optimize svg cache and add retry (#752)

skyvern/exceptions.py

@@ -293,6 +293,11 @@ class StepUnableToExecuteError(SkyvernException):
         super().__init__(f"Step {step_id} cannot be executed and task execution is stopped. Reason: {reason}")
 
 
+class SVGConversionFailed(SkyvernException):
+    def __init__(self, svg_html: str) -> None:
+        super().__init__(f"Failed to convert SVG after max retries. svg_html={svg_html}")
+
+
 class UnsupportedActionType(SkyvernException):
     def __init__(self, action_type: str):
         super().__init__(f"Unsupport action type: {action_type}")

skyvern/forge/agent_functions.py

@@ -1,3 +1,4 @@
+import asyncio
 import copy
 import hashlib
 from typing import Awaitable, Callable, Dict, List
@@ -6,7 +7,7 @@ import structlog
 from playwright.async_api import Page
 
 from skyvern.constants import SKYVERN_ID_ATTR
-from skyvern.exceptions import StepUnableToExecuteError
+from skyvern.exceptions import StepUnableToExecuteError, SVGConversionFailed
 from skyvern.forge import app
 from skyvern.forge.async_operations import AsyncOperation
 from skyvern.forge.prompts import prompt_engine
@@ -19,6 +20,9 @@ CleanupElementTreeFunc = Callable[[str, list[dict]], Awaitable[list[dict]]]
 
 LOG = structlog.get_logger()
 
+USELESS_SVG_ATTRIBUTE = [SKYVERN_ID_ATTR, "id", "aria-describedby"]
+SVG_RETRY_ATTEMPT = 3
+
 
 def _remove_rect(element: dict) -> None:
     if "rect" in element:
@@ -38,8 +42,11 @@ def _remove_skyvern_attributes(element: Dict) -> Dict:
         if element_copied.get(attr):
             del element_copied[attr]
 
-    if element_copied.get("attributes") and SKYVERN_ID_ATTR in element_copied.get("attributes", {}):
+    if "attributes" in element_copied:
-        del element_copied["attributes"][SKYVERN_ID_ATTR]
+        attributes: dict = copy.deepcopy(element_copied.get("attributes", {}))
+        for key in attributes.keys():
+            if key in USELESS_SVG_ATTRIBUTE:
+                del element_copied["attributes"][key]
 
     children: List[Dict] | None = element_copied.get("children", None)
     if children is None:
@@ -80,6 +87,8 @@ async def _convert_svg_to_string(task: Task, step: Step, organization: Organizat
     else:
         LOG.debug("call LLM to convert SVG to string shape", element_id=element_id)
         svg_convert_prompt = prompt_engine.load_prompt("svg-convert", svg_element=svg_html)
+
+        for retry in range(SVG_RETRY_ATTEMPT):
             try:
                 json_response = await app.SECONDARY_LLM_API_HANDLER(prompt=svg_convert_prompt, step=step)
                 svg_shape = json_response.get("shape", "")
@@ -87,13 +96,16 @@ async def _convert_svg_to_string(task: Task, step: Step, organization: Organizat
                     raise Exception("Empty SVG shape replied by secondary llm")
                 LOG.info("SVG converted by LLM", element_id=element_id, shape=svg_shape)
                 await app.CACHE.set(svg_key, svg_shape)
+                break
             except Exception:
                 LOG.exception(
-                "Failed to convert SVG to string shape by secondary llm",
+                    "Failed to convert SVG to string shape by secondary llm. Will retry if haven't met the max try attempt after 3s.",
-                element=element,
+                    element_id=element_id,
-                svg_html=svg_html,
+                    retry=retry,
                 )
-            return
+                await asyncio.sleep(3)
+        else:
+            raise SVGConversionFailed(svg_html=svg_html)
 
     element["attributes"] = dict()
     element["attributes"]["alt"] = svg_shape