Merge pull request #840 from Abirdcfly/agent-history

feat: agent can get history
kubeagi · Mar 13, 2024 · 1beca6d · 1beca6d
2 parents 6323c69 + 3a5a1c0
commit 1beca6d
Show file tree

Hide file tree

Showing 14 changed files with 71 additions and 35 deletions.
diff --git a/api/app-node/agent/v1alpha1/agent_types.go b/api/app-node/agent/v1alpha1/agent_types.go
@@ -51,7 +51,8 @@ type Options struct {
 	// +kubebuilder:validation:Maximum=10
 	// +kubebuilder:default=5
 	MaxIterations int `json:"maxIterations,omitempty"`
-
+	// Memory for chain memory
+	Memory node.Memory `json:"memory,omitempty"`
 	// The options below might be used later
 	// prompt prompts.PromptTemplate
 	// outputKey               string

diff --git a/api/app-node/agent/v1alpha1/zz_generated.deepcopy.go b/api/app-node/agent/v1alpha1/zz_generated.deepcopy.go
diff --git a/api/app-node/chain/v1alpha1/llmchain_types.go b/api/app-node/chain/v1alpha1/llmchain_types.go
@@ -37,7 +37,7 @@ type CommonChainConfig struct {
 	MaxNumberOfConccurent int `json:"maxNumberOfConccurent,omitempty"`
 
 	// for memory
-	Memory Memory `json:"memory,omitempty"`
+	Memory node.Memory `json:"memory,omitempty"`
 
 	// Model is the model to use in an llm call.like `gpt-3.5-turbo` or `chatglm_turbo`
 	// Usually this value is just empty
@@ -68,16 +68,6 @@ type CommonChainConfig struct {
 	RepetitionPenalty float64 `json:"repetitionPenalty,omitempty"`
 }
 
-type Memory struct {
-	// 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.
-	// +kubebuilder:validation:Minimum=0
-	// +kubebuilder:validation:Maximum=30
-	// +kubebuilder:default=5
-	ConversionWindowSize int `json:"conversionWindowSize,omitempty"`
-}
-
 // LLMChainStatus defines the observed state of LLMChain
 type LLMChainStatus struct {
 	// ObservedGeneration is the last observed generation.

diff --git a/api/app-node/chain/v1alpha1/zz_generated.deepcopy.go b/api/app-node/chain/v1alpha1/zz_generated.deepcopy.go
diff --git a/api/app-node/common_type.go b/api/app-node/common_type.go
@@ -27,6 +27,16 @@ const (
 	OutputLengthAnnotationKey = v1alpha1.Group + `/output-rules`
 )
 
+type Memory struct {
+	// 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.
+	// +kubebuilder:validation:Minimum=0
+	// +kubebuilder:validation:Maximum=30
+	// +kubebuilder:default=5
+	ConversionWindowSize int `json:"conversionWindowSize,omitempty"`
+}
+
 type Ref struct {
 	Kind   string `json:"kind,omitempty"`
 	Group  string `json:"group,omitempty"`

diff --git a/apiserver/pkg/application/application.go b/apiserver/pkg/application/application.go
@@ -31,6 +31,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/controller/controllerutil"
 
+	apinode "github.com/kubeagi/arcadia/api/app-node"
 	apiagent "github.com/kubeagi/arcadia/api/app-node/agent/v1alpha1"
 	apichain "github.com/kubeagi/arcadia/api/app-node/chain/v1alpha1"
 	apidocumentloader "github.com/kubeagi/arcadia/api/app-node/documentloader/v1alpha1"
@@ -440,7 +441,7 @@ func UpdateApplicationConfig(ctx context.Context, c client.Client, input generat
 					Description: "qachain",
 				},
 				CommonChainConfig: apichain.CommonChainConfig{
-					Memory: apichain.Memory{
+					Memory: apinode.Memory{
 						ConversionWindowSize: pointer.IntDeref(input.ConversionWindowSize, 0),
 					},
 					Model:       pointer.StringDeref(input.Model, ""),
@@ -473,7 +474,7 @@ func UpdateApplicationConfig(ctx context.Context, c client.Client, input generat
 					Description: "qachain",
 				},
 				CommonChainConfig: apichain.CommonChainConfig{
-					Memory: apichain.Memory{
+					Memory: apinode.Memory{
 						ConversionWindowSize: pointer.IntDeref(input.ConversionWindowSize, 0),
 					},
 					Model:       pointer.StringDeref(input.Model, ""),

