5DFMSSWK5TFNZ2LK4DSL5RW63LNMVXHIX3POBSW4X3DNRUWG2Y#issuecomment-4269020146> LGTM. Now that I've added some things I can't review though 😄 — Reply to this email directly, view it on GitHub <#110?email_source=notifications&email_token=AA4C4TFCZ6H3ZLY3H23HNJ34WI7RFA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMRWHEYDEMBRGQ3KM4TFMFZW63VQOJSXM2LFO5PXEZLROVSXG5DFMSSWK5TFNZ2LK4DSL5RW63LNMVXHIX3POBSW4X3DNRUWG2Y#issuecomment-4269020146>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA4C4TBTQHIYEX7FIKWIVCD4WI7RFAVCNFSM6AAAAACWCPXEWOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DENRZGAZDAMJUGY> . You are receiving this because your review was requested.Message ID: ***@***.***>

loriab · 2026-04-17T17:13:36Z

LGTM. Are we ready to merge?

The only thing I've though of that I haven't tested lately is new optking with old/current psi4. Though probably your CI covers that ... I have so many versions running around that I've lost reasoning capacity. I've just done a quick test locally, and it looks good.

I think it's ready for merge and tag. And if we're lucky it'll need zero-to-minimal tuneups for the next Psi4 release. Thanks!

loriab added 7 commits February 26, 2026 10:48

allow optking.optimize_psi4() to work with QCSchema v1 or v2

617375e

allow psi4.optimize() calling optking to work with existing or QCSche…

d72782b

…ma v2-adapted psi4

allow qcengine.compute(optking) to work with old v0.3x QCSchema v1-ba…

8530e28

…sed QCEngine and new v0.5x QCSchema v2-based QCEngine

mol_input tests working for v1/v2

7484b30

test suite changes to allow QCSchema v1 or v2

95188ca

allows full optking test suite to work in three v1/v2 envs (test_inpu…

eae7343

…t_through_json excluded)

make psi4 1.9.1 work again

8c0e7a5

AlexHeide reviewed Feb 28, 2026

View reviewed changes

Comment thread optking/optwrapper.py Outdated

AlexHeide reviewed Feb 28, 2026

View reviewed changes

Comment thread optking/tests/test_2_hessians.py

Lori A. Burns added 5 commits March 8, 2026 01:18

now mbe opt too

1fc425c

properties

393e20f

pass through protocols

086bf12

one more fix and a few docs

f7fa2af

docs?

19d9c37

loriab marked this pull request as ready for review March 10, 2026 02:03

Lori A. Burns added 4 commits March 11, 2026 03:03

program fixes

101397f

hopefully program compromise

1d88bfb

misc

7be5e62

until qcng rc2

fb75449

loriab requested review from AlexHeide and psi-rking March 13, 2026 06:21

Lori A. Burns added 2 commits March 14, 2026 02:06

convert docs v2

facc6d1

fix NRE

c6f4e2e

psi-rking reviewed Mar 18, 2026

View reviewed changes

psi-rking previously approved these changes Mar 18, 2026

View reviewed changes

loriab mentioned this pull request Mar 21, 2026

Py314 and QCSchema v2 psi4/psi4#3341

Merged

40 tasks

fixes and extensions for rc4

014d589

loriab dismissed psi-rking’s stale review via 014d589 April 1, 2026 16:40

add optking

c707b49

Update docs to match v1. Fixes tables, links, and highlighting

83e5e8d

AlexHeide previously approved these changes Apr 16, 2026

View reviewed changes

Adds xfail if pubhcem cannot be reached (only checks for validationer…

555d1d0

…ror)

AlexHeide dismissed their stale review via 555d1d0 April 17, 2026 14:31

AlexHeide requested a review from psi-rking April 17, 2026 14:43

psi-rking approved these changes Apr 17, 2026

View reviewed changes

AlexHeide merged commit 753c7c9 into psi-rking:master Apr 17, 2026
6 checks passed

loriab deleted the py314_v1v2 branch April 29, 2026 06:40

Conversation

loriab commented Feb 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Goals

Notes

Known Problems

Next Steps

Added After Above

Uh oh!

codecov Bot commented Feb 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

Uh oh!

psi-rking left a comment

Choose a reason for hiding this comment

Uh oh!

psi-rking Mar 18, 2026

Choose a reason for hiding this comment

Uh oh!

loriab Mar 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

psi-rking Mar 18, 2026

Choose a reason for hiding this comment

Uh oh!

psi-rking Mar 18, 2026

Choose a reason for hiding this comment

Uh oh!

AlexHeide commented Apr 16, 2026

Uh oh!

AlexHeide commented Apr 16, 2026

Uh oh!

AlexHeide left a comment

Choose a reason for hiding this comment

Uh oh!

AlexHeide commented Apr 17, 2026

Uh oh!

psi-rking commented Apr 17, 2026 via email

Uh oh!

loriab commented Apr 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

loriab commented Feb 28, 2026 •

edited

Loading

codecov Bot commented Feb 28, 2026 •

edited

Loading

loriab Mar 21, 2026 •

edited

Loading