Simon-Liu/gemma-4-e4b-kubectl-mcp-server-mcp-sft-it
Text Generation • 8B • Updated • 33
Datasets:
id stringlengths 32 32 | query stringlengths 38 198 | query_zhtw stringlengths 16 104 | tools stringclasses 296
values | think stringlengths 46 295 | answer stringlengths 40 313 | messages listlengths 3 3 |
|---|---|---|---|---|---|---|
bc70782e7f20422ba0d73dd1ab91418f | Check the current replica count for the identity-service in the auth namespace and scale it up to 5. | 幫我確認一下 auth namespace 裡面的 identity-service 目前有幾個副本，然後把它擴充到 5 個。 | [{"name": "get_deployments", "description": "Get all deployments in the specified namespace.\n\nArgs:\n namespace: Namespace to list deployments from (all namespaces if not specified)\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "pr... | 為了確認 identity-service 的副本數，我需要先列出 auth namespace 下所有的 deployments 來查看現狀。確認後，再使用 scale_deployment 工具將該服務的副本數調整為 5。 | [{"name": "get_deployments", "arguments": {"namespace": "auth"}}, {"name": "scale_deployment", "arguments": {"name": "identity-service", "replicas": 5, "namespace": "auth"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_deploy... |
c182ef45db6a4686bbf1872cee624a5f | List all pods that are currently pending in the backend namespace and explain why they won't start. | 列出現在 backend namespace 中狀態為 pending 的 pods，順便幫我解釋一下為什麼它們啟動不了。 | [{"name": "get_pods", "description": "Get all pods in the specified namespace.\n\nArgs:\n namespace: Namespace to list pods from (all namespaces if not specified)\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "properties": {"namespac... | 使用者想要查看 backend 這個 namespace 中處於 pending 狀態的 pods，並了解它們無法啟動的原因。我可以直接使用 detect_pending_pods 工具，它不僅能篩選出 pending 的 pods，還會自動分析無法排程的原因，完全符合使用者的需求。 | [{"name": "detect_pending_pods", "arguments": {"namespace": "backend"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_pods\"... |
719fbb84c1444ff3b6aa2e2a8f5970d1 | Can you list all the nodes in the cluster and give me a summary of their current health status? | 可以列出叢集裡所有的節點，並給我一份目前的健康狀態摘要嗎？ | [{"name": "get_nodes", "description": "Get all nodes in the cluster.\n\nArgs:\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "properties": {"context": {"default": "", "type": "string"}}, "type": "object"}, "read_only": true}, {"name": "g... | 為了滿足使用者的需求，我需要先獲取叢集中所有節點的詳細列表，接著再取得一份節點的健康狀態摘要。這兩個需求分別對應到 get_nodes 和 get_nodes_summary 這兩個工具，我將會依序呼叫它們來完成任務。 | [{"name": "get_nodes", "arguments": {"context": ""}}, {"name": "get_nodes_summary", "arguments": {"context": ""}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_nodes\... |
38149b3fef11463cbbf3dc2481180b5f | Why is the cart-api pod in the production namespace stuck in a CrashLoopBackOff? | 為什麼 production namespace 底下的 cart-api pod 一直處於 CrashLoopBackOff 重啟狀態？ | [{"name": "diagnose_pod_crash", "description": "Automated diagnosis of pod crash loops and failures.\n\nArgs:\n pod_name: Name of the pod\n namespace: Namespace of the pod\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "properties"... | 使用者詢問 production namespace 下的 cart-api pod 發生 CrashLoopBackOff 的原因。為了找出根本問題，我應該先使用 diagnose_pod_crash 工具進行自動化診斷，同時獲取該 pod 的日誌以分析錯誤訊息。 | [{"name": "diagnose_pod_crash", "arguments": {"pod_name": "cart-api", "namespace": "production"}}, {"name": "get_logs", "arguments": {"pod_name": "cart-api", "namespace": "production"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"diagnose_p... |
