feat: new api for creating and connecting datasets/robots

README.md

@@ -103,18 +103,19 @@ nc.connect_robot(
 )
 
 # Create a dataset for recording
-nc.create_dataset(
+dataset = nc.create_dataset(
     name="My Robot Dataset",
     description="Example dataset with multiple data types"
 )
+dataset = nc.connect_dataset(dataset)
 
 # Recording and streaming data
-nc.start_recording()
+dataset.start_recording()
 t = time.time()
 nc.log_joint_positions(positions={'joint1': 0.5, 'joint2': -0.3}, timestamp=t)
 nc.log_rgb(name="top_camera", rgb=image_array, timestamp=t)
 # Stop recording, the dataset is automatically uploaded to the cloud
-nc.stop_recording()
+dataset.stop_recording()
 
 # Kick off cloud training
 job_data = nc.start_training_run(
@@ -169,4 +170,4 @@ If you use Neuracore in your research, please consider citing:
   year = {2026},
   url = {https://github.com/NeuracoreAI/neuracore}
 }
-```
+```

docs/data_daemon.md

@@ -87,11 +87,12 @@ import neuracore as nc
 
 def main():
     nc.login()
+    dataset = nc.connect_dataset("My Robot Dataset")
 
     # The daemon starts automatically when needed
-    nc.start_recording()
+    dataset.start_recording()
     # ...
-    nc.stop_recording()
+    dataset.stop_recording()
 ```
 
 Choosing a profile when using auto-start:

docs/tutorial.md

@@ -51,13 +51,14 @@ nc.connect_robot(
 import time
 
 # Create a dataset for recording
-nc.create_dataset(
+dataset = nc.create_dataset(
     name="My Robot Dataset",
     description="Example dataset with multiple data types"
 )
+dataset = nc.connect_dataset(dataset)
 
 # Start recording
-nc.start_recording()
+dataset.start_recording()
 
 # Log various data types with timestamps
 t = time.time()
@@ -80,7 +81,7 @@ custom_sensor_data = np.array([1.2, 3.4, 5.6])
 nc.log_custom_1d("force_sensor", custom_sensor_data, timestamp=t)
 
 # Stop recording
-nc.stop_recording()
+dataset.stop_recording()
 ```
 
 #### Live Data Streaming Control

examples/example_data_collection_bigym.py

@@ -33,7 +33,12 @@
 
 
 def run_episode(
-    episode_idx: int, record: bool, demo: Demo, env: Any, render: bool
+    episode_idx: int,
+    record: bool,
+    demo: Demo,
+    env: Any,
+    render: bool,
+    dataset: nc.Dataset | None = None,
 ) -> bool:
     """Run one demonstration episode and optionally record it."""
     print(f"\n=== Starting Episode {episode_idx} ===")
@@ -45,7 +50,7 @@ def run_episode(
     try:
         # Start neuracore recording if required
         if record:
-            nc.start_recording()
+            dataset.start_recording()
 
         for step in demo._steps:
             # Log current joint positions and velocities
@@ -93,10 +98,10 @@ def run_episode(
         if record:
             if success:
                 print("Episode successful → finalizing recording...")
-                nc.stop_recording(wait=True)
+                dataset.stop_recording(wait=True)
             else:
                 print("Episode failed → cancelling recording...")
-                nc.cancel_recording()
+                dataset.cancel_recording()
 
     print(f"=== Episode {episode_idx} done | success={success} ===")
     return success
@@ -120,11 +125,14 @@ def main(num_episodes: int, record: bool, recording_name: str, render: bool) ->
 
     if record:
         # Create a dataset to start recording episodes into
-        nc.create_dataset(
+        created_dataset = nc.create_dataset(
             name=recording_name,
             description="Example Bigym data collection on the ReachTarget environment",
         )
+        dataset = nc.connect_dataset(created_dataset)
         print("Created dataset.")
+    else:
+        dataset = None
 
     # Create ReachTarget Bigym environment
     env = make_env()
@@ -144,6 +152,7 @@ def main(num_episodes: int, record: bool, recording_name: str, render: bool) ->
                 demo=demos[episode_idx],
                 env=env,
                 render=render,
+                dataset=dataset,
             )
 
             if success:
@@ -152,8 +161,8 @@ def main(num_episodes: int, record: bool, recording_name: str, render: bool) ->
 
     except KeyboardInterrupt:
         print("\nInterrupted by user.")
-        if record:
-            nc.cancel_recording()
+        if dataset is not None:
+            dataset.cancel_recording()
     finally:
         env.close()
         print(

examples/example_data_collection_vx300s.py

@@ -25,12 +25,14 @@ def main(args):
     # Setup parameters
     record = args["record"]
     num_episodes = args["num_episodes"]
+    dataset = None
 
     if record:
         nc.create_dataset(
             name="My Example Dataset",
             description="This is an example dataset",
         )
+        dataset = nc.connect_dataset("My Example Dataset")
         print("Created Dataset...")
 
     try:
@@ -46,7 +48,7 @@ def main(args):
 
             # Start recording if enabled
             if record:
-                nc.start_recording()
+                dataset.start_recording()
 
             # Log initial state
             t = time.time()
@@ -83,13 +85,13 @@ def main(args):
             # Stop recording if enabled
             if record:
                 print("Finishing recording...")
-                nc.stop_recording()
+                dataset.stop_recording()
                 print("Finished recording!")
 
             print(f"Episode {episode_idx} done")
     except KeyboardInterrupt:
         if record:
-            nc.cancel_recording()
+            dataset.cancel_recording()
         raise
 
