update DPA4 support, add melting point calculation

.gitignore

@@ -49,6 +49,8 @@ confs
 # Generated MatMaster skill archive
 apex-flow.zip
 
-# Large DeePMD checkpoints (fetch with scripts/fetch_models.py; keep frozen *.pb in git)
+# Exclude arbitrary DeePMD training/source checkpoints, but keep the one
+# authoritative single-task DPA4 model bundled by apex-flow.
 apex/skills/apex-flow/models/**/*.pt
+!apex/skills/apex-flow/models/DPA4-alloytongqi/model.pt
 apex/skills/apex-flow/models/**/*.partial

.pre-commit-config.yaml

@@ -0,0 +1,36 @@
+# Local pre-commit checks (no remote hook repos; works behind restricted proxies).
+# Install once:
+#   pip install pre-commit && pre-commit install
+repos:
+  - repo: local
+    hooks:
+      - id: check-merge-conflict
+        name: check merge conflict markers
+        entry: python3 -c
+        args:
+          - |
+            import pathlib, re, sys
+            pat = re.compile(r"^(<<<<<<< |>>>>>>> )")
+            bad = False
+            for raw in sys.argv[1:]:
+                path = pathlib.Path(raw)
+                if not path.is_file():
+                    continue
+                try:
+                    text = path.read_text(encoding="utf-8", errors="ignore")
+                except Exception:
+                    continue
+                for i, line in enumerate(text.splitlines(), 1):
+                    if pat.match(line):
+                        print(f"{path}:{i}: merge conflict marker")
+                        bad = True
+            raise SystemExit(1 if bad else 0)
+        language: system
+        types: [text]
+
+      - id: apex-unit-smoke
+        name: APEX unit smoke (gamma + ticket + failed-result)
+        entry: bash -c 'cd tests && python3 -m unittest -v test_gamma.TestGamma.test_compute_lower test_skill_scripts.TestGenerateConfigHelpers.test_get_bohrium_ticket_success test_ops.TestTaskStatusHelpers.test_is_failed_task_result_accepts_dict_and_mapping_like'
+        language: system
+        pass_filenames: false
+        types: [python]

CHANGELOG-1.3.0.md

@@ -15,6 +15,8 @@ Major additions include:
 - Finite-temperature lattice workflow improvements
 - Finite-temperature elasticity workflow support and example documentation
 - Annealing workflow support for LAMMPS
+- Melting-point workflow support for solid-liquid coexistence calculations
+- Image-resident DPA4/PT2 runtime support with a fail-closed T4 qualification profile
 - Shape-controlled automatic supercell generation
 - `req_calc`-based relaxation/property selection for joint workflows
 - GUI-side batched submission for large configuration sets
@@ -74,6 +76,21 @@ Major additions include:
 - Added LAMMPS annealing workflow support.
 - Improved generated annealing input scripts for heating, cooling, RDF output, and holding stages.
 
+### Melting Point Workflow
+
+- Added a LAMMPS solid-liquid coexistence workflow for estimating melting points across multiple temperatures and replicas.
+- Added restart-file staging, interface-axis controls, per-temperature restart validation, and generated skill templates.
+- Added input validation and regression coverage for melting-specific temperature, replica, restart, and backend constraints.
+
+### DPA4/PT2 Runtime and Qualification
+
+- Added image-resident DeePMD PT2 model handling so immutable runtime artifacts are not copied into task upload packages.
+- Added DPA4-aware LAMMPS input generation, including the required atom maps, runtime plugin auto-loading, and rejection of incompatible legacy plugin commands.
+- Added an audited `DPA4-alloytongqi` source checkpoint, container wrapper templates, runtime manifest template, and a reproducible CPU/GPU and phonoLAMMPS benchmark harness.
+- Added a single-source DPA4 runtime profile used by generation, recommendation, standalone validation, and submission. The bundled profile remains fail-closed until an immutable image `ref@digest` passes the recorded post-snapshot T4 qualification.
+- Restricted the qualified production contract to one MPI rank on one `c4_m15_1 * NVIDIA T4`, with exact image, wrapper, model hash, and dispatcher validation.
+- Added complete automatic `type_map` expansion across every resolved structure and every effective LAMMPS `overwrite_interaction`.
+
 ### Workflow Selection and Submission
 
 - Added `req_calc`-based workflow selection for relaxation and property calculations.
@@ -94,6 +111,7 @@ Major additions include:
   - `.debug.log`
   - `.debug.stdout`
   - `.debug.stderr`
+- Added bounded retries for transient LAMMPS remote-startup failures, including header-only logs, retry evidence preservation, and explicit retry classification in task status files.
 
 ### Reporting
 
@@ -126,4 +144,5 @@ Major additions include:
 - Added runnable examples for RSS, GammaSurface, and finite-temperature elasticity.
 - Added GUI developer documentation.
 - Added `monty` to the package dependencies and constrained supported Python versions to `<3.13`.
-- Added support for Phonopy v4 in terms of phonon calculation.
+- Added Phonopy v4-compatible setup, force-constant, and band-generation fallbacks.
+- Expanded the bundled `apex-flow` skill with backend-aware generation and validation for Bohrium, local debug, local cluster, VASP, LAMMPS, DPA4/PT2, melting-point, gamma, and finite-temperature workflows.