feat: add more config for app

Signed-off-by: Abirdcfly <fp544037857@gmail.com>

.github/workflows/example_test.yaml

@@ -86,7 +86,7 @@ jobs:
         run: tests/example-test.sh
       - name: Upload logs if test fail
         if: failure()
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
           name: ${{ github.sha }}-${{ matrix.no }}.logs
           path: ${{ env.LOG_DIR }}

api/app-node/chain/v1alpha1/llmchain_types.go

@@ -42,13 +42,36 @@ type Output struct {
 }
 
 type CommonChainConfig struct {
-	// 记忆相关参数
+	// for memory
 	Memory Memory `json:"memory,omitempty"`
+
+	// MaxTokens is the maximum number of tokens to generate to use in a llm call.
+	MaxTokens int `json:"maxTokens,omitempty"`
+	// Temperature is the temperature for sampling to use in a llm call, between 0 and 1.
+	//+kubebuilder:validation:Minimum=0
+	//+kubebuilder:validation:Maximum=1
+	Temperature float64 `json:"temperature,omitempty"`
+	// StopWords is a list of words to stop on to use in a llm call.
+	StopWords []string `json:"stopWords,omitempty"`
+	// TopK is the number of tokens to consider for top-k sampling in a llm call.
+	TopK int `json:"topK,omitempty"`
+	// TopP is the cumulative probability for top-p sampling in a llm call.
+	TopP float64 `json:"topP,omitempty"`
+	// Seed is a seed for deterministic sampling in a llm call.
+	Seed int `json:"seed,omitempty"`
+	// MinLength is the minimum length of the generated text in a llm call.
+	MinLength int `json:"minLength,omitempty"`
+	// MaxLength is the maximum length of the generated text in a llm call.
+	MaxLength int `json:"maxLength,omitempty"`
+	// RepetitionPenalty is the repetition penalty for sampling in a llm call.
+	RepetitionPenalty float64 `json:"repetitionPenalty,omitempty"`
 }
 
 type Memory struct {
-	// 能记住的最大 token 数
+	// MaxTokenLimit is the maximum number of tokens to keep in memory. Can only use MaxTokenLimit or ConversionWindowSize.
 	MaxTokenLimit int `json:"maxTokenLimit,omitempty"`
+	// ConversionWindowSize is the maximum number of conversation rounds in memory.Can only use MaxTokenLimit or ConversionWindowSize.
+	ConversionWindowSize int `json:"conversionWindowSize,omitempty"`
 }
 
 // LLMChainStatus defines the observed state of LLMChain

api/app-node/retriever/v1alpha1/knowledgebaseretriever_types.go

@@ -25,9 +25,10 @@ import (
 
 // KnowledgeBaseRetrieverSpec defines the desired state of KnowledgeBaseRetriever
 type KnowledgeBaseRetrieverSpec struct {
-	v1alpha1.CommonSpec `json:",inline"`
-	Input               Input  `json:"input,omitempty"`
-	Output              Output `json:"output,omitempty"`
+	v1alpha1.CommonSpec   `json:",inline"`
+	Input                 Input  `json:"input,omitempty"`
+	Output                Output `json:"output,omitempty"`
+	CommonRetrieverConfig `json:",inline"`
 }
 
 type Input struct {
@@ -38,6 +39,10 @@ type Output struct {
 	node.CommonOrInPutOrOutputRef `json:",inline"`
 }
 
+type CommonRetrieverConfig struct {
+	ScoreThreshold float32 `json:"scoreThreshold,omitempty"` // TODO
+}
+
 // KnowledgeBaseRetrieverStatus defines the observed state of KnowledgeBaseRetriever
 type KnowledgeBaseRetrieverStatus struct {
 	// ObservedGeneration is the last observed generation.

