Add possibility to pass on_artifacts for a specific conversation (lan…

…gchain-ai#12687)

Possibility to pass on_artifacts to a conversation. It can be then
achieved by adding this way:

```python
result = agent.run(
    input=message.text,
    metadata={
        "on_artifact": CALLBACK_FUNCTION
    },
)
```

libs/langchain/langchain/tools/e2b_data_analysis/tool.py

@@ -9,6 +9,7 @@
 
 from langchain.callbacks.manager import (
     AsyncCallbackManagerForToolRun,
+    CallbackManager,
     CallbackManagerForToolRun,
 )
 from langchain.pydantic_v1 import BaseModel, Field, PrivateAttr
@@ -151,14 +152,26 @@ def uploaded_files_description(self) -> str:
         return "\n".join(lines)
 
     def _run(
-        self, python_code: str, run_manager: Optional[CallbackManagerForToolRun] = None
+        self,
+        python_code: str,
+        run_manager: Optional[CallbackManagerForToolRun] = None,
+        callbacks: Optional[CallbackManager] = None,
     ) -> str:
         python_code = add_last_line_print(python_code)
-        stdout, stderr, _ = self.session.run_python(python_code)
+
+        if callbacks is not None:
+            on_artifact = getattr(callbacks.metadata, "on_artifact", None)
+        else:
+            on_artifact = None
+
+        stdout, stderr, artifacts = self.session.run_python(
+            python_code, on_artifact=on_artifact
+        )
 
         out = {
             "stdout": stdout,
             "stderr": stderr,
+            "artifacts": list(map(lambda artifact: artifact.name, artifacts)),
         }
         return json.dumps(out)