Misc

hupe1980 · Jun 28, 2023 · c12b8a0 · c12b8a0
1 parent d4e500f
commit c12b8a0
Show file tree

Hide file tree

Showing 35 changed files with 105 additions and 166 deletions.
diff --git a/agent/conversational.go b/agent/conversational.go
@@ -70,10 +70,7 @@ func NewConversationalReactDescription(llm schema.LLM, tools []schema.Tool) (*Co
 		AIPrefix:     "AI",
 	}
 
-	prompt, err := createConversationalPrompt(tools, opts.Prefix, opts.Instructions, opts.Suffix)
-	if err != nil {
-		return nil, err
-	}
+	prompt := createConversationalPrompt(tools, opts.Prefix, opts.Instructions, opts.Suffix)
 
 	llmChain, err := chain.NewLLM(llm, prompt, func(o *chain.LLMOptions) {
 		o.Memory = memory.NewConversationBuffer()
@@ -166,7 +163,7 @@ func (a *ConversationalReactDescription) parseOutput(output string) ([]schema.Ag
 	}, nil, nil
 }
 
-func createConversationalPrompt(tools []schema.Tool, prefix, instructions, suffix string) (*prompt.Template, error) {
+func createConversationalPrompt(tools []schema.Tool, prefix, instructions, suffix string) *prompt.Template {
 	return prompt.NewTemplate(strings.Join([]string{prefix, instructions, suffix}, "\n\n"), func(o *prompt.TemplateOptions) {
 		o.PartialValues = prompt.PartialValues{
 			"toolNames":        toolNames(tools),

diff --git a/agent/react.go b/agent/react.go
@@ -59,10 +59,7 @@ func NewReactDescription(llm schema.LLM, tools []schema.Tool) (*ReactDescription
 		OutputKey:    "output",
 	}
 
-	prompt, err := createReactDescriptioPrompt(tools, opts.Prefix, opts.Instructions, opts.Suffix)
-	if err != nil {
-		return nil, err
-	}
+	prompt := createReactDescriptioPrompt(tools, opts.Prefix, opts.Instructions, opts.Suffix)
 
 	llmChain, err := chain.NewLLM(llm, prompt)
 	if err != nil {
@@ -153,7 +150,7 @@ func (a *ReactDescription) parseOutput(output string) ([]schema.AgentAction, *sc
 	}, nil, nil
 }
 
-func createReactDescriptioPrompt(tools []schema.Tool, prefix, instructions, suffix string) (*prompt.Template, error) {
+func createReactDescriptioPrompt(tools []schema.Tool, prefix, instructions, suffix string) *prompt.Template {
 	return prompt.NewTemplate(strings.Join([]string{prefix, instructions, suffix}, "\n\n"), func(o *prompt.TemplateOptions) {
 		o.PartialValues = prompt.PartialValues{
 			"toolNames":        toolNames(tools),

diff --git a/callback/callback.go b/callback/callback.go
@@ -27,7 +27,7 @@ func (h *handler) OnModelNewToken(token string) error {
 	return nil
 }
 
-func (h *handler) OnModelEnd(result schema.LLMResult) error {
+func (h *handler) OnModelEnd(result schema.ModelResult) error {
 	return nil
 }
 

diff --git a/callback/manager.go b/callback/manager.go
@@ -109,7 +109,7 @@ func (m *manager) OnModelNewToken(token string) error {
 	return nil
 }
 
-func (m *manager) OnModelEnd(result schema.LLMResult) error {
+func (m *manager) OnModelEnd(result schema.ModelResult) error {
 	for _, c := range m.callbacks {
 		if m.verbose || c.AlwaysVerbose() {
 			if err := c.OnModelEnd(result); err != nil {

diff --git a/callback/openai_info.go b/callback/openai_info.go
@@ -58,7 +58,7 @@ func (o *OpenAIHandler) AlwaysVerbose() bool {
 	return true
 }
 
