Skip to content

Commit 164dfea

Browse files
committed
Merge origin/main into fix/policy-versionless-merge
2 parents 6fc79a4 + b79b7e8 commit 164dfea

252 files changed

Lines changed: 23491 additions & 6319 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Lines changed: 59 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,59 @@
1+
---
2+
name: nemoclaw-configure-inference
3+
description: Changes the active inference model without restarting the sandbox. Use when change inference runtime, inference routing, openclaw, openshell, switch nemoclaw inference model, switch nemoclaw inference models.
4+
---
5+
6+
# Nemoclaw Configure Inference
7+
8+
Change the active inference model without restarting the sandbox.
9+
10+
## Prerequisites
11+
12+
- A running NemoClaw sandbox.
13+
- The OpenShell CLI on your `PATH`.
14+
15+
Change the active inference model while the sandbox is running.
16+
No restart is required.
17+
18+
## Step 1: Switch to a Different Model
19+
20+
Set the provider to `nvidia-nim` and specify a model from [build.nvidia.com](https://build.nvidia.com):
21+
22+
```console
23+
$ openshell inference set --provider nvidia-nim --model nvidia/nemotron-3-super-120b-a12b
24+
```
25+
26+
This requires the `NVIDIA_API_KEY` environment variable.
27+
The `nemoclaw onboard` command stores this key in `~/.nemoclaw/credentials.json` on first run.
28+
29+
## Step 2: Verify the Active Model
30+
31+
Run the status command to confirm the change:
32+
33+
```console
34+
$ nemoclaw <name> status
35+
```
36+
37+
Add the `--json` flag for machine-readable output:
38+
39+
```console
40+
$ nemoclaw <name> status --json
41+
```
42+
43+
The output includes the active provider, model, and endpoint.
44+
45+
## Step 3: Available Models
46+
47+
The following table lists the models registered with the `nvidia-nim` provider.
48+
You can switch to any of these models at runtime.
49+
50+
| Model ID | Label | Context Window | Max Output |
51+
|---|---|---|---|
52+
| `nvidia/nemotron-3-super-120b-a12b` | Nemotron 3 Super 120B | 131,072 | 8,192 |
53+
| `nvidia/llama-3.1-nemotron-ultra-253b-v1` | Nemotron Ultra 253B | 131,072 | 4,096 |
54+
| `nvidia/llama-3.3-nemotron-super-49b-v1.5` | Nemotron Super 49B v1.5 | 131,072 | 4,096 |
55+
| `nvidia/nemotron-3-nano-30b-a3b` | Nemotron 3 Nano 30B | 131,072 | 4,096 |
56+
57+
## Related Skills
58+
59+
- `nemoclaw-reference` — Inference Profiles for full profile configuration details
Lines changed: 144 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,144 @@
1+
---
2+
name: nemoclaw-deploy-remote
3+
description: Provisions a remote GPU VM with NemoClaw using Brev deployment. Also covers forwards messages between Telegram and the sandboxed OpenClaw agent. Use when deploy nemoclaw remote gpu, deployment, gpu, nemoclaw, nemoclaw brev cloud deployment, nemoclaw telegram bridge, openclaw, openshell.
4+
---
5+
6+
# Nemoclaw Deploy Remote
7+
8+
Provision a remote GPU VM with NemoClaw using Brev deployment.
9+
10+
## Prerequisites
11+
12+
- The [Brev CLI](https://brev.nvidia.com) installed and authenticated.
13+
- An NVIDIA API key from [build.nvidia.com](https://build.nvidia.com).
14+
- NemoClaw installed locally. Follow the Quickstart (see the `nemoclaw-get-started` skill) install steps.
15+
- A running NemoClaw sandbox, either local or remote.
16+
- A Telegram bot token from [BotFather](https://t.me/BotFather).
17+
18+
Run NemoClaw on a remote GPU instance through [Brev](https://brev.nvidia.com).