c130fad6237c474196f57c0b8e6c4b92 | I need to see the values file used for the nginx-ingress release in the ingress-nginx namespace. | 我想要看一下 ingress-nginx 這個 namespace 裡面 nginx-ingress 的部署用了哪些設定檔（values）。 | [{"name": "helm_get_values", "description": "Get the values used for a Helm release.\n\nArgs:\n release_name: Name of the release\n namespace: Kubernetes namespace\n all_values: Include computed (default + user) values\n revision: Get values for a specific revision\n context: Kubernetes context to use (o... | 使用者想要查詢位於 'ingress-nginx' 命名空間下，名為 'nginx-ingress' 的 Helm release 所使用的設定值。我可以使用 'helm_get_values' 工具，並將 'release_name' 設定為 'nginx-ingress'，'namespace' 設定為 'ingress-nginx' 來取得這些資訊。 | [{"name": "helm_get_values", "arguments": {"release_name": "nginx-ingress", "namespace": "ingress-nginx"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"helm_get_v... |
4e01021afad449819b553203e26d59f3 | The payment-gateway service seems slow. Can you trace the connection chain from the service to its backing pods? | payment-gateway 服務好像有點慢，可以幫我追蹤一下從這個 service 到後端 pod 的連線鏈路嗎？ | [{"name": "trace_service_chain", "description": "Trace the connection chain from service to pods.\n\nArgs:\n service_name: Name of the service to trace\n namespace: Namespace of the service\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": f... | 使用者反映 payment-gateway 服務回應緩慢，需要檢查該服務到後端 Pod 的連線鏈路。我可以使用 trace_service_chain 工具來追蹤該服務的網路路徑，確認是否有異常。 | [{"name": "trace_service_chain", "arguments": {"service_name": "payment-gateway"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"trace_serv... |
af94428b30a5436faf66beb69fe33539 | Search for any installed Helm charts related to redis in the repo and give me the latest version info. | 幫我在 repo 裡搜尋一下有沒有跟 redis 相關的 Helm charts，順便告訴我最新的版本號。 | [{"name": "helm_search_repo", "description": "Search for charts in configured Helm repositories.\n\nArgs:\n keyword: Search keyword\n regexp: Use regular expression for searching\n versions: Show all chart versions (not just latest)\n version: Version constraint (e.g., '>1.0.0')\n max_results: Maximum nu... | 使用者想要搜尋與 redis 相關的 Helm charts，並且需要確認最新的版本號。我可以使用 helm_search_repo 工具，將關鍵字設為 'redis'，並將 versions 參數設為 true，這樣就能列出所有版本，方便確認最新的版本資訊。 | [{"name": "helm_search_repo", "arguments": {"keyword": "redis", "versions": true}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"helm_searc... |
011dc020b25f4d6b9fb41107442f0b43 | Check the resource usage metrics for the payment-processor pod in the checkout namespace. | 幫我查一下 checkout namespace 裡面的 payment-processor pod 目前的資源耗用數據。 | [{"name": "get_pod_metrics", "description": "Get pod resource usage metrics (requires metrics-server).\n\nArgs:\n namespace: Target namespace (optional, all namespaces if not specified)\n pod_name: Filter by specific pod name (optional)\n context: Kubernetes context to use (optional, uses current context if no... | 使用者想要查詢 checkout 命名空間中 payment-processor 這個 Pod 的資源耗用情況，我可以使用 get_pod_metrics 工具，並將 namespace 設為 'checkout'，pod_name 設為 'payment-processor' 來直接獲取該 Pod 的數據。 | [{"name": "get_pod_metrics", "arguments": {"namespace": "checkout", "pod_name": "payment-processor"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_pod_me... |
86252658863d419a810a777cbff1cbc7 | I want to verify if the cert-manager certificates in the secure-ingress namespace are valid. | 我想要確認一下 secure-ingress namespace 裡面的 cert-manager 憑證是不是都還有效。 | [{"name": "list_certs", "description": "List cert-manager certificates with status.", "input_schema": {"additionalProperties": false, "properties": {"namespace": {"default": "", "type": "string"}, "context": {"default": "", "type": "string"}, "label_selector": {"default": "", "type": "string"}}, "type": "object"}, "rea... | 為了確認 secure-ingress 命名空間下的 cert-manager 憑證狀態，我需要列出該命名空間中所有的憑證資源，以便檢查它們是否有效。 | [{"name": "list_certs", "arguments": {"namespace": "secure-ingress"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"list_certs... |
