fix: use navigation_goal/data_extraction_goal for self-hosted Skyvern API

Cloud API uses 'prompt', self-hosted uses 'navigation_goal' and
'data_extraction_goal'. Pass them separately.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

cron/browser_executor.py

@@ -16,16 +16,17 @@ POLL_INTERVAL = 5  # seconds
 MAX_POLL_TIME = 300  # 5 minutes
 
 
-async def _create_task(url: str, goal: str, extraction_schema: dict | None = None,
+async def _create_task(url: str, goal: str, extraction_goal: str = "",
+                       extraction_schema: dict | None = None,
                        credential_id: str | None = None, totp_identifier: str | None = None) -> str:
-    """Create a Skyvern task and return the run_id."""
+    """Create a Skyvern task and return the task_id."""
     payload: dict = {
-        "prompt": goal,
         "url": url,
-        "engine": "skyvern-v2",
+        "navigation_goal": goal,
+        "data_extraction_goal": extraction_goal or goal,
     }
     if extraction_schema:
-        payload["data_extraction_schema"] = extraction_schema
+        payload["extracted_information_schema"] = extraction_schema
     if credential_id:
         payload["credential_id"] = credential_id
     if totp_identifier:
@@ -82,10 +83,8 @@ async def execute_browser_scrape(job: dict, send_text, **_kwargs) -> dict:
     target_room = job["targetRoom"]
     config = job.get("config", {})
     url = config.get("url", "")
-    extraction_goal = config.get("extractionGoal", "")
     goal = config.get("goal", config.get("query", f"Scrape content from {url}"))
-    if extraction_goal:
+    extraction_goal = config.get("extractionGoal", "") or goal
-        goal += f"\n\nExtract the following: {extraction_goal}"
     extraction_schema = config.get("extractionSchema")
     browser_profile = job.get("browserProfile")
 
@@ -118,6 +117,7 @@ async def execute_browser_scrape(job: dict, send_text, **_kwargs) -> dict:
         run_id = await _create_task(
             url=url,
             goal=goal,
+            extraction_goal=extraction_goal,
             extraction_schema=extraction_schema,
             credential_id=credential_id,
             totp_identifier=totp_identifier,

pipelines/steps/skyvern.py

@@ -42,20 +42,15 @@ async def execute_skyvern(config: dict, send_text=None, target_room: str = "", *
     if not url or not goal:
         raise ValueError("Skyvern step requires 'url' and 'goal' in config")
 
-    # Build prompt combining goal and extraction goal
-    prompt = goal
-    if data_extraction_goal:
-        prompt += f"\n\nExtract the following: {data_extraction_goal}"
-
     payload: dict = {
-        "prompt": prompt,
         "url": url,
-        "engine": "skyvern-v2",
+        "navigation_goal": goal,
+        "data_extraction_goal": data_extraction_goal or goal,
     }
     if extraction_schema:
         if isinstance(extraction_schema, str):
             extraction_schema = json.loads(extraction_schema)
-        payload["data_extraction_schema"] = extraction_schema
+        payload["extracted_information_schema"] = extraction_schema
     if credential_id:
         payload["credential_id"] = credential_id
     if totp_identifier: