diff --git a/README.md b/README.md
index f438800..fe5e136 100644
--- a/README.md
+++ b/README.md
@@ -140,7 +140,7 @@ NVIDIA offers free inference endpoints for powerful models like Moonshot AI's Ki
   - Copy the entire Python code snippet.
 - Paste the code into a chat with OpenClaw and use this prompt:
 ```
-Hey, here's the Python code from NVIDIA Build for the Kimi model (with my API key already included). 
+Hey, here's the Python code from NVIDIA Build to call an API with a model (my API key already included). 
 1. Test the code to make sure it works.
 2. Configure this model as a fallback for my primary model.
 ```
diff --git a/docker/Dockerfile b/docker/Dockerfile
index e47efe1..3413e5b 100644
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@@ -1,5 +1,5 @@
-ARG OPENCLAW_VERSION=2026.3.22
-ARG OLLAMA_VERSION=0.18.2
+ARG OPENCLAW_VERSION=2026.3.28
+ARG OLLAMA_VERSION=0.19.0
 ARG NODE_VERSION=24
 
 # Use the official Ollama image to get the binary