19+
The deploy command provisions the VM, installs dependencies, and connects you to a running sandbox.
20+
21+
## Step 1: Deploy the Instance
22+
23+
> **Warning:** The `nemoclaw deploy` command is experimental and may not work as expected.
24+
25+
Create a Brev instance and run the NemoClaw setup:
26+
27+
```console
28+
$ nemoclaw deploy <instance-name>
29+
```
30+
31+
Replace `<instance-name>` with a name for your remote instance, for example `my-gpu-box`.
32+
33+
The deploy script performs the following steps on the VM:
34+
35+
1. Installs Docker and the NVIDIA Container Toolkit if a GPU is present.
36+
2. Installs the OpenShell CLI.
37+
3. Runs the nemoclaw setup to create the gateway, register providers, and launch the sandbox.
38+
4. Starts auxiliary services, such as the Telegram bridge and cloudflared tunnel.
39+
40+
## Step 2: Connect to the Remote Sandbox
41+
42+
After deployment finishes, the deploy command opens an interactive shell inside the remote sandbox.
43+
To reconnect after closing the session, run the deploy command again:
44+
45+
```console
46+
$ nemoclaw deploy <instance-name>
47+
```
48+
49+
## Step 3: Monitor the Remote Sandbox
50+
51+
SSH to the instance and run the OpenShell TUI to monitor activity and approve network requests:
52+
53+
```console
54+
$ ssh <instance-name> 'cd /home/ubuntu/nemoclaw && set -a && . .env && set +a && openshell term'
55+
```
56+
57+
## Step 4: Verify Inference
58+
59+
Run a test agent prompt inside the remote sandbox:
60+
61+
```console
62+
$ openclaw agent --agent main --local -m "Hello from the remote sandbox" --session-id test
63+
```
64+
65+
## Step 5: GPU Configuration
66+
67+
The deploy script uses the `NEMOCLAW_GPU` environment variable to select the GPU type.
68+
The default value is `a2-highgpu-1g:nvidia-tesla-a100:1`.
69+
Set this variable before running `nemoclaw deploy` to use a different GPU configuration:
70+
71+
```console
72+
$ export NEMOCLAW_GPU="a2-highgpu-1g:nvidia-tesla-a100:2"
73+
$ nemoclaw deploy <instance-name>
74+
```
75+
76+
---
77+
78+
Forward messages between a Telegram bot and the OpenClaw agent running inside the sandbox.
79+
The Telegram bridge is an auxiliary service managed by `nemoclaw start`.
80+
81+
## Step 6: Create a Telegram Bot
82+
83+
Open Telegram and send `/newbot` to [@BotFather](https://t.me/BotFather).
84+
Follow the prompts to create a bot and receive a bot token.
85+
86+
## Step 7: Set the Environment Variable
87+
88+
Export the bot token as an environment variable:
89+
90+
```console
91+
$ export TELEGRAM_BOT_TOKEN=<your-bot-token>
92+
```
93+
94+
## Step 8: Start Auxiliary Services
95+
96+
Start the Telegram bridge and other auxiliary services:
97+
98+
```console
99+
$ nemoclaw start
100+
```
101+
102+
The `start` command launches the following services:
103+
104+
- The Telegram bridge forwards messages between Telegram and the agent.
105+
- The cloudflared tunnel provides external access to the sandbox.
106+
107+
The Telegram bridge starts only when the `TELEGRAM_BOT_TOKEN` environment variable is set.
108+
109+
## Step 9: Verify the Services
110+
111+
Check that the Telegram bridge is running:
112+
113+
```console
114+
$ nemoclaw status
115+
```
116+
117+
The output shows the status of all auxiliary services.
118+
119+
## Step 10: Send a Message
120+
121+
Open Telegram, find your bot, and send a message.
122+
The bridge forwards the message to the OpenClaw agent inside the sandbox and returns the agent response.
