feat: added dynamic variable support

[sauravpanda]

Summary by cubic

Adds automatic variable identification and parameterization to workflows, generating an input_schema and replacing hardcoded values with placeholders. Integrates into the semantic converter and healing flow with a fast, pattern-based pass and optional YAML cleanup.

• New Features

  • New VariableIdentifier with pattern/context heuristics and optional LLM; exposes identify_variables_in_workflow.
  • Auto-generates input_schema and metadata. Presets via VariableConfigPresets.
  • SemanticWorkflowConverter now identifies variables (enable_variable_identification, variable_config).
  • Healing service post-process adds pattern-based identification (enable_pattern_variable_identification, pattern_variable_confidence) and YAML cleanup (cleanup_yaml, remove_descriptions, remove_verification_checks, remove_expected_outcomes).
  • Tests added for variable detection, presets, and converter integration.
  • Implements Linear 134 by introducing an AI-powered variable identifier with a zero-cost pattern mode.

• Migration

  • YAML cleanup defaults to on; disable via cleanup_yaml=False to preserve descriptions/verification/expected_outcome.
  • Pattern-based variable identification defaults to on; disable via enable_pattern_variable_identification=False.
  • Adjust confidence via pattern_variable_confidence or converter variable_config.min_confidence.

^{Written for commit 1fe6e86. Summary will update automatically on new commits.}

[cubic-dev-ai]

3 issues found across 8 files

Prompt for AI agents (all 3 issues)

Understand the root cause of the following 3 issues and fix them.


<file name="workflows/workflow_use/workflow/tests/test_variable_identifier.py">

<violation number="1" location="workflows/workflow_use/workflow/tests/test_variable_identifier.py:183">
`email_schema[&#39;type&#39;]` is asserted to equal `&#39;email&#39;`, but `_generate_input_schema` maps email variables to schema type `&#39;string&#39;` and uses the `&#39;format&#39;` field for email specificity. Update the test to check for `&#39;string&#39;` (and optionally the `&#39;format&#39;`) so it reflects the actual schema.</violation>
</file>

<file name="workflows/workflow_use/workflow/variable_config.py">

<violation number="1" location="workflows/workflow_use/workflow/variable_config.py:93">
`domain_config` should use `typing.Any` instead of the builtin `any`; otherwise static type checkers will flag invalid annotations.</violation>
</file>

<file name="workflows/workflow_use/workflow/variable_identifier.py">

<violation number="1" location="workflows/workflow_use/workflow/variable_identifier.py:287">
Context-based matches can return a None variable name, which ends up as `{None}` in the workflow and name=None in input_schema; this breaks the generated workflow defaults and schema consumers. Please fall back to a type-derived name when context is missing.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

domain_config should use typing.Any instead of the builtin any; otherwise static type checkers will flag invalid annotations.

Prompt for AI agents

Address the following comment on workflows/workflow_use/workflow/variable_config.py at line 93:

<comment>`domain_config` should use `typing.Any` instead of the builtin `any`; otherwise static type checkers will flag invalid annotations.</comment>

<file context>
@@ -0,0 +1,254 @@
+	group_related_fields: bool = True
+
+	# Domain-specific configurations
+	domain_config: Dict[str, any] = field(default_factory=dict)
+
+
</file context>

[cubic-dev-ai]

Context-based matches can return a None variable name, which ends up as {None} in the workflow and name=None in input_schema; this breaks the generated workflow defaults and schema consumers. Please fall back to a type-derived name when context is missing.

Prompt for AI agents

Address the following comment on workflows/workflow_use/workflow/variable_identifier.py at line 287:

<comment>Context-based matches can return a None variable name, which ends up as `{None}` in the workflow and name=None in input_schema; this breaks the generated workflow defaults and schema consumers. Please fall back to a type-derived name when context is missing.</comment>

<file context>
@@ -0,0 +1,526 @@
+		for keyword, var_type in self.VARIABLE_KEYWORDS.items():
+			if keyword in combined_hints:
+				# Generate variable name from context
+				var_name = self._generate_variable_name_from_context(context, var_type)
+				confidence = 0.85  # Good confidence for context matches
+				logger.debug(f&quot;Context match: keyword &#39;{keyword}&#39; → {var_type} (name: {var_name})&quot;)
</file context>

✅ Addressed in 1fe6e86

[cubic-dev-ai]

Reviewed changes from recent commits (found 1 issue).

1 issue found across 1 file

Prompt for AI agents (all 1 issues)

Understand the root cause of the following 1 issues and fix them.


<file name="workflows/workflow_use/workflow/tests/test_variable_identifier.py">

<violation number="1" location="workflows/workflow_use/workflow/tests/test_variable_identifier.py:183">
This line is indented with spaces while the rest of the block uses tabs, which will trigger a TabError when the test file is executed. Please switch the indentation to match the surrounding tabs.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

This line is indented with spaces while the rest of the block uses tabs, which will trigger a TabError when the test file is executed. Please switch the indentation to match the surrounding tabs.

Prompt for AI agents

Address the following comment on workflows/workflow_use/workflow/tests/test_variable_identifier.py at line 183:

<comment>This line is indented with spaces while the rest of the block uses tabs, which will trigger a TabError when the test file is executed. Please switch the indentation to match the surrounding tabs.</comment>

<file context>
@@ -180,7 +180,7 @@ def test_input_schema_generation(self):
 		email_schema = next((s for s in schema if &#39;email&#39; in s[&#39;name&#39;]), None)
 		assert email_schema is not None
-		assert email_schema[&#39;type&#39;] == &#39;email&#39;
+        assert email_schema[&#39;type&#39;] == &#39;string&#39;
 		assert email_schema[&#39;required&#39;] is True
 
</file context>

Suggested change

	assert email_schema['type'] == 'string'
	assert email_schema['type'] == 'string'

[cubic-dev-ai]

No issues found across 6 files