diff --git a/join-use-case/start-training.mdx b/join-use-case/start-training.mdx
index 7183db7..53b5bdf 100644
--- a/join-use-case/start-training.mdx
+++ b/join-use-case/start-training.mdx
@@ -36,9 +36,15 @@ Then, install requirements:
 
 ```bash
 python -m pip install --upgrade pip
-pip install tracebloc_package
+
+# Pick the extra that matches your ML framework:
+pip install "tracebloc_package[pytorch]>=0.6.33"      # most common
+# pip install "tracebloc_package[tensorflow]>=0.6.33" # TensorFlow
+# pip install "tracebloc_package[all]>=0.6.33"        # everything
 ```
 
+See the [tracebloc Package page](/tools-help/tracebloc-package) for all available extras.
+
 ## Install and Launch Jupyter Notebook
 
 Install Jupyter into your environments:
diff --git a/tools-help/tracebloc-package.mdx b/tools-help/tracebloc-package.mdx
index 367226c..564d72c 100644
--- a/tools-help/tracebloc-package.mdx
+++ b/tools-help/tracebloc-package.mdx
@@ -9,9 +9,19 @@ The `tracebloc_package` is a Python library for uploading models, linking them w
 
 ## Installation
 
-```bash
-pip install tracebloc_package>=0.6.32
-```
+ML framework dependencies are optional — pick the extra that matches your setup. The default install is ~140 MB (up from ~8 GB in older versions).
+
+| You use | Install command |
+|---|---|
+| PyTorch (most common) | `pip install "tracebloc_package[pytorch]>=0.6.33"` |
+| TensorFlow | `pip install "tracebloc_package[tensorflow]>=0.6.33"` |
+| scikit-learn only | `pip install "tracebloc_package[sklearn]>=0.6.33"` |
+| Hugging Face LoRA / PEFT (requires PyTorch) | `pip install "tracebloc_package[pytorch,huggingface]>=0.6.33"` |
+| XGBoost / CatBoost / LightGBM | `pip install "tracebloc_package[boosting]>=0.6.33"` |
+| Survival analysis | `pip install "tracebloc_package[survival]>=0.6.33"` |
+| Everything | `pip install "tracebloc_package[all]>=0.6.33"` |
+
+Upgrading from 0.6.32 or earlier? See the [migration guide](https://github.com/tracebloc/tracebloc-py-package/blob/develop/MIGRATION.md).
 
 ## Key Features