123+
124+
## Step 11: Restrict Access by Chat ID
125+
126+
To restrict which Telegram chats can interact with the agent, set the `ALLOWED_CHAT_IDS` environment variable to a comma-separated list of Telegram chat IDs:
127+
128+
```console
129+
$ export ALLOWED_CHAT_IDS="123456789,987654321"
130+
$ nemoclaw start
131+
```
132+
133+
## Step 12: Stop the Services
134+
135+
To stop the Telegram bridge and all other auxiliary services:
136+
137+
```console
138+
$ nemoclaw stop
139+
```
140+
141+
## Related Skills
142+
143+
- `nemoclaw-monitor-sandbox` — Monitor Sandbox Activity for sandbox monitoring tools
144+
- `nemoclaw-reference` — Commands for the full `deploy` command reference
Lines changed: 148 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,148 @@
1+
---
2+
name: nemoclaw-get-started
3+
description: Installs NemoClaw, launch a sandbox, and run your first agent prompt. Use when inference routing, install nemoclaw openclaw sandbox, nemoclaw, nemoclaw quickstart, nemoclaw quickstart install launch, openclaw, openshell, sandboxing.
4+
---
5+
6+
# Nemoclaw Get Started
7+
8+
Install NemoClaw, launch a sandbox, and run your first agent prompt.
9+
10+
> **Alpha software:** NemoClaw is in alpha, available as an early preview since March 16, 2026.
11+
> APIs, configuration schemas, and runtime behavior are subject to breaking changes between releases.
12+
> Do not use this software in production environments.
13+
> File issues and feedback through the GitHub repository as the project continues to stabilize.
14+
15+
Follow these steps to get started with NemoClaw and your first sandboxed OpenClaw agent.
16+
17+
> **Note:** NemoClaw currently requires a fresh installation of OpenClaw.
18+
19+
## Prerequisites
20+
21+
Check the prerequisites before you start to ensure you have the necessary software and hardware to run NemoClaw.
22+
23+
### Hardware
24+
25+
| Resource | Minimum | Recommended |
26+
|----------|----------------|------------------|
27+
| CPU | 4 vCPU | 4+ vCPU |
28+
| RAM | 8 GB | 16 GB |
29+
| Disk | 20 GB free | 40 GB free |
30+
31+
The sandbox image is approximately 2.4 GB compressed. During image push, the Docker daemon, k3s, and the OpenShell gateway run alongside the export pipeline, which buffers decompressed layers in memory. On machines with less than 8 GB of RAM, this combined usage can trigger the OOM killer. If you cannot add memory, configuring at least 8 GB of swap can work around the issue at the cost of slower performance.
32+
33+
#### Software
34+
35+
| Dependency | Version |
36+
|------------|----------------------------------|
37+
| Linux | Ubuntu 22.04 LTS or later |
38+
| Node.js | 20 or later |
39+
| npm | 10 or later |
40+
| Container runtime | Supported runtime installed and running |
41+
| [OpenShell](https://github.com/NVIDIA/OpenShell) | Installed |
42+
43+
#### Container Runtime Support
44+
45+
| Platform | Supported runtimes | Notes |
46+
|----------|--------------------|-------|
47+
| Linux | Docker | Primary supported path today |
48+
| macOS (Apple Silicon) | Colima, Docker Desktop | Recommended runtimes for supported macOS setups |
49+
| macOS | Podman | Not supported yet. NemoClaw currently depends on OpenShell support for Podman on macOS. |
50+
| Windows WSL | Docker Desktop (WSL backend) | Supported target path |
51+
52+
> **💡 Tip**
53+
>
54+
> For DGX Spark, follow the [DGX Spark setup guide](https://github.com/NVIDIA/NemoClaw/blob/main/spark-install.md). It covers Spark-specific prerequisites, such as cgroup v2 and Docker configuration, before running the standard installer.
55+
56+
### Install NemoClaw and Onboard OpenClaw Agent
57+
58+
Download and run the installer script.