-func (o *OpenAIHandler) OnModelEnd(result schema.LLMResult) error {
+func (o *OpenAIHandler) OnModelEnd(result schema.ModelResult) error {
 	if result.LLMOutput == nil {
 		return nil
 	}

diff --git a/chain/conversation.go b/chain/conversation.go
@@ -58,12 +58,7 @@ func NewConversation(llm schema.LLM, optFns ...func(o *ConversationOptions)) (*C
 	}
 
 	if opts.Prompt == nil {
-		var pErr error
-
-		opts.Prompt, pErr = prompt.NewTemplate(conversationTemplate)
-		if pErr != nil {
-			return nil, pErr
-		}
+		opts.Prompt = prompt.NewTemplate(conversationTemplate)
 	}
 
 	return &Conversation{
@@ -147,7 +142,7 @@ func (c *Conversation) OutputKeys() []string {
 	return []string{c.opts.OutputKey}
 }
 
-func (c *Conversation) createOutputs(llmResult *schema.LLMResult) ([]map[string]any, error) {
+func (c *Conversation) createOutputs(llmResult *schema.ModelResult) ([]map[string]any, error) {
 	result := make([]map[string]any, len(llmResult.Generations)-1)
 
 	for _, generation := range llmResult.Generations {

diff --git a/chain/conversational_retrieval.go b/chain/conversational_retrieval.go
@@ -71,12 +71,7 @@ func NewConversationalRetrieval(llm schema.LLM, retriever schema.Retriever, optF
 	}
 
 	if opts.CondenseQuestionPrompt == nil {
-		p, err := prompt.NewTemplate(defaultcondenseQuestionPromptTemplate)
-		if err != nil {
-			return nil, err
-		}
-
-		opts.CondenseQuestionPrompt = p
+		opts.CondenseQuestionPrompt = prompt.NewTemplate(defaultcondenseQuestionPromptTemplate)
 	}
 
 	condenseQuestionChain, err := NewLLM(llm, opts.CondenseQuestionPrompt)

diff --git a/chain/llm.go b/chain/llm.go
@@ -130,10 +130,10 @@ func (c *LLM) OutputKeys() []string {
 	return []string{c.opts.OutputKey}
 }
 
-func (c *LLM) createOutputs(llmResult *schema.LLMResult) ([]map[string]any, error) {
-	result := make([]map[string]any, len(llmResult.Generations)-1)
+func (c *LLM) createOutputs(modelResult *schema.ModelResult) ([]map[string]any, error) {
+	result := make([]map[string]any, len(modelResult.Generations)-1)
 
-	for _, generation := range llmResult.Generations {
+	for _, generation := range modelResult.Generations {
 		parsed, err := c.opts.OutputParser.ParseResult(generation)
 		if err != nil {
 			return nil, err

diff --git a/chain/llm_bash.go b/chain/llm_bash.go
@@ -57,12 +57,9 @@ func NewLLMBash(llm schema.LLM, optFns ...func(o *LLMBashOptions)) (*LLMBash, er
 		fn(&opts)
 	}
 
-	prompt, err := prompt.NewTemplate(llmBashTemplate, func(o *prompt.TemplateOptions) {
+	prompt := prompt.NewTemplate(llmBashTemplate, func(o *prompt.TemplateOptions) {
 		o.OutputParser = outputparser.NewFencedCodeBlock("```bash")
 	})
-	if err != nil {
-		return nil, err
-	}
 
 	llmChain, err := NewLLM(llm, prompt)
 	if err != nil {

diff --git a/chain/llm_math.go b/chain/llm_math.go
@@ -63,12 +63,9 @@ func NewLLMMath(llm schema.LLM, optFns ...func(o *LLMMathOptions)) (*LLMMath, er
 		fn(&opts)
 	}
 