config/crd/bases/chain.arcadia.kubeagi.k8s.com.cn_llmchains.yaml

@@ -88,13 +88,30 @@ spec:
                 - llm
                 - prompt
                 type: object
+              maxLength:
+                description: MaxLength is the maximum length of the generated text
+                  in a llm call.
+                type: integer
+              maxTokens:
+                description: MaxTokens is the maximum number of tokens to generate
+                  to use in a llm call.
+                type: integer
               memory:
-                description: 记忆相关参数
+                description: for memory
                 properties:
+                  conversionWindowSize:
+                    description: ConversionWindowSize is the maximum number of conversation
+                      rounds in memory.Can only use MaxTokenLimit or ConversionWindowSize.
+                    type: integer
                   maxTokenLimit:
-                    description: 能记住的最大 token 数
+                    description: MaxTokenLimit is the maximum number of tokens to
+                      keep in memory. Can only use MaxTokenLimit or ConversionWindowSize.
                     type: integer
                 type: object
+              minLength:
+                description: MinLength is the minimum length of the generated text
+                  in a llm call.
+                type: integer
               output:
                 properties:
                   apiGroup:
@@ -111,6 +128,33 @@ spec:
                     description: Name is the name of resource being referenced
                     type: string
                 type: object
+              repetitionPenalty:
+                description: RepetitionPenalty is the repetition penalty for sampling
+                  in a llm call.
+                type: number
+              seed:
+                description: Seed is a seed for deterministic sampling in a llm call.
+                type: integer
+              stopWords:
+                description: StopWords is a list of words to stop on to use in a llm
+                  call.
+                items:
+                  type: string
+                type: array
+              temperature:
+                description: Temperature is the temperature for sampling to use in
+                  a llm call, between 0 and 1.
+                maximum: 1
+                minimum: 0
+                type: number
+              topK:
+                description: TopK is the number of tokens to consider for top-k sampling
+                  in a llm call.
+                type: integer
+              topP:
+                description: TopP is the cumulative probability for top-p sampling
+                  in a llm call.
+                type: number
             required:
             - input
             - output

config/crd/bases/chain.arcadia.kubeagi.k8s.com.cn_retrievalqachains.yaml

@@ -109,13 +109,30 @@ spec:
                 - prompt
                 - retriever
                 type: object
+              maxLength:
+                description: MaxLength is the maximum length of the generated text
+                  in a llm call.
+                type: integer
+              maxTokens:
+                description: MaxTokens is the maximum number of tokens to generate
+                  to use in a llm call.
+                type: integer
               memory:
-                description: 记忆相关参数
+                description: for memory
                 properties:
+                  conversionWindowSize:
+                    description: ConversionWindowSize is the maximum number of conversation
+                      rounds in memory.Can only use MaxTokenLimit or ConversionWindowSize.
+                    type: integer
                   maxTokenLimit:
-                    description: 能记住的最大 token 数
+                    description: MaxTokenLimit is the maximum number of tokens to
+                      keep in memory. Can only use MaxTokenLimit or ConversionWindowSize.
                     type: integer
                 type: object
+              minLength:
+                description: MinLength is the minimum length of the generated text
+                  in a llm call.
+                type: integer
               output:
                 properties:
                   apiGroup:
@@ -132,6 +149,33 @@ spec:
                     description: Name is the name of resource being referenced
                     type: string
                 type: object
+              repetitionPenalty:
+                description: RepetitionPenalty is the repetition penalty for sampling
+                  in a llm call.
+                type: number
+              seed:
+                description: Seed is a seed for deterministic sampling in a llm call.
+                type: integer
+              stopWords:
+                description: StopWords is a list of words to stop on to use in a llm
+                  call.
+                items:
+                  type: string
+                type: array
+              temperature:
+                description: Temperature is the temperature for sampling to use in
+                  a llm call, between 0 and 1.
+                maximum: 1
+                minimum: 0
+                type: number
+              topK:
+                description: TopK is the number of tokens to consider for top-k sampling
+                  in a llm call.
+                type: integer
+              topP:
+                description: TopP is the cumulative probability for top-p sampling
+                  in a llm call.
+                type: number
             required:
             - input
             - output

config/crd/bases/retriever.arcadia.kubeagi.k8s.com.cn_knowledgebaseretrievers.yaml

