Onboarding General and API Spec Review channels (#12801)

Author: JiaqiZhang-Dev

tools/sdk-ai-bots/azure-sdk-qa-bot-backend/CHANGELOG.md

@@ -1,5 +1,18 @@
 # Release History
 
+## 0.7.0 (unreleased)
+
+### Features Added
+
+* Add new tenants General, API Review, JS, Java and .Net
+* Add new knowledge sources from JS, Java and .Net
+* Support routing tenant based on question domain
+
+## Other changes
+
+* Refine prompt for onboarding channel to improve answer accuracy
+* Refactor post process of merging and sorting and complete chunk logic, to give more better context to LLM
+
 ## 0.6.0 (2025-12-16)
 
 ### Features Added

tools/sdk-ai-bots/azure-sdk-qa-bot-backend/config/env.go

@@ -23,19 +23,21 @@ const (
 )
 
 type Config struct {
-	AOAI_CHAT_REASONING_MODEL      string
-	AOAI_CHAT_COMPLETIONS_MODEL    string
-	AOAI_CHAT_COMPLETIONS_TOP_P    float32
-	AOAI_CHAT_MAX_TOKENS           int
-	AOAI_CHAT_CONTEXT_MAX_TOKENS   int
-	AOAI_CHAT_COMPLETIONS_ENDPOINT string
+	AOAI_CHAT_REASONING_MODEL             string
+	AOAI_CHAT_REASONING_MODEL_TEMPERATURE float32
+	AOAI_CHAT_COMPLETIONS_MODEL           string
+	AOAI_CHAT_COMPLETIONS_TEMPERATURE     float32
+	AOAI_CHAT_MAX_TOKENS                  int
+	AOAI_CHAT_CONTEXT_MAX_TOKENS          int
+	AOAI_CHAT_COMPLETIONS_ENDPOINT        string
 
 	AI_SEARCH_BASE_URL           string
 	AI_SEARCH_INDEX              string
 	AI_SEARCH_AGENT              string
 	AI_SEARCH_KNOWLEDGE_BASE     string
 	AI_SEARCH_KNOWLEDGE_SOURCE   string
 	AI_SEARCH_KNOWLEDGE_BASE_API string
+	AI_SEARCH_TOPK               int
 
 	STORAGE_BASE_URL            string
 	STORAGE_KNOWLEDGE_CONTAINER string

tools/sdk-ai-bots/azure-sdk-qa-bot-backend/config/tenant.go

@@ -9,6 +9,7 @@ type TenantConfig struct {
 	AgenticSearchPrompt     string
 	IntentionPromptTemplate string
 	KeywordReplaceMap       map[string]string
+	EnableRouting           bool
 }
 