-	prompt, err := prompt.NewTemplate(llmMathTemplate, func(o *prompt.TemplateOptions) {
+	prompt := prompt.NewTemplate(llmMathTemplate, func(o *prompt.TemplateOptions) {
 		o.OutputParser = outputparser.NewFencedCodeBlock("```text")
 	})
-	if err != nil {
-		return nil, err
-	}
 
 	llmChain, err := NewLLM(llm, prompt)
 	if err != nil {

diff --git a/chain/prompt_selector_test.go b/chain/prompt_selector_test.go
@@ -12,9 +12,9 @@ import (
 
 func TestConditionalPromptSelector(t *testing.T) {
 	t.Run("GetPrompt", func(t *testing.T) {
-		defaultPrompt, _ := prompt.NewTemplate("Default Prompt")
-		llmPrompt, _ := prompt.NewTemplate("LLM Prompt")
-		chatModelPrompt, _ := prompt.NewTemplate("ChatModel Prompt")
+		defaultPrompt := prompt.NewTemplate("Default Prompt")
+		llmPrompt := prompt.NewTemplate("LLM Prompt")
+		chatModelPrompt := prompt.NewTemplate("ChatModel Prompt")
 
 		conditional1 := Conditional{
 			Condition: func(model schema.Model) bool {

diff --git a/chain/refine_documents.go b/chain/refine_documents.go
@@ -45,12 +45,7 @@ func NewRefineDocuments(llmChain *LLM, refineLLMChain *LLM, optFns ...func(o *Re
 	}
 
 	if opts.DocumentPrompt == nil {
-		p, err := prompt.NewTemplate("{{.pageContent}}")
-		if err != nil {
-			return nil, err
-		}
-
-		opts.DocumentPrompt = p
+		opts.DocumentPrompt = prompt.NewTemplate("{{.pageContent}}")
 	}
 
 	return &RefineDocuments{

diff --git a/chain/retrieval_qa.go b/chain/retrieval_qa.go
@@ -43,12 +43,7 @@ func NewRetrievalQA(llm schema.LLM, retriever schema.Retriever, optFns ...func(o
 	}
 
 	if opts.StuffQAPrompt == nil {
-		p, err := prompt.NewTemplate(defaultStuffQAPromptTemplate)
-		if err != nil {
-			return nil, err
-		}
-
-		opts.StuffQAPrompt = p
+		opts.StuffQAPrompt = prompt.NewTemplate(defaultStuffQAPromptTemplate)
 	}
 
 	llmChain, err := NewLLM(llm, opts.StuffQAPrompt)

diff --git a/chain/summarization.go b/chain/summarization.go
@@ -29,10 +29,7 @@ func NewStuffSummarization(llm schema.LLM, optFns ...func(o *StuffSummarizationO
 		fn(&opts)
 	}
 
-	stuffPrompt, err := prompt.NewTemplate(stuffSummarizationTemplate)
-	if err != nil {
-		return nil, err
-	}
+	stuffPrompt := prompt.NewTemplate(stuffSummarizationTemplate)
 
 	llmChain, err := NewLLM(llm, stuffPrompt, func(o *LLMOptions) {
 		o.CallbackOptions = opts.CallbackOptions
@@ -74,10 +71,7 @@ func NewRefineSummarization(llm schema.LLM, optFns ...func(o *RefineSummarizatio
 		fn(&opts)
 	}
 
-	stuffPrompt, err := prompt.NewTemplate(stuffSummarizationTemplate)
-	if err != nil {
-		return nil, err
-	}
+	stuffPrompt := prompt.NewTemplate(stuffSummarizationTemplate)
 
 	llmChain, err := NewLLM(llm, stuffPrompt, func(o *LLMOptions) {
 		o.CallbackOptions = opts.CallbackOptions
@@ -86,10 +80,7 @@ func NewRefineSummarization(llm schema.LLM, optFns ...func(o *RefineSummarizatio
 		return nil, err
 	}
 
