Remove roadmap doc and fix embedder model deployment issue

ROADMAP.md

@@ -1,67 +1,3 @@
 ## Roadmap of KubeAGI
 
-### v0.1.0 - 2023 Q4 Released
-
-* Dataset Management - manage data, including local files, integrate with object storage(s3), data editing, version control, and file download
-* Data Processing - data cleaning, text splitting (e.g., text segmentation, QA splitting), file labeling
-* Knowledge Base - data embedding
-* Model Management - manage the lifecycle of models.
-* Model Serving
-  - Support CPU & GPU Model Serving
-  - Support both remote and local model inference services, and associate with the knowledge base
-  - Support local embedding service (bge, m3e)
-  - Support vLLM inference engine
-* LLM Applications - prompt engineering, initial implementation of LLM application orchestration capabilities. Manage and orchestrate Prompt, LLM/Retriever Chain nodes, and provide relevant example applications (based on streamlit)
-* Guided walkthroughs and example scenarios - let the user get started to build LLM application quickly, add momre built-in chat example applications
-
-### v0.2.0 - 2024 Feb. Ongoing
-* Support evaluation of Prompts under different LLMs and generate test reports.
-* RAG evaluation and RAG Question Generation
-  - Optimize question generation, analyze question quality, filter out low-similarity questions
-  - Evaluation metrics: retrieval evaluation - Hit Rate, MRR; answer evaluation - fairness, relevance, consistency, etc
-  - Other evaluation capabilities
-
-* Data lineage - Understand the origin and flow of data, e.g., support mapping between answers and original documents
-* Perform similarity analysis on QA pairs generated by large models, allowing manual processing (deletion, merging, etc.) by users
-* Playground for datasets, knowledge base, model services, etc., based on streamlit.
-* LLM applications support Get/Post API Chain, enabling typical LLM application development (non-workflow mode)
-* Visualization of various data types, based on streamlit
-
-### v0.3.0 - 2024 Mar.
-* Data Processing - Introduce text annotation (automated + manual) to improve data quality through assisted fine-tuning
-
-* Data Security - Support data anonymization (e.g., masking sensitive information like ID numbers, phone numbers, and bank account numbers)
-
-* Enhanced Data Integration - Increase the capability to integrate with various data sources (databases, APIs, etc.) and support data synchronization strategies (automatic synchronization)
-
-* Support manual evaluation to ensure quality control before deploying to production. Additionally, incorporate manual feedback into the monitoring system
-
-* Enable user feedback on the question-answering system to facilitate optimization of LLM applications (data processing, prompt optimization, etc.)
-
-* Integration of GPU management, scheduling, and resource monitoring capabilities for containerized environments
-
-* Integration of API gateway to govern model service APIs, including monitoring, analysis, and security measures, and construct AI gateway
-
-
-### v0.5.0 - 2024 Apr.
-* Support low-resource large model fine-tuning, including RLHF (Reinforcement Learning from Human Feedback), SFT (Semi-Supervised Fine-Tuning) techniques such as Adapter, P-tuning, and LoRA. This improves model quality while reducing performance requirements for model serving (e.g., reducing inference costs, latency issues related to long prompts or slow inference)
-* Model compression techniques
-* Conduct testing and evaluation of model services and embeddings (QA evaluation, metric collection)
-
-* Implement "scale to zero" capability (integrating with Arbiter) for cold start scenarios, enabling models and applications to evolve towards a Serverless architecture
-
-* Support orchestration of additional node types such as Agent, Cache, etc
-
-* Add more best practices for prompt engineering
-  - Few-shot learning techniques
-  - Chain-of-Thought (CoT) approach
-  - Mind-mapping techniques
-
-### v1.0 - 2024 Jun.
-* Automatically constructing prompt templates based on data annotations
-* Enhance the monitoring capabilities of LLMOps, monitoring the pipeline from dataset and feature data to model inference, with call chain tracing based on langchain-go
-* Implement a pipeline from data source -> dataset -> data processing -> data versioning -> knowledge base -> model service
-* Strengthen the Python SDK to handle basic capabilities such as dataset manipulation, data processing, and vectorization. These operations can be performed in a notebook environment.
-  - Refer Databricks to enhance the developer experience
-* Implement gray release for LLM applications based on AI gateway
-
+Refer to our [online documentation](http://kubeagi.k8s.com.cn/docs/Release&Plan/roadmap)