47a798febab3405b9d3ec017b13aee94 | Get the rollout history for the web-dashboard deployment in the frontend namespace to see what happened. | 我想查一下 frontend namespace 裡 web-dashboard deployment 的發布歷史，看看之前更新發生了什麼事。 | [{"name": "kubectl_rollout", "description": "Manage rollouts (restart, status, history, undo, pause, resume).\n\nArgs:\n action: Rollout action (status, history, restart, undo, pause, resume)\n resource_type: Type of resource (deployment, statefulset, daemonset)\n name: Name of the resource\n namespace: Tar... | 使用者想要查看特定命名空間中某個 deployment 的發布歷史，這可以直接透過 kubectl_rollout 工具的 history 動作來達成。我需要指定 resource_type 為 'deployment'，name 為 'web-dashboard'，namespace 為 'frontend'，並將 action 設定為 'history'。 | [{"name": "kubectl_rollout", "arguments": {"action": "history", "resource_type": "deployment", "name": "web-dashboard", "namespace": "frontend"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"kubectl_ro... |
5b3c3084f0f9472294cd0d8592b9f4b7 | Show me the logs of the previous container instance for the user-profile-service pod before it crashed. | user-profile-service 這個 pod 剛才掛掉了，可以給我看它在崩潰前的上一個 container 實例的 log 嗎？ | [{"name": "get_previous_logs", "description": "Get logs from the previous container instance (useful for crash debugging).\n\nArgs:\n pod_name: Name of the pod\n namespace: Namespace of the pod\n container: Container name (for multi-container pods)\n tail: Number of lines to show from end of logs\n conte... | 使用者想要查看 user-profile-service 這個 Pod 在崩潰前的日誌，這需要調用 get_previous_logs 工具。我需要指定 pod_name 為 'user-profile-service'，並使用預設的 namespace 'default' 來獲取該 Pod 的歷史日誌。 | [{"name": "get_previous_logs", "arguments": {"pod_name": "user-profile-service", "namespace": "default"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_previo... |
7eb85c1605ea4f19a88259ae526d7252 | Execute a 'curl -v' command inside the order-worker-0 pod to test connectivity to the internal database. | 幫我在 order-worker-0 這個 pod 裡面執行一次 curl -v 指令，測試一下能不能連到內部資料庫。 | [{"name": "exec_in_pod", "description": "Execute a command inside a pod.\n\nArgs:\n pod_name: Name of the pod\n command: Command to execute\n namespace: Namespace of the pod\n container: Container name (for multi-container pods)\n context: Kubernetes context to use (uses current context if not specified)... | 使用者想要在名為 order-worker-0 的 Pod 中執行 curl -v 指令來測試資料庫連線。我可以使用 exec_in_pod 工具，將 pod_name 設定為 'order-worker-0'，並將 command 設定為 'curl -v' 來達成這個需求。 | [{"name": "exec_in_pod", "arguments": {"pod_name": "order-worker-0", "command": "curl -v"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"exec_in_po... |
06da988019af498588481a68f5a19a87 | Identify all the Custom Resource Definitions currently available in the cluster and tell me if we have any monitoring CRDs. | 列出現在叢集裡所有的 CRD，順便幫我看看有沒有跟監控系統（monitoring）相關的定義。 | [{"name": "list_crds", "description": "List all Custom Resource Definitions in the cluster.\n\nArgs:\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "properties": {"context": {"default": "", "type": "string"}}, "type": "object"}, "read_on... | 為了完成使用者的需求，我需要執行兩個步驟：首先列出叢集內所有的 CRD 以滿足「列出所有 CRD」的要求，接著利用分類功能查詢「monitoring」類別的 CRD，以確認是否有監控系統相關的定義。 | [{"name": "list_crds", "arguments": {"context": ""}}, {"name": "discover_crds", "arguments": {"category": "monitoring", "context": ""}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"list_crds\... |
6034aeadd5ee448ba0c19a0164b8798e | Create a new deployment named 'debug-tool' using the 'alpine:latest' image in the default namespace with 1 replica. | 在 default namespace 建立一個名稱為 debug-tool 的 deployment，使用 alpine:latest 映像檔，數量設為 1 個。 | [{"name": "create_deployment", "description": "Create a new deployment.\n\nArgs:\n name: Name of the deployment\n image: Container image to deploy\n replicas: Number of replicas\n namespace: Namespace to create deployment in (default: \"default\")\n context: Kubernetes context to use (uses current contex... | 使用者想要在 Kubernetes 的 default namespace 中建立一個名為 debug-tool 的 deployment，使用的映像檔為 alpine:latest，且副本數量為 1。這可以直接使用 create_deployment 工具來完成，將對應參數填入即可。 | [{"name": "create_deployment", "arguments": {"name": "debug-tool", "image": "alpine:latest", "replicas": 1, "namespace": "default"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"create_dep... |