-	refinePrompt, err := prompt.NewTemplate(refineSummarizationTemplate)
-	if err != nil {
-		return nil, err
-	}
+	refinePrompt := prompt.NewTemplate(refineSummarizationTemplate)
 
 	refineLLMChain, err := NewLLM(llm, refinePrompt, func(o *LLMOptions) {
 		o.CallbackOptions = opts.CallbackOptions

diff --git a/evaluation/context_qa_eval_chain.go b/evaluation/context_qa_eval_chain.go
@@ -56,12 +56,7 @@ func NewContextQAEvalChain(llm schema.LLM, optFns ...func(o *ContextQAEvalChainO
 	}
 
 	if opts.Prompt == nil {
-		contextQAEvalPrompt, err := prompt.NewTemplate(contextQAEvalTemplate)
-		if err != nil {
-			return nil, err
-		}
-
-		opts.Prompt = contextQAEvalPrompt
+		opts.Prompt = prompt.NewTemplate(contextQAEvalTemplate)
 	}
 
 	llmChain, err := chain.NewLLM(llm, opts.Prompt)

diff --git a/evaluation/cot_qa_eval_chain.go b/evaluation/cot_qa_eval_chain.go
@@ -47,12 +47,7 @@ func NewCOTQAEvalChain(llm schema.LLM, optFns ...func(o *COTQAEvalChainOptions))
 	}
 
 	if opts.Prompt == nil {
-		cotQAEvalPrompt, err := prompt.NewTemplate(cotQAEvalTemplate)
-		if err != nil {
-			return nil, err
-		}
-
-		opts.Prompt = cotQAEvalPrompt
+		opts.Prompt = prompt.NewTemplate(cotQAEvalTemplate)
 	}
 
 	contextQAEvalChain, err := NewContextQAEvalChain(llm, func(o *ContextQAEvalChainOptions) {

diff --git a/evaluation/qa_eval_chain.go b/evaluation/qa_eval_chain.go
@@ -41,13 +41,8 @@ type QAEvalChain struct {
 }
 
 func NewQAEvalChain(llm schema.LLM, optFns ...func(o *QAEvalChainOptions)) (*QAEvalChain, error) {
-	qaEvalPrompt, err := prompt.NewTemplate(qaEvalTemplate)
-	if err != nil {
-		return nil, err
-	}
-
 	opts := QAEvalChainOptions{
-		Prompt:        qaEvalPrompt,
+		Prompt:        prompt.NewTemplate(qaEvalTemplate),
 		QuestionKey:   "query",
 		AnswerKey:     "answer",
 		PredictionKey: "result",

diff --git a/model/chatmodel/anthropic.go b/model/chatmodel/anthropic.go
@@ -56,7 +56,7 @@ func NewAnthropic(apiKey string, optFns ...func(o *AnthropicOptions)) (*Anthropi
 	}, nil
 }
 
-func (cm *Anthropic) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (cm *Anthropic) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	res, err := cm.client.Complete(ctx, &anthropic.CompletionRequest{
 		Model:     cm.opts.ModelName,
 		MaxTokens: cm.opts.MaxTokens,
@@ -65,7 +65,7 @@ func (cm *Anthropic) Generate(ctx context.Context, messages schema.ChatMessages,
 		return nil, err
 	}
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{newChatGeneraton(res.Completion)}},
 		LLMOutput:   map[string]any{},
 	}, nil

diff --git a/model/chatmodel/fake.go b/model/chatmodel/fake.go
@@ -20,8 +20,8 @@ func NewFake(response string) *Fake {
 	}
 }
 
-func (cm *Fake) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
-	return &schema.LLMResult{
+func (cm *Fake) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{newChatGeneraton(cm.response)}},
 		LLMOutput:   map[string]any{},
 	}, nil

diff --git a/model/chatmodel/openai.go b/model/chatmodel/openai.go
@@ -74,7 +74,7 @@ func newOpenAI(client *openai.Client, opts OpenAIOptions) (*OpenAI, error) {
 	}, nil
 }
 