59+
The script installs Node.js if it is not already present, then runs the guided onboard wizard to create a sandbox, configure inference, and apply security policies.
60+
61+
```bash
62+
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
63+
```
64+
65+
If you use nvm or fnm to manage Node.js, the installer may not update your current shell's PATH.
66+
If `nemoclaw` is not found after install, run `source ~/.bashrc` (or `source ~/.zshrc` for zsh) or open a new terminal.
67+
68+
When the install completes, a summary confirms the running environment:
69+
70+
```text
71+
──────────────────────────────────────────────────
72+
Sandbox my-assistant (Landlock + seccomp + netns)
73+
Model nvidia/nemotron-3-super-120b-a12b (NVIDIA Endpoint API)
74+
──────────────────────────────────────────────────
75+
Run: nemoclaw my-assistant connect
76+
Status: nemoclaw my-assistant status
77+
Logs: nemoclaw my-assistant logs --follow
78+
──────────────────────────────────────────────────
79+
80+
[INFO] === Installation complete ===
81+
```
82+
83+
### Chat with the Agent
84+
85+
Connect to the sandbox, then chat with the agent through the TUI or the CLI.
86+
87+
#### Connect to the Sandbox
88+
89+
Run the following command to connect to the sandbox:
90+
91+
```bash
92+
nemoclaw my-assistant connect
93+
```
94+
95+
This connects you to the sandbox shell `sandbox@my-assistant:~$` where you can run `openclaw` commands.
96+
97+
#### OpenClaw TUI
98+
99+
In the sandbox shell, run the following command to open the OpenClaw TUI, which opens an interactive chat interface.
100+
101+
```bash
102+
openclaw tui
103+
```
104+
105+
Send a test message to the agent and verify you receive a response.
106+
107+
> **ℹ️ Note**
108+
>
109+
> The TUI is best for interactive back-and-forth. If you need the full text of a long response such as a large code generation output, use the CLI instead.
110+
111+
#### OpenClaw CLI
112+
113+
In the sandbox shell, run the following command to send a single message and print the response:
114+
115+
```bash
116+
openclaw agent --agent main --local -m "hello" --session-id test
117+
```
118+
119+
This prints the complete response directly in the terminal and avoids relying on the TUI view for long output.
120+
121+
### Uninstall
122+
123+
To remove NemoClaw and all resources created during setup, in the terminal outside the sandbox, run:
124+
125+
```bash
126+
curl -fsSL https://raw.githubusercontent.com/NVIDIA/NemoClaw/refs/heads/main/uninstall.sh | bash
127+
```
128+
129+
The script removes sandboxes, the NemoClaw gateway and providers, related Docker images and containers, local state directories, and the global `nemoclaw` npm package. It does not remove shared system tooling such as Docker, Node.js, npm, or Ollama.
130+
131+
| Flag | Effect |
132+
|--------------------|-----------------------------------------------------|
133+
| `--yes` | Skip the confirmation prompt. |
134+
| `--keep-openshell` | Leave the `openshell` binary installed. |
135+
| `--delete-models` | Also remove NemoClaw-pulled Ollama models. |
136+
137+
For example, to skip the confirmation prompt:
138+
139+
```bash
140+
curl -fsSL https://raw.githubusercontent.com/NVIDIA/NemoClaw/refs/heads/main/uninstall.sh | bash -s -- --yes
141+
```
142+
143+
## Related Skills
144+
145+
- `nemoclaw-configure-inference` — Switch inference providers to use a different model or endpoint
146+
- `nemoclaw-manage-policy` — Approve or deny network requests when the agent tries to reach external hosts
147+
- `nemoclaw-deploy-remote` — Deploy to a remote GPU instance for always-on operation
148+
- `nemoclaw-monitor-sandbox` — Monitor sandbox activity through the OpenShell TUI

0 commit comments

Comments
 (0)