@@ -82,6 +82,8 @@ spec:
                     description: Name is the name of resource being referenced
                     type: string
                 type: object
+              scoreThreshold:
+                type: number
             type: object
           status:
             description: KnowledgeBaseRetrieverStatus defines the observed state of

config/samples/app_llmchain_chat_with_bot.yaml

@@ -0,0 +1,117 @@
+apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1
+kind: Application
+metadata:
+  name: base-chat-with-bot
+  namespace: arcadia
+spec:
+  displayName: "对话机器人"
+  description: "和AI对话，品赛博人生"
+  prologue: "Hello, I am KubeAGI Bot🤖, Tell me something?"
+  nodes:
+    - name: Input
+      displayName: "用户输入"
+      description: "用户输入节点，必须"
+      ref:
+        kind: Input
+        name: Input
+      nextNodeName: ["prompt-node"]
+    - name: prompt-node
+      displayName: "prompt"
+      description: "设定prompt，template中可以使用{{xx}}来替换变量"
+      ref:
+        apiGroup: prompt.arcadia.kubeagi.k8s.com.cn
+        kind: Prompt
+        name: base-chat-with-bot
+      nextNodeName: ["chain-node"]
+    - name: llm-node
+      displayName: "zhipu大模型服务"
+      description: "设定质谱大模型的访问信息"
+      ref:
+        apiGroup: arcadia.kubeagi.k8s.com.cn
+        kind: LLM
+        name: base-chat-with-bot
+      nextNodeName: ["chain-node"]
+    - name: chain-node
+      displayName: "llm chain"
+      description: "chain是langchain的核心概念，llmChain用于连接prompt和llm"
+      ref:
+        apiGroup: chain.arcadia.kubeagi.k8s.com.cn
+        kind: LLMChain
+        name: base-chat-with-bot
+      nextNodeName: ["Output"]
+    - name: Output
+      displayName: "最终输出"
+      description: "最终输出节点，必须"
+      ref:
+        kind: Output
+        name: Output
+---
+apiVersion: prompt.arcadia.kubeagi.k8s.com.cn/v1alpha1
+kind: Prompt
+metadata:
+  name: base-chat-with-bot
+  namespace: arcadia
+spec:
+  displayName: "设定对话的prompt"
+  description: "设定对话的prompt"
+  userMessage: |
+    The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.
+
+    Current conversation:
+    {{.history}}
+    Human: {{.question}}
+    AI:
+  input:
+    kind: "Input"
+    name: "Input"
+  output:
+    apiGroup: chain.arcadia.kubeagi.k8s.com.cn
+    kind: LLMChain
+    name: base-chat-with-bot
+---
+apiVersion: chain.arcadia.kubeagi.k8s.com.cn/v1alpha1
+kind: LLMChain
+metadata:
+  name: base-chat-with-bot
+  namespace: arcadia
+spec:
+  displayName: "llm chain"
+  description: "llm chain"
+  memory:
+    conversionWindowSize: 2
+  input:
+    llm:
+      apiGroup: arcadia.kubeagi.k8s.com.cn
+      kind: LLM
+      name: base-chat-with-bot
+    prompt:
+      apiGroup: prompt.arcadia.kubeagi.k8s.com.cn
+      kind: Prompt
+      name: base-chat-with-bot
+  output:
+    apiGroup: "arcadia.kubeagi.k8s.com.cn"
+    kind: "Output"
+    name: "output-node"
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: base-chat-with-bot
+  namespace: arcadia
+type: Opaque
+data:
+  apiKey: "MTZlZDcxYzcwMDE0NGFiMjIyMmI5YmEwZDFhMTBhZTUuUTljWVZtWWxmdjlnZGtDeQ==" # replace this with your API key
+---
+apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1
+kind: LLM
+metadata:
+  name: base-chat-with-bot
+  namespace: arcadia
+spec:
+  type: "zhipuai"
+  provider:
+    endpoint:
+      url: "https://open.bigmodel.cn/api/paas/v3/model-api" # replace this with your LLM URL(Zhipuai use predefined url https://open.bigmodel.cn/api/paas/v3/model-api)
+      authSecret:
+        kind: secret
+        name: base-chat-with-bot