update api for subagent protocol and delete hermes agent

browser_env/browser_use_runner.py

@@ -1,15 +1,60 @@
 import asyncio
+import inspect
 import json
 import os
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
-from typing import Any
+from typing import Any, Literal
 from urllib import error, request
 
 from browser_use import Agent, Browser, ChatOpenAI
+from pydantic import BaseModel, Field, ValidationError, field_validator
 
 
-def _json_response(handler, status_code, payload):
-    data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+class RunTaskRequest(BaseModel):
+    """RPC payload для запуска browser-use задачи."""
+
+    task: str = Field(..., min_length=1)
+
+    @field_validator("task")
+    @classmethod
+    def validate_task(cls, value: str) -> str:
+        normalized = value.strip()
+        if not normalized:
+            raise ValueError("Field 'task' is required")
+        return normalized
+
+
+class HistoryEvent(BaseModel):
+    """Нормализованное событие из history агента."""
+
+    step: int
+    kind: str
+    content: str | None = None
+    data: dict[str, Any] = Field(default_factory=dict)
+
+
+class RunTaskSuccessResponse(BaseModel):
+    """Успешный ответ RPC раннера."""
+
+    success: Literal[True] = True
+    result: str | None = None
+    history: list[HistoryEvent] = Field(default_factory=list)
+    browser_view: str = ""
+
+
+class RunTaskErrorResponse(BaseModel):
+    """Ошибка выполнения задачи в RPC раннере."""
+
+    success: Literal[False] = False
+    error: str
+
+
+def _json_response(handler, status_code: int, payload: dict[str, Any] | BaseModel) -> None:
+    if isinstance(payload, BaseModel):
+        body = payload.model_dump(mode="json")
+    else:
+        body = payload
+    data = json.dumps(body, ensure_ascii=False).encode("utf-8")
     handler.send_response(status_code)
     handler.send_header("Content-Type", "application/json; charset=utf-8")
     handler.send_header("Content-Length", str(len(data)))
@@ -17,7 +62,7 @@ def _json_response(handler, status_code, payload):
     handler.wfile.write(data)
 
 
-async def run_browser_task(task):
+async def run_browser_task(task: str) -> RunTaskSuccessResponse | RunTaskErrorResponse:
     cdp_url = os.getenv("BROWSER_CDP_URL", "http://127.0.0.1:9222")
     browser_view_url = os.getenv("BROWSER_VIEW_URL", "")
 
@@ -34,17 +79,20 @@ async def run_browser_task(task):
 
     try:
         history = await agent.run()
-        return {
-            "success": True,
-            "result": history.final_result(),
-            "history": _extract_history_events(history),
-            "browser_view": browser_view_url,
-        }
+        return RunTaskSuccessResponse(
+            result=history.final_result(),
+            history=[HistoryEvent.model_validate(item) for item in _extract_history_events(history)],
+            browser_view=browser_view_url,
+        )
     except Exception as err:
-        return {"success": False, "error": f"Browser automation failed: {err}"}
+        return RunTaskErrorResponse(error=f"Browser automation failed: {err}")
     finally:
         try:
-            await browser.close()
+            close_method = getattr(browser, "close", None)
+            if callable(close_method):
+                close_result = close_method()
+                if inspect.isawaitable(close_result):
+                    await close_result
         except Exception:
             pass
 
@@ -156,20 +204,19 @@ class BrowserUseRPCHandler(BaseHTTPRequestHandler):
             content_length = int(self.headers.get("Content-Length", "0"))
             raw = self.rfile.read(content_length)
             payload = json.loads(raw.decode("utf-8") if raw else "{}")
-            task = payload.get("task", "")
-            if not isinstance(task, str) or not task.strip():
-                _json_response(self, 400, {"success": False, "error": "Field 'task' is required"})
-                return
+            request_model = RunTaskRequest.model_validate(payload)
 
-            result = asyncio.run(run_browser_task(task.strip()))
-            code = 200 if result.get("success") else 500
-            _json_response(self, code, result)
+            result_model = asyncio.run(run_browser_task(request_model.task))
+            code = 200 if result_model.success else 500
+            _json_response(self, code, result_model)
+        except ValidationError as err:
+            _json_response(self, 400, RunTaskErrorResponse(error=f"Invalid request payload: {err.errors()}"))
         except json.JSONDecodeError:
-            _json_response(self, 400, {"success": False, "error": "Invalid JSON payload"})
+            _json_response(self, 400, RunTaskErrorResponse(error="Invalid JSON payload"))
         except error.URLError as err:
-            _json_response(self, 503, {"success": False, "error": f"Transport error: {err}"})
+            _json_response(self, 503, RunTaskErrorResponse(error=f"Transport error: {err}"))
         except Exception as err:
-            _json_response(self, 500, {"success": False, "error": f"Internal error: {err}"})
+            _json_response(self, 500, RunTaskErrorResponse(error=f"Internal error: {err}"))
 
     def log_message(self, format_str, *args):
         return
@@ -178,7 +225,7 @@ class BrowserUseRPCHandler(BaseHTTPRequestHandler):
 def main():
     host = os.getenv("BROWSER_USE_RPC_HOST", "0.0.0.0")
     port = int(os.getenv("BROWSER_USE_RPC_PORT", "8787"))
-    server = ThreadingHTTPServer((host, port), BrowserUseRPCHandler)
+    server = ThreadingHTTPServer((host, port), BrowserUseRPCHandler)  # type: ignore[arg-type]
     print(f"browser-use RPC listening on {host}:{port}")
     server.serve_forever()