Validate Bohrium project IDs before submission

Prevent agents from serializing project IDs as strings by clarifying generated config usage and rejecting malformed Bohrium settings during input validation.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: Zikkying

apex/skills/apex-flow/reference/submission.md

@@ -138,18 +138,23 @@ dflow validates workflow names against RFC 1123 subdomain regex. Names like `"Cu
 
 `program_id` and `bohrium_config.project_id` **must** come from the environment
 variable `BOHRIUM_PROJECT_ID` (or `--project-id`). Never hardcode a personal
-project ID in examples or generated configs.
+project ID in examples or generated configs. **Do not write `global.json`
+manually and do not copy the placeholders below into a real file.** Run
+`scripts/generate_config.py`; it converts the environment string to an integer
+and writes both ID fields as JSON numbers.
 
-```json
+The following is a type-annotated shape, not valid JSON:
+
+```text
 {
     "dflow_host": "https://workflows.deepmodeling.com",
     "k8s_api_server": "https://workflows.deepmodeling.com",
     "batch_type": "Bohrium",
     "context_type": "Bohrium",
-    "program_id": "<BOHRIUM_PROJECT_ID>",
+    "program_id": <BOHRIUM_PROJECT_ID as an unquoted integer>,
     "bohrium_config": {
         "ticket": "<UUID from API conversion — auto-filled by generate_config.py>",
-        "project_id": "<BOHRIUM_PROJECT_ID>"
+        "project_id": <the same unquoted integer>
     },
     "apex_image_name": "registry.dp.tech/dptech/dp/native/prod-397637/apex-flow:1.3.0.post2",
     "lammps_image_name": "registry.dp.tech/dptech/dp/native/prod-397637/deepmd-kit-phonolammps:3.1.3",
@@ -160,6 +165,15 @@ project ID in examples or generated configs.
 }
 ```
 
+Quoted digits such as `"program_id": "..."` are invalid for DPDispatcher even
+though they look numeric. Before submission, always run:
+
+```bash
+python scripts/validate_inputs.py --param param.json --global global.json
+```
+
+Do not submit unless validation reports `Validation PASSED`.
+
 > For GPU potentials (DeePMD, MACE, NEP), change `scass_type` to `"c8_m31_1 * NVIDIA T4"`.
 > Before submitting, run `scripts/validate_apex_combo.py check` on the chosen image × scass_type.
 

apex/skills/apex-flow/scripts/validate_inputs.py

@@ -40,21 +40,81 @@ def validate_global(global_config: dict) -> list:
     errors = []
     warnings = []
 
-    # Check machine config
-    if "machine" not in global_config:
-        errors.append("Missing 'machine' section in global.json")
+    # generate_config.py writes the current, top-level APEX schema. Keep
+    # accepting the older nested-machine schema for existing user configs.
+    is_current_schema = any(
+        key in global_config
+        for key in ("dflow_host", "context_type", "bohrium_config", "program_id")
+    )
+
+    if is_current_schema:
+        for key in ("batch_type", "context_type"):
+            if not global_config.get(key):
+                errors.append(f"Missing '{key}' in global.json")
+
+        program_id = global_config.get("program_id")
+        if not isinstance(program_id, int) or isinstance(program_id, bool):
+            errors.append(
+                "'program_id' must be an unquoted JSON integer, not a string; "
+                "generate global.json with "
+                "generate_config.py"
+            )
+        elif program_id <= 0:
+            errors.append("'program_id' must be a positive integer")
+
+        bohrium_config = global_config.get("bohrium_config")
+        if not isinstance(bohrium_config, dict):
+            errors.append("Missing 'bohrium_config' section in global.json")
+        else:
+            project_id = bohrium_config.get("project_id")
+            if not isinstance(project_id, int) or isinstance(project_id, bool):
+                errors.append(
+                    "'bohrium_config.project_id' must be a JSON integer, "
+                    "not a quoted string"
+                )
+            elif project_id <= 0:
+                errors.append(
+                    "'bohrium_config.project_id' must be a positive integer"
+                )
+            elif isinstance(program_id, int) and not isinstance(program_id, bool):
+                if project_id != program_id:
+                    errors.append(
+                        "'program_id' and 'bohrium_config.project_id' must match"
+                    )
+
+            ticket = bohrium_config.get("ticket")
+            if not isinstance(ticket, str) or not ticket.strip():
+                errors.append(
+                    "Missing non-empty 'bohrium_config.ticket'; regenerate "
+                    "global.json with generate_config.py"
+                )
+
+        if not global_config.get("scass_type"):
+            errors.append("Missing 'scass_type' in global.json")
+
+        run_commands = (
+            "lammps_run_command", "abacus_run_command", "vasp_run_command"
+        )
+        if not any(global_config.get(key) for key in run_commands):
+            errors.append(
+                "Missing calculator run command in global.json "
+                "(expected one of lammps_run_command, abacus_run_command, "
+                "vasp_run_command)"
+            )
     else:
-        machine = global_config["machine"]
-        if "batch_type" not in machine:
-            errors.append("Missing 'machine.batch_type'")
+        # Legacy nested-machine schema.
+        if "machine" not in global_config:
+            errors.append("Missing 'machine' section in global.json")
+        else:
+            machine = global_config["machine"]
+            if "batch_type" not in machine:
+                errors.append("Missing 'machine.batch_type'")
 
-    # Check resources
-    if "resources" not in global_config:
-        warnings.append("No 'resources' section - will use defaults")
+        if "resources" not in global_config:
+            warnings.append("No 'resources' section - will use defaults")
 
-    # Check run_command
-    if "run_command" not in global_config:
-        errors.append("Missing 'run_command' in global.json")
+        if "run_command" not in global_config:
+            errors.append("Missing 'run_command' in global.json")
 
     return errors, warnings
 

tests/test_skill_scripts.py

@@ -311,6 +311,34 @@ def test_validate_global(self):
             ([], []),
         )
 
+    def test_validate_current_global_requires_integer_matching_project_ids(self):
+        config = {
+            "dflow_host": "https://workflows.deepmodeling.com",
+            "batch_type": "Bohrium",
+            "context_type": "Bohrium",
+            "program_id": "42",
+            "bohrium_config": {
+                "ticket": "ticket",
+                "project_id": "42",
+            },
+            "scass_type": "c8_m31_1 * NVIDIA T4",
+            "lammps_run_command": "lmp -in in.lammps",
+        }
+        errors, warnings = self.validator.validate_global(config)
+        self.assertTrue(any("'program_id'" in error for error in errors))
+        self.assertTrue(
+            any("'bohrium_config.project_id'" in error for error in errors)
+        )
+        self.assertFalse(warnings)
+
+        config["program_id"] = 42
+        config["bohrium_config"]["project_id"] = 43
+        errors, _ = self.validator.validate_global(config)
+        self.assertTrue(any("must match" in error for error in errors))
+
+        config["bohrium_config"]["project_id"] = 42
+        self.assertEqual(self.validator.validate_global(config), ([], []))
+
     def test_validate_interaction(self):
         cases = (
             ({}, "Missing 'interaction.type'"),