init blog for model booster controller

Signed-off-by: HunterChen <[email protected]>

Author: huntersman

docs/kthena/blog/model-controller-manager/index.md

@@ -0,0 +1,86 @@
+---
+slug: model-booster-controller-blog-post
+title: All you should know about model-booster controller
+authors: [ HunterChen ]
+tags: [ ]
+---
+
+# All you should know about model-booster controller
+
+## 1. Brief Introduction
+
+Model-booster controller is a k8s [operator](https://kubernetes.io/docs/concepts/extend-kubernetes/operator/) that helps
+you manage your model deployments. It automates the deployment of models in a Kubernetes
+cluster. If you don't use model-booster to manager models in Kthena, you need to create `ModelServing`,
+`AutoscalingPolicy`,`AutoscalingPolicyBinding`, `ModelRoute`, `ModelServer` one by one.
+Model-booster controller will help you create these resources automatically based on the `ModelBooster` custom resource.
+
+There are three kinds of conditions for `ModelBooster` custom resource:
+
+- `Initialized`: The `ModelBooster` CR passed validation and is processing.
+- `Active`: Everything is ok, now you can use the model.
+- `Failed`: There is something wrong, check the message for details.
+
+We use `true`, `false` and `unknown` to represent the status of each condition. `true` means the condition is met,
+`false` means the
+condition is not met, and `unknown` means we don't know for yet.
+
+In kthena version v0.1.0, we only support `vLLM` as the model backend. In the future, we will support `SGLang`, `MindIE`.
+
+## 2. Core Features
+
+### 2.1 Automated Deployment
+
+With model-booster controller, you can create/update/delete models with a single
+`ModelBooster` custom resource, the controller will create/update/delete the corresponding `ModelServing`,
+`AutoscalingPolicy`,`AutoscalingPolicyBinding`, `ModelRoute`, `ModelServer` resources automatically.
+
+`ModelBooster` supports models from different sources, including:
+- `HuggingFace`(recommend): Models hosted on HuggingFace.
+- `S3`: Models stored in S3-compatible storage.
+- `PVC`: Models stored in Persistent Volume Claims.
+
+When you first deploy a model using `ModelBooster`, the controller will download the model from the sources into the `cache` you specified, next time when you deploy the same model, the controller will use the cached model to speed up the deployment.
+
+### 2.2 Validation
+
+The webhook and the controller will validate the `ModelBooster` custom resource before
+creating/updating/deleting the corresponding resources to ensure that the resource is valid.
+
+You can find the validations of CR in
+the [CRD Reference](https://kthena.volcano.sh/docs/reference/crd/workload.serving.volcano.sh#modelbooster)
+
+### 2.3 Reconciliation
+
+The controller will continuously monitor the state of the `ModelBooster` custom resource and the
+corresponding resources, and will take action to ensure that the desired state is always maintained. And when all the
+resources are ready, the controller will update the condition of the `ModelBooster` custom resource to `Active`.
+
+### 2.4 Best Practice templates
+
+We provide some best practice templates for `ModelBooster` custom resource to help you get started quickly. Including:
+- DeepSeek-R1
+- DeepSeek-R1-Distill-Qwen-7B
+- DeepSeek-R1-Distill-Qwen-32B
+- gemma-2-2b-it
+- gemma-2-27b-it
+- gemma-3-4b-it
+- gemma-3-27b-it
+- Llama-3.2-1B-Instruct
+- Llama-3.3-70B-Instruct
+- Llama-4-Maverick-17B-128E-Instruct-FP8
+- Llama-4-Scout-17B-16E-Instruct
+- Meta-Llama-3-8B
+- Mistral-Small-24B-Instruct-2501
+- Mixtral-8x7B-Instruct-v0.1
+- Mixtral-8x22B-Instruct-v0.1
+- gpt-oss-20b
+- gpt-oss-120b
+- Qwen3-32B
+
+For more details, please refer to the [CLI documentation](/cli/kthena/README.md).
+
+## 3. Conclusion
+
+Model-booster controller simplifies the management of model deployments in Kthena. It automates the creation and
+management of multiple resources, ensuring that your models are always deployed and running as expected.