diff --git a/config/crd/bases/arcadia.kubeagi.k8s.com.cn_agents.yaml b/config/crd/bases/arcadia.kubeagi.k8s.com.cn_agents.yaml
@@ -67,6 +67,22 @@ spec:
                     maximum: 10
                     minimum: 1
                     type: integer
+                  memory:
+                    description: Memory for chain memory
+                    properties:
+                      conversionWindowSize:
+                        default: 5
+                        description: ConversionWindowSize is the maximum number of
+                          conversation rounds in memory.Can only use MaxTokenLimit
+                          or ConversionWindowSize.
+                        maximum: 30
+                        minimum: 0
+                        type: integer
+                      maxTokenLimit:
+                        description: MaxTokenLimit is the maximum number of tokens
+                          to keep in memory. Can only use MaxTokenLimit or ConversionWindowSize.
+                        type: integer
+                    type: object
                   showToolAction:
                     default: false
                     description: Whether to show tool action in the streaming output

diff --git a/deploy/charts/arcadia/crds/arcadia.kubeagi.k8s.com.cn_agents.yaml b/deploy/charts/arcadia/crds/arcadia.kubeagi.k8s.com.cn_agents.yaml
@@ -67,6 +67,22 @@ spec:
                     maximum: 10
                     minimum: 1
                     type: integer
+                  memory:
+                    description: Memory for chain memory
+                    properties:
+                      conversionWindowSize:
+                        default: 5
+                        description: ConversionWindowSize is the maximum number of
+                          conversation rounds in memory.Can only use MaxTokenLimit
+                          or ConversionWindowSize.
+                        maximum: 30
+                        minimum: 0
+                        type: integer
+                      maxTokenLimit:
+                        description: MaxTokenLimit is the maximum number of tokens
+                          to keep in memory. Can only use MaxTokenLimit or ConversionWindowSize.
+                        type: integer
+                    type: object
                   showToolAction:
                     default: false
                     description: Whether to show tool action in the streaming output

diff --git a/pkg/appruntime/agent/executor.go b/pkg/appruntime/agent/executor.go
@@ -24,12 +24,14 @@ import (
 	"github.com/tmc/langchaingo/agents"
 	"github.com/tmc/langchaingo/callbacks"
 	"github.com/tmc/langchaingo/llms"
+	langchaingoschema "github.com/tmc/langchaingo/schema"
 	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/klog/v2"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
 	"github.com/kubeagi/arcadia/api/app-node/agent/v1alpha1"
 	"github.com/kubeagi/arcadia/pkg/appruntime/base"
+	"github.com/kubeagi/arcadia/pkg/appruntime/chain"
 	"github.com/kubeagi/arcadia/pkg/appruntime/log"
 	"github.com/kubeagi/arcadia/pkg/appruntime/tools"
 )
@@ -59,6 +61,13 @@ func (p *Executor) Run(ctx context.Context, cli client.Client, args map[string]a
 	}
 	allowedTools := tools.InitTools(ctx, instance.Spec.AllowedTools)
 
+	var history langchaingoschema.ChatMessageHistory
+	if v3, ok := args[base.LangchaingoChatMessageHistoryKeyInArg]; ok && v3 != nil {
+		history, ok = v3.(langchaingoschema.ChatMessageHistory)
+		if !ok {
+			return args, errors.New("history not memory.ChatMessageHistory")
+		}
+	}
 	// Initialize executor using langchaingo
 	executorOptions := func(o *agents.CreationOptions) {
 		agents.WithCallbacksHandler(log.KLogHandler{LogLevel: 3})(o)
@@ -70,6 +79,7 @@ func (p *Executor) Run(ctx context.Context, cli client.Client, args map[string]a
 				agents.WithCallbacksHandler(streamHandler)(o)
 			}
 		}
+		agents.WithMemory(chain.GetMemory(llm, instance.Spec.AgentConfig.Options.Memory, history, "", ""))(o)
 	}
 	executor, err := agents.Initialize(llm, allowedTools, agents.ZeroShotReactDescription, executorOptions)
 	if err != nil {

diff --git a/pkg/appruntime/chain/apichain.go b/pkg/appruntime/chain/apichain.go
@@ -88,8 +88,8 @@ func (l *APIChain) Run(ctx context.Context, _ client.Client, args map[string]any
 	options := GetChainOptions(instance.Spec.CommonChainConfig)
 