deploy/charts/arcadia/crds/arcadia.kubeagi.k8s.com.cn_workers.yaml

@@ -42,6 +42,115 @@ spec:
           spec:
             description: WorkerSpec defines the desired state of Worker
             properties:
+              additionalEnvs:
+                description: Additional env to use
+                items:
+                  description: EnvVar represents an environment variable present in
+                    a Container.
+                  properties:
+                    name:
+                      description: Name of the environment variable. Must be a C_IDENTIFIER.
+                      type: string
+                    value:
+                      description: 'Variable references $(VAR_NAME) are expanded using
+                        the previously defined environment variables in the container
+                        and any service environment variables. If a variable cannot
+                        be resolved, the reference in the input string will be unchanged.
+                        Double $$ are reduced to a single $, which allows for escaping
+                        the $(VAR_NAME) syntax: i.e. "$$(VAR_NAME)" will produce the
+                        string literal "$(VAR_NAME)". Escaped references will never
+                        be expanded, regardless of whether the variable exists or
+                        not. Defaults to "".'
+                      type: string
+                    valueFrom:
+                      description: Source for the environment variable's value. Cannot
+                        be used if value is not empty.
+                      properties:
+                        configMapKeyRef:
+                          description: Selects a key of a ConfigMap.
+                          properties:
+                            key:
+                              description: The key to select.
+                              type: string
+                            name:
+                              description: 'Name of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#names
+                                TODO: Add other useful fields. apiVersion, kind, uid?'
+                              type: string
+                            optional:
+                              description: Specify whether the ConfigMap or its key
+                                must be defined
+                              type: boolean
+                          required:
+                          - key
+                          type: object
+                          x-kubernetes-map-type: atomic
+                        fieldRef:
+                          description: 'Selects a field of the pod: supports metadata.name,
+                            metadata.namespace, `metadata.labels[''<KEY>'']`, `metadata.annotations[''<KEY>'']`,
+                            spec.nodeName, spec.serviceAccountName, status.hostIP,
+                            status.podIP, status.podIPs.'
+                          properties:
+                            apiVersion:
+                              description: Version of the schema the FieldPath is
+                                written in terms of, defaults to "v1".
+                              type: string
+                            fieldPath:
+                              description: Path of the field to select in the specified
+                                API version.
+                              type: string
+                          required:
+                          - fieldPath
+                          type: object
+                          x-kubernetes-map-type: atomic
+                        resourceFieldRef:
+                          description: 'Selects a resource of the container: only
+                            resources limits and requests (limits.cpu, limits.memory,
+                            limits.ephemeral-storage, requests.cpu, requests.memory
+                            and requests.ephemeral-storage) are currently supported.'
+                          properties:
+                            containerName:
+                              description: 'Container name: required for volumes,
+                                optional for env vars'
+                              type: string
+                            divisor:
+                              anyOf:
+                              - type: integer
+                              - type: string
+                              description: Specifies the output format of the exposed
+                                resources, defaults to "1"
+                              pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                              x-kubernetes-int-or-string: true
+                            resource:
+                              description: 'Required: resource to select'
+                              type: string
+                          required:
+                          - resource
+                          type: object
+                          x-kubernetes-map-type: atomic
+                        secretKeyRef:
+                          description: Selects a key of a secret in the pod's namespace
+                          properties:
+                            key:
+                              description: The key of the secret to select from.  Must
+                                be a valid secret key.
+                              type: string
+                            name:
+                              description: 'Name of the referent. More info: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#names
+                                TODO: Add other useful fields. apiVersion, kind, uid?'
+                              type: string
+                            optional:
+                              description: Specify whether the Secret or its key must
+                                be defined
+                              type: boolean
+                          required:
+                          - key
+                          type: object
+                          x-kubernetes-map-type: atomic
+                      type: object
+                  required:
+                  - name
+                  type: object
+                type: array
               creator:
                 description: Creator defines datasource creator (AUTO-FILLED by webhook)
                 type: string
@@ -51,6 +160,35 @@ spec:
               displayName:
                 description: DisplayName defines datasource display name
                 type: string
+              matchExpressions:
+                description: NodeSelectorRequirement to schedule this worker
+                items:
+                  description: A node selector requirement is a selector that contains
+                    values, a key, and an operator that relates the key and values.
+                  properties:
+                    key:
+                      description: The label key that the selector applies to.
+                      type: string
+                    operator:
+                      description: Represents a key's relationship to a set of values.
+                        Valid operators are In, NotIn, Exists, DoesNotExist. Gt, and
+                        Lt.
+                      type: string
+                    values:
+                      description: An array of string values. If the operator is In
+                        or NotIn, the values array must be non-empty. If the operator
+                        is Exists or DoesNotExist, the values array must be empty.
+                        If the operator is Gt or Lt, the values array must have a
+                        single element, which will be interpreted as an integer. This
+                        array is replaced during a strategic merge patch.
+                      items:
+                        type: string
+                      type: array
+                  required:
+                  - key
+                  - operator
+                  type: object
+                type: array
               model:
                 description: Model this worker wants to use
                 properties:

pkg/worker/devices.go

@@ -43,8 +43,8 @@ const (
 
 // DeviceBasedOnResource returns the device type based on the resource list
 func DeviceBasedOnResource(resource corev1.ResourceList) Device {
-	_, ok := resource[ResourceNvidiaGPU]
-	if ok {
+	value, ok := resource[ResourceNvidiaGPU]
+	if ok && value.Value() > 0 {
 		return CUDA
 	}
 	return CPU

pkg/worker/worker.go

@@ -257,7 +257,7 @@ func (podWorker *PodWorker) Model() *arcadiav1alpha1.Model {
 }
 
 // BeforeStart will create resources which are related to this Worker
-// Now we have a pvc(if configured),service,LLM(if a llm model),Embedder(if a embedding model)
+// Now we have a pvc(if configured), service, LLM(if a llm model), Embedder(if a embedding model)
 func (podWorker *PodWorker) BeforeStart(ctx context.Context) error {
 	var err error
 	// If the local directory is mounted, there is no need to create the pvc