de26daaf33c74aba90999e727fd0371a | Find any underutilized pods in the analytics namespace that use less than 10% of their requested CPU. | 幫我找一下 analytics namespace 裡面有沒有資源配置過剩的 pod，CPU 使用率低於 10% 的那種。 | [{"name": "get_idle_resources", "description": "Find underutilized pods using less than threshold percentage of requested resources.\n\nArgs:\n namespace: Target namespace (optional, all namespaces if not specified)\n cpu_threshold: CPU usage threshold percentage\n memory_threshold: Memory usage threshold perc... | 使用者想要檢查 analytics 命名空間中資源配置過剩的 Pod，特別是 CPU 使用率低於 10% 的情況。我可以使用 get_idle_resources 工具，將 namespace 設定為 'analytics'，並將 cpu_threshold 設定為 10.0 來篩選出符合條件的 Pod。 | [{"name": "get_idle_resources", "arguments": {"namespace": "analytics", "cpu_threshold": 10.0}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_idle_r... |
1e77f36574104746a32b57947b515d55 | List all configured Helm repositories and update their indexes so we have the latest chart versions. | 列出目前設定好的 Helm repositories，並把它們的 index 更新一下，確保我們有抓到最新的 chart 版本。 | [{"name": "helm_repo_list", "description": "List all configured Helm repositories.", "input_schema": {"additionalProperties": false, "properties": {}, "type": "object"}, "read_only": true}, {"name": "helm_repo_update", "description": "Update Helm repository indexes.\n\nArgs:\n repos: Specific repositories to update ... | 為了完成使用者的要求，我首先需要查看目前系統中已經設定好的 Helm repositories，接著再執行更新操作，以確保所有儲存庫的索引都是最新的。 | [{"name": "helm_repo_list", "arguments": {}}, {"name": "helm_repo_update", "arguments": {"repos": null}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"helm_repo_... |
29f1e7025e924fba8c2a2044bad41ef8 | Check if there are any pending policy violations in the security namespace reported by Kyverno. | 幫我檢查一下 security namespace 裡有沒有 Kyverno 報出來的尚未處理的政策違規。 | [{"name": "get_policy_violations", "description": "List policy violations from PolicyReports or Gatekeeper.", "input_schema": {"additionalProperties": false, "properties": {"namespace": {"default": "", "type": "string"}, "context": {"default": "", "type": "string"}, "engine": {"default": "all", "type": "string"}, "seve... | 使用者想要檢查 security namespace 中的政策違規，我可以使用 get_policy_violations 工具，並將 namespace 參數設定為 'security' 來過濾出該命名空間下的違規事項。 | [{"name": "get_policy_violations", "arguments": {"namespace": "security"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_policy... |
56bf1fe5763743d586a3783484e09fc5 | Diagnose why the 'database-migration' job in the migration namespace has failed. | migration namespace 裡面的 database-migration 這個 job 失敗了，幫我診斷一下原因。 | [{"name": "get_jobs", "description": "Get Jobs and optionally CronJobs.\n\nArgs:\n namespace: Namespace to list Jobs from (all namespaces if not specified)\n include_cronjobs: Whether to include CronJobs in the result\n context: Kubernetes context to use (uses current context if not specified)", "input_schema"... | 為了診斷 migration namespace 下 database-migration 這個 Job 失敗的原因，我需要先確認該 Job 對應的 Pod 名稱，接著查看該 Pod 的事件紀錄與日誌，以便找出具體的錯誤訊息。 | [{"name": "get_jobs", "arguments": {"namespace": "migration"}}, {"name": "get_pod_events", "arguments": {"pod_name": "database-migration", "namespace": "migration"}}, {"name": "get_logs", "arguments": {"pod_name": "database-migration", "namespace": "migration"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_jobs\"... |