 	chain := chains.NewAPIChain(llm, http.DefaultClient)
-	chain.RequestChain.Memory = getMemory(llm, instance.Spec.Memory, history, "", "")
-	chain.AnswerChain.Memory = getMemory(llm, instance.Spec.Memory, history, "input", "")
+	chain.RequestChain.Memory = GetMemory(llm, instance.Spec.Memory, history, "", "")
+	chain.AnswerChain.Memory = GetMemory(llm, instance.Spec.Memory, history, "input", "")
 	l.APIChain = chain
 	apiDoc := instance.Spec.APIDoc
 	if apiDoc == "" {

diff --git a/pkg/appruntime/chain/common.go b/pkg/appruntime/chain/common.go
@@ -27,6 +27,7 @@ import (
 	langchaingoschema "github.com/tmc/langchaingo/schema"
 	"k8s.io/klog/v2"
 
+	appnode "github.com/kubeagi/arcadia/api/app-node"
 	"github.com/kubeagi/arcadia/api/app-node/chain/v1alpha1"
 	"github.com/kubeagi/arcadia/pkg/appruntime/base"
 )
@@ -85,7 +86,7 @@ func GetChainOptions(config v1alpha1.CommonChainConfig) []chains.ChainCallOption
 	return options
 }
 
-func getMemory(llm llms.Model, config v1alpha1.Memory, history langchaingoschema.ChatMessageHistory, inputKey, outputKey string) langchaingoschema.Memory {
+func GetMemory(llm llms.Model, config appnode.Memory, history langchaingoschema.ChatMessageHistory, inputKey, outputKey string) langchaingoschema.Memory {
 	if inputKey == "" {
 		inputKey = "question"
 	}

diff --git a/pkg/appruntime/chain/llmchain.go b/pkg/appruntime/chain/llmchain.go
@@ -112,7 +112,7 @@ func (l *LLMChain) Run(ctx context.Context, _ client.Client, args map[string]any
 	}
 	chain := chains.NewLLMChain(llm, prompt)
 	if history != nil {
-		chain.Memory = getMemory(llm, instance.Spec.Memory, history, "", "")
+		chain.Memory = GetMemory(llm, instance.Spec.Memory, history, "", "")
 	}
 	l.LLMChain = *chain
 

diff --git a/pkg/appruntime/chain/retrievalqachain.go b/pkg/appruntime/chain/retrievalqachain.go
@@ -113,9 +113,9 @@ func (l *RetrievalQAChain) Run(ctx context.Context, _ client.Client, args map[st
 
 	llmChain := chains.NewLLMChain(llm, prompt)
 	if history != nil {
-		llmChain.Memory = getMemory(llm, instance.Spec.Memory, history, "", "")
+		llmChain.Memory = GetMemory(llm, instance.Spec.Memory, history, "", "")
 	}
-	chain := chains.NewConversationalRetrievalQA(chains.NewStuffDocuments(llmChain), chains.LoadCondenseQuestionGenerator(llm), retriever, getMemory(llm, instance.Spec.Memory, history, "", ""))
+	chain := chains.NewConversationalRetrievalQA(chains.NewStuffDocuments(llmChain), chains.LoadCondenseQuestionGenerator(llm), retriever, GetMemory(llm, instance.Spec.Memory, history, "", ""))
 	l.ConversationalRetrievalQA = chain
 	args["query"] = args["question"]
 	var out string

diff --git a/tests/example-test.sh b/tests/example-test.sh
@@ -587,6 +587,11 @@ info "8.6 tool test"
 kubectl apply -f config/samples/app_llmchain_chat_with_bot_tool.yaml
 waitCRDStatusReady "Application" "arcadia" "base-chat-with-bot-tool"
 sleep 3
+info "8.6.1 conversation test"
+info "23*34 结果应该是 782, 结果再乘2是 1564, 再减去564是 1000"
+getRespInAppChat "base-chat-with-bot-tool" "arcadia" "计算 23*34 的结果" "" "false"
+getRespInAppChat "base-chat-with-bot-tool" "arcadia" "结果再乘2" ${resp_conversation_id} "false"
+getRespInAppChat "base-chat-with-bot-tool" "arcadia" "结果再减去564" ${resp_conversation_id} "false"
 #	info "8.6.1 bingsearch test"
 #	getRespInAppChat "base-chat-with-bot-tool" "arcadia" "用30字介绍一下时速云" "" "true"
 #	if [ -z "$references" ] || [ "$references" = "null" ]; then