-func (cm *OpenAI) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (cm *OpenAI) Generate(ctx context.Context, messages schema.ChatMessages, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	opts := schema.GenerateOptions{}
 
 	openAIMessages := []openai.ChatCompletionMessage{}
@@ -114,7 +114,7 @@ func (cm *OpenAI) Generate(ctx context.Context, messages schema.ChatMessages, op
 	text := res.Choices[0].Message.Content
 	role := res.Choices[0].Message.Role
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{schema.Generation{
 			Text:    text,
 			Message: openAIResponseToChatMessage(role, text),

diff --git a/model/llm/cohere.go b/model/llm/cohere.go
@@ -58,7 +58,7 @@ func NewCohere(apiKey string, optFns ...func(o *CohereOptions)) (*Cohere, error)
 	}, nil
 }
 
-func (l *Cohere) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (l *Cohere) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	opts := schema.GenerateOptions{}
 
 	res, err := l.client.Generate(cohere.GenerateOptions{
@@ -70,7 +70,7 @@ func (l *Cohere) Generate(ctx context.Context, prompts []string, optFns ...func(
 		return nil, err
 	}
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{schema.Generation{Text: res.Generations[0].Text}}},
 		LLMOutput:   map[string]any{},
 	}, nil

diff --git a/model/llm/fake.go b/model/llm/fake.go
@@ -20,8 +20,8 @@ func NewFake(response string) *Fake {
 	}
 }
 
-func (l *Fake) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
-	return &schema.LLMResult{
+func (l *Fake) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{schema.Generation{Text: l.response}}},
 		LLMOutput:   map[string]any{},
 	}, nil

diff --git a/model/llm/hugging_face_hub.go b/model/llm/hugging_face_hub.go
@@ -59,7 +59,7 @@ func NewHuggingFaceHub(apiToken string, optFns ...func(o *HuggingFaceHubOptions)
 	}, nil
 }
 
-func (l *HuggingFaceHub) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (l *HuggingFaceHub) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	var (
 		text string
 		err  error
@@ -79,7 +79,7 @@ func (l *HuggingFaceHub) Generate(ctx context.Context, prompts []string, optFns
 		return nil, err
 	}
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: [][]schema.Generation{{schema.Generation{Text: text}}},
 		LLMOutput:   map[string]any{},
 	}, nil

diff --git a/model/llm/openai.go b/model/llm/openai.go
@@ -77,7 +77,7 @@ func NewOpenAI(apiKey string, optFns ...func(o *OpenAIOptions)) (*OpenAI, error)
 	}, nil
 }
 
-func (l *OpenAI) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (l *OpenAI) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	opts := schema.GenerateOptions{}
 
 	subPromps := util.ChunkBy(prompts, l.opts.BatchSize)
@@ -159,7 +159,7 @@ func (l *OpenAI) Generate(ctx context.Context, prompts []string, optFns ...func(
 		})
 	})
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: generations,
 		LLMOutput: map[string]any{
 			"ModelName":  l.opts.ModelName,

diff --git a/model/llm/sagemaker_endpoint.go b/model/llm/sagemaker_endpoint.go
@@ -101,7 +101,7 @@ func NewSagemakerEndpoint(client *sagemakerruntime.Client, endpointName string,
 	}, nil
 }
 
-func (l *SagemakerEndpoint) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.LLMResult, error) {
+func (l *SagemakerEndpoint) Generate(ctx context.Context, prompts []string, optFns ...func(o *schema.GenerateOptions)) (*schema.ModelResult, error) {
 	generations := [][]schema.Generation{}
 
 	for _, prompt := range prompts {
@@ -130,7 +130,7 @@ func (l *SagemakerEndpoint) Generate(ctx context.Context, prompts []string, optF
 		}})
 	}
 
-	return &schema.LLMResult{
+	return &schema.ModelResult{
 		Generations: generations,
 		LLMOutput:   map[string]any{},
 	}, nil