a943cf4b42b94412a14cf6b95f9eb51e | Rollback the current 'mobile-api' release in the prod namespace to revision 3. | 將 prod namespace 裡面的 mobile-api release 回滾到第 3 個版本。 | [{"name": "helm_rollback", "description": "Rollback a Helm release to a previous revision.\n\nArgs:\n release_name: Name of the release\n revision: Revision number to rollback to (use helm_history to see revisions)\n namespace: Kubernetes namespace\n force: Force resource updates through delete/recreate\n ... | 使用者希望將位於 'prod' 命名空間中的 'mobile-api' Helm release 回滾到第 3 個版本。我可以使用 'helm_rollback' 工具來執行此操作，將 release_name 設為 'mobile-api'，namespace 設為 'prod'，並將 revision 指定為 3。 | [{"name": "helm_rollback", "arguments": {"release_name": "mobile-api", "revision": 3, "namespace": "prod"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"helm_rollb... |
6196ab99f36b4679aa0272858d68051a | Get the status of the 'auth-server' ingress in the production namespace. | 幫我確認一下 production namespace 裡面 auth-server 這個 ingress 的運作狀態。 | [{"name": "get_ingress", "description": "Get Ingress resources in a namespace or cluster-wide.\n\nArgs:\n namespace: Namespace to list ingresses from (all namespaces if not specified)\n context: Kubernetes context to use (uses current context if not specified)", "input_schema": {"additionalProperties": false, "pr... | 為了確認 production namespace 中 auth-server 這個 Ingress 的運作狀態，我需要先獲取該資源的詳細資訊。雖然 `get_ingress` 可以列出 Ingress，但 `kubectl_describe` 能提供更完整的狀態描述，因此我將使用 `kubectl_describe` 來查詢該資源。 | [{"name": "kubectl_describe", "arguments": {"resource_type": "ingress", "name": "auth-server", "namespace": "production"}}] | [
{
"role": "system",
"content": "You are a function calling AI model. You are provided with function signatures within <tools></tools> XML tags. You may call one or more functions to assist with the user query. Don't make assumptions about what values to plug into functions.\n<tools>\n{\"name\": \"get_ingres... |
End of preview. Expand in Data Studio
Kubernetes MCP Function-Calling SFT Dataset
本資料集是一個專為 Kubernetes 操作設計的高品質 Function-Calling 微調資料集，包含 1,500 筆指令與工具呼叫對應樣本，旨在提升大語言模型在 Kubernetes 集群管理與診斷任務中的自動化執行能力。
資料來源與生成方式
本資料集基於 MCP (Model Context Protocol) 定義的 Kubernetes 工具集生成。我們使用 google/gemini-3.1-flash-lite 作為生成模型，針對各類 Kubernetes 運維場景（如網路策略診斷、資源備份、Pod 健康檢查等）進行合成數據建構。
資料集格式嚴格對齊 twinkle-ai/tw-function-call-reasoning-10k 標準，確保模型在執行工具前具備良好的思維鏈（Chain-of-Thought）推理能力。
品質控管與驗證流程
為了確保資料集的可靠性與可用性，我們執行了以下嚴格的過濾程序：
- JSON Schema 驗證：所有生成的工具參數皆經過 Schema 檢查，確保參數型別與必要欄位符合定義。
- 執行驗證：所有樣本均經過實際環境模擬呼叫，僅保留能夠成功執行且返回有效結果的樣本。
- LLM Faithfulness 評審：透過額外的 LLM 審查機制，確認
think過程與最終answer（Function Call）之間的邏輯一致性，排除幻覺與無效指令。
欄位說明
| 欄位名稱 | 型別 | 說明 |
|---|---|---|
id |
string | 唯一識別碼 |
query |
string | 原始英文指令 |
query_zhtw |
string | 繁體中文指令 |
tools |
list | 可用的 MCP 工具定義 (JSON Schema) |
think |
string | 繁體中文思維鏈 (CoT)，用於引導模型推理 |
answer |
string | 最終的 Function Call 結構 |
messages |
list | 遵循 Hermes 格式的對話列表，可直接用於 SFT |
使用方式
您可以透過 HuggingFace datasets 函式庫輕鬆載入：
from datasets import load_dataset
dataset = load_dataset("your-username/your-dataset-name")
# 查看第一筆資料
print(dataset['train'][0]['messages'])
用於 SFT 微調
本資料集已預處理為 messages 格式，直接支援主流的指令微調框架（如 Axolotl, LLaMA-Factory）。建議在微調時使用適當的 System Prompt 來引導模型進行工具呼叫。
涵蓋工具範疇
本資料集涵蓋 30 種 Kubernetes 核心操作，包括但不限於：
- 網路與安全：
analyze_network_policies,audit_rbac_permissions,check_secrets_security - 診斷與監控：
diagnose_pod_crash,check_pod_health,diagnose_network_connectivity - 資源管理：
create_deployment,delete_resource,backup_resource,cleanup_pods - Cilium 整合：
cilium_get_status,cilium_list_identities,cilium_list_policies
限制與免責聲明
- 合成資料：本資料集由 AI 生成，儘管經過驗證，仍可能存在極少數邏輯瑕疵。
- 版本依賴：工具定義基於特定的 MCP 協議版本，若 Kubernetes API 或 MCP 工具定義更新，資料集中的參數可能需要同步調整。
- 非官方：本資料集為社群開發，與 Kubernetes 官方或 Google 無直接關聯。請務必在測試環境中驗證模型生成的指令，切勿直接在生產環境執行未經審核的自動化操作。
- Downloads last month
- 48
Models trained or fine-tuned on Simon-Liu/kubectl-mcp-server-tool-call-reasoning-1k