 

examples/getting_started_with_neuracore.ipynb

@@ -31,7 +31,7 @@
         "2. **Register a robot** with Neuracore by providing an MJCF model, and see it show up in your web dashboard overview.\n",
         "3. **Real-time monitoring** of your robot behavior in simulation.\n",
         "4. **Create a new dataset** in the \"Data\" page of your web dashboard.\n",
-        "5. **Record and upload data**: upload robot demonstrations to the cloud with `nc.start_recording()`.\n",
+        "5. **Record and upload data**: upload robot demonstrations to the cloud with `dataset.start_recording()`.\n",
         "6. **Visualize collected data** on the web dashboard.\n",
         "7. **Train your first policy** on your freshly collected dataset.\n",
         "8. **Deploy your policy** in simulation and track its behavior in real time."
@@ -293,6 +293,7 @@
         "    demo: Optional[Demo] = None,\n",
         "    policy=None,\n",
         "    record: bool = False,\n",
+        "    dataset=None,\n",
         ") -> bool:\n",
         "    \"\"\"Run one simulation episode and optionally record it.\n",
         "\n",
@@ -316,7 +317,7 @@
         "    try:\n",
         "        # Start a new recording to upload logs on Neuracore\n",
         "        if record:\n",
-        "            nc.start_recording()\n",
+        "            dataset.start_recording()\n",
         "\n",
         "        for step_idx in range(max_steps):\n",
         "            # Log current joint positions and velocities\n",
@@ -366,10 +367,10 @@
         "        if record:\n",
         "            if success:\n",
         "                print(\"Episode successful → finalizing recording...\")\n",
-        "                nc.stop_recording(wait=True)\n",
+        "                dataset.stop_recording(wait=True)\n",
         "            else:\n",
         "                print(\"Episode failed → cancelling recording...\")\n",
-        "                nc.cancel_recording()\n",
+        "                dataset.cancel_recording()\n",
         "\n",
         "    print(f\"=== Episode {episode_idx} done | success={success} ===\")\n",
         "    return success\n",
@@ -532,10 +533,11 @@
         "\n",
         "DATASET_NAME = \"Getting started Bigym Example\"\n",
         "\n",
-        "nc.create_dataset(\n",
+        "created_dataset = nc.create_dataset(\n",
         "    name=DATASET_NAME,\n",
         "    description=\"Data collection on the Bigym simulation environments\",\n",
         ")\n",
+        "dataset = nc.connect_dataset(created_dataset)\n",
         "print(\"Dataset created.\")"
       ]
     },
@@ -571,13 +573,14 @@
         "            episode_idx=episode_idx,\n",
         "            env=env,\n",
         "            demo=demos[episode_idx],\n",
-        "            record=RECORD\n",
+        "            record=RECORD,\n",
+        "            dataset=dataset,\n",
         "        )\n",
         "\n",
         "except KeyboardInterrupt:\n",
         "    print(\"\\nInterrupted by user.\")\n",
         "    if RECORD:\n",
-        "        nc.cancel_recording()"
+        "        dataset.cancel_recording()"
       ]
     },
     {
@@ -785,10 +788,11 @@
         "robot = nc.connect_robot(robot_name=\"Mujoco UnitreeH1 Example\")\n",
         "\n",
         "# Create a dataset for collecting episodes at inference time\n",
-        "nc.create_dataset(\n",
+        "inference_dataset = nc.create_dataset(\n",
         "    name=DATASET_NAME+\"_Inference\",\n",
         "    description=\"Inference rollouts for the Getting started Bigym example\",\n",
         ")\n",
+        "inference_dataset = nc.connect_dataset(inference_dataset)\n",
         "\n",
         "USE_PRETRAINED_MODEL = True\n",
         "if USE_PRETRAINED_MODEL:\n",
@@ -821,7 +825,8 @@
         "            episode_idx=episode_idx,\n",
         "            env=env,\n",
         "            policy=policy,\n",
-        "            record=RECORD\n",
+        "            record=RECORD,\n",
+        "            dataset=inference_dataset,\n",
         "        )\n",
         "\n",
         "        if success:\n",
@@ -832,7 +837,7 @@
         "except KeyboardInterrupt:\n",
         "    print(\"\\nInterrupted by user.\")\n",
         "    if RECORD:\n",
-        "        nc.cancel_recording()\n",
+        "        inference_dataset.cancel_recording()\n",
         "finally:\n",
         "    print(f\"\\nFinished running {NUM_EPISODES} episodes → {success_count} succeeded.\")\n",
         "    policy.disconnect()"

examples/ros_example/ros_example/simulation_node.py

@@ -64,6 +64,7 @@ def __init__(self):
         self.action_traj = []
         self.current_action_traj = []
         self.record = True
+        self.dataset = None
 
         # Initialize environment first, before setting up timers
         self.initialize_environment()
@@ -73,9 +74,10 @@ def __init__(self):
             nc.create_dataset(
                 name=dataset_name, description="ROS2 distributed data collection"
             )
+            self.dataset = nc.connect_dataset(dataset_name)
             self.get_logger().info(f"Created dataset: {dataset_name}")
 
-            nc.start_recording()
+            self.dataset.start_recording()
             self.get_logger().info("Started recording")
 
         # Create all timers in a specific order, with simulation_step last
@@ -128,7 +130,7 @@ def simulation_step(self):
                 self.obs, _, _ = self.env.step(np.array(list(action.values())))
             else:
                 if self.record:
-                    nc.stop_recording()
+                    self.dataset.stop_recording()
                     self.get_logger().info("Recording stopped")
 
                 # Increment episode counter
@@ -148,7 +150,7 @@ def simulation_step(self):
                 self.current_action_traj = self.action_traj.copy()
 
                 if self.record:
-                    nc.start_recording()
+                    self.dataset.start_recording()
                     self.get_logger().info("Recording Started")
 
         except Exception as e: