Skip to content

Automate browser based workflows with AI

support pdf parser block to fix script gen error (#3646)

Shuchang Zheng16736686dfb1
diff --git a/skyvern/__init__.py b/skyvern/__init__.py
index d4ac57e49..59618edec 100644
--- a/skyvern/__init__.py
+++ b/skyvern/__init__.py
@@ -38,6 +38,7 @@ from skyvern.services.script_service import ( # noqa: E402
login, # noqa: E402
loop, # noqa: E402
parse_file, # noqa: E402
+ parse_pdf, # noqa: E402
prompt, # noqa: E402
render_list, # noqa: E402
render_template, # noqa: E402
@@ -64,6 +65,7 @@ __all__ = [
"login",
"loop",
"parse_file",
+ "parse_pdf",
"prompt",
"render_list",
"render_template",
diff --git a/skyvern/core/script_generations/generate_script.py b/skyvern/core/script_generations/generate_script.py
index dd6cc8860..e0f92d7d9 100644
--- a/skyvern/core/script_generations/generate_script.py
+++ b/skyvern/core/script_generations/generate_script.py
@@ -1044,6 +1044,55 @@ def _build_file_upload_statement(block: dict[str, Any]) -> cst.SimpleStatementLi
return cst.SimpleStatementLine([cst.Expr(cst.Await(call))])
+def _build_pdf_parser_statement(block: dict[str, Any]) -> cst.SimpleStatementLine:
+ """Build a skyvern.parse_pdf statement."""
+ args = [
+ cst.Arg(
+ keyword=cst.Name("file_url"),
+ value=_value(block.get("file_url", "")),
+ whitespace_after_arg=cst.ParenthesizedWhitespace(
+ indent=True,
+ last_line=cst.SimpleWhitespace(INDENT),
+ ),
+ ),
+ ]
+
+ if block.get("json_schema") is not None:
+ args.append(
+ cst.Arg(
+ keyword=cst.Name("schema"),
+ value=_value(block.get("json_schema")),
+ whitespace_after_arg=cst.ParenthesizedWhitespace(
+ indent=True,
+ last_line=cst.SimpleWhitespace(INDENT),
+ ),
+ )
+ )
+
+ if block.get("label") is not None:
+ args.append(
+ cst.Arg(
+ keyword=cst.Name("label"),
+ value=_value(block.get("label")),
+ whitespace_after_arg=cst.ParenthesizedWhitespace(
+ indent=True,
+ last_line=cst.SimpleWhitespace(INDENT),
+ ),
+ )
+ )
+ _mark_last_arg_as_comma(args)
+
+ call = cst.Call(
+ func=cst.Attribute(value=cst.Name("skyvern"), attr=cst.Name("parse_pdf")),
+ args=args,
+ whitespace_before_args=cst.ParenthesizedWhitespace(
+ indent=True,
+ last_line=cst.SimpleWhitespace(INDENT),
+ ),
+ )
+ return cst.SimpleStatementLine([cst.Expr(cst.Await(call))])
+
+
def _build_file_url_parser_statement(block: dict[str, Any]) -> cst.SimpleStatementLine:
"""Build a skyvern.parse_file statement."""
args = [
@@ -1531,6 +1580,8 @@ def _build_block_statement(block: dict[str, Any], data_variable_name: str | None
stmt = _build_file_url_parser_statement(block)
elif block_type == "http_request":
stmt = _build_http_request_statement(block)
+ elif block_type == "pdf_parser":
+ stmt = _build_pdf_parser_statement(block)
else:
# Default case for unknown block types
stmt = cst.SimpleStatementLine([cst.Expr(cst.SimpleString(f"# Unknown block type: {block_type}"))])
diff --git a/skyvern/services/script_service.py b/skyvern/services/script_service.py
index 3d6f68210..9ba6250ff 100644
--- a/skyvern/services/script_service.py
+++ b/skyvern/services/script_service.py
@@ -37,6 +37,7 @@ from skyvern.forge.sdk.workflow.models.block import (
ForLoopBlock,
HttpRequestBlock,
LoginBlock,
+ PDFParserBlock,
SendEmailBlock,
TaskBlock,
TextPromptBlock,
@@ -1822,6 +1823,29 @@ async def send_email(
)
+async def parse_pdf(
+ file_url: str,
+ schema: dict[str, Any] | None = None,
+ label: str | None = None,
+ parameters: list[PARAMETER_TYPE] | None = None,
+) -> None:
+ block_validation_output = await _validate_and_get_output_parameter(label)
+ file_url = _render_template_with_label(file_url, label)
+ pdf_parser_block = PDFParserBlock(
+ file_url=file_url,
+ json_schema=schema,
+ label=block_validation_output.label,
+ output_parameter=block_validation_output.output_parameter,
+ parameters=parameters or [],
+ )
+ await pdf_parser_block.execute_safe(
+ workflow_run_id=block_validation_output.workflow_run_id,
+ parent_workflow_run_block_id=block_validation_output.context.parent_workflow_run_block_id,
+ organization_id=block_validation_output.organization_id,
+ browser_session_id=block_validation_output.browser_session_id,
+ )
+
+
async def parse_file(
file_url: str,
file_type: FileType,