 var typespecSources = []model.Source{
@@ -41,18 +42,48 @@ var tenantConfigMap = map[model.TenantID]TenantConfig{
 		IntentionPromptTemplate: "language_python/intention.md",
 		AgenticSearchPrompt:     "language_python/agentic_search.md",
 	},
+	model.TenantID_DotnetChannelQaBot: {
+		Sources: append([]model.Source{model.Source_AzureSDKForNetDocs, model.Source_AzureSDKGuidelines, model.Source_AzureSDKDocsEng}, typespecSources...),
+		SourceFilter: map[model.Source]string{
+			model.Source_AzureSDKGuidelines: "search.ismatch('dotnet_*', 'title')",
+		},
+		PromptTemplate:          "language_channel/qa.md",
+		IntentionPromptTemplate: "language_channel/intention.md",
+		AgenticSearchPrompt:     "language_channel/agentic_search.md",
+	},
 	model.TenantID_GolangChannelQaBot: {
-		Sources: append([]model.Source{model.Source_AzureSDKForGo, model.Source_AzureSDKGuidelines}, typespecSources...),
+		Sources: append([]model.Source{model.Source_AzureSDKForGo, model.Source_AzureSDKGuidelines, model.Source_AzureSDKDocsEng}, typespecSources...),
 		SourceFilter: map[model.Source]string{
 			model.Source_AzureSDKGuidelines: "search.ismatch('golang_*', 'title')",
 		},
-		PromptTemplate:          "common/language_channel.md",
-		IntentionPromptTemplate: "prompt_template/intention.md",
-		AgenticSearchPrompt:     "common/agentic_search.md",
+		PromptTemplate:          "language_channel/qa.md",
+		IntentionPromptTemplate: "language_channel/intention.md",
+		AgenticSearchPrompt:     "language_channel/agentic_search.md",
+	},
+	model.TenantID_JavaChannelQaBot: {
+		Sources: append([]model.Source{model.Source_AzureSDKForJava, model.Source_AzureSDKForJavaWiki, model.Source_AzureSDKGuidelines, model.Source_AutorestJava, model.Source_AzureSDKDocsEng}, typespecSources...),
+		SourceFilter: map[model.Source]string{
+			model.Source_AzureSDKGuidelines: "search.ismatch('java_*', 'title')",
+		},
+		PromptTemplate:          "language_channel/qa.md",
+		IntentionPromptTemplate: "language_channel/intention.md",
+		AgenticSearchPrompt:     "language_channel/agentic_search.md",
+	},
+	model.TenantID_JavaScriptChannelQaBot: {
+		Sources: append([]model.Source{model.Source_AzureSDKForJavaScript, model.Source_AzureSDKForJavaScriptWiki, model.Source_AzureSDKGuidelines, model.Source_AzureSDKDocsEng}, typespecSources...),
+		SourceFilter: map[model.Source]string{
+			model.Source_AzureSDKGuidelines: "search.ismatch('typescript_*', 'title')",
+		},
+		PromptTemplate:          "language_channel/qa.md",
+		IntentionPromptTemplate: "language_channel/intention.md",
+		AgenticSearchPrompt:     "language_channel/agentic_search.md",
 	},
 	model.TenantID_AzureSDKQaBot: {
-		PromptTemplate:          "typespec/qa.md",
-		Sources:                 typespecSources,
+		PromptTemplate: "typespec/qa.md",
+		Sources:        append(typespecSources, model.Source_AzureSDKDocsEng),
+		SourceFilter: map[model.Source]string{
+			model.Source_AzureSDKDocsEng: "search.ismatch('design*', 'title')",
+		},
 		IntentionPromptTemplate: "typespec/intention.md",
 		AgenticSearchPrompt:     "typespec/agentic_search.md",
 	},
@@ -62,6 +93,22 @@ var tenantConfigMap = map[model.TenantID]TenantConfig{
 		AgenticSearchPrompt:     "azure_sdk_onboarding/agentic_search.md",
 		IntentionPromptTemplate: "azure_sdk_onboarding/intention.md",
 	},
+	model.TenantID_GeneralQaBot: {
+		PromptTemplate:          "general/qa.md",
+		IntentionPromptTemplate: "general/intention.md",
+		AgenticSearchPrompt:     "general/agentic_search.md",
+		EnableRouting:           true,
+	},
+	model.TenantID_APISpecReviewBot: {
+		PromptTemplate: "api_spec_review/qa.md",
+		Sources:        []model.Source{model.Source_StaticAzureDocs, model.Source_AzureRestAPISpec, model.Source_AzureSDKDocsEng},
+		SourceFilter: map[model.Source]string{
+			model.Source_AzureSDKDocsEng: "search.ismatch('design*', 'title')",
+		},
+		IntentionPromptTemplate: "api_spec_review/intention.md",
+		AgenticSearchPrompt:     "api_spec_review/agentic_search.md",
+		EnableRouting:           true,
+	},
 }
 
 func GetTenantConfig(tenantID model.TenantID) (TenantConfig, bool) {

tools/sdk-ai-bots/azure-sdk-qa-bot-backend/model/completion.go

@@ -3,11 +3,16 @@ package model
 type TenantID string
 
 const (
-	TenantID_AzureSDKQaBot      TenantID = "azure_sdk_qa_bot" // default as TypeSpec QA bot
-	TenantID_TypeSpecExtension  TenantID = "typespec_extension"
-	TenantID_PythonChannelQaBot TenantID = "python_channel_qa_bot"
-	TenantID_AzureSDKOnboarding TenantID = "azure_sdk_onboarding"
-	TenantID_GolangChannelQaBot TenantID = "golang_channel_qa_bot"
+	TenantID_AzureSDKQaBot          TenantID = "azure_sdk_qa_bot" // default as TypeSpec QA bot
+	TenantID_TypeSpecExtension      TenantID = "typespec_extension"
+	TenantID_PythonChannelQaBot     TenantID = "python_channel_qa_bot"
+	TenantID_DotnetChannelQaBot     TenantID = "dotnet_channel_qa_bot"
+	TenantID_AzureSDKOnboarding     TenantID = "azure_sdk_onboarding"
+	TenantID_GolangChannelQaBot     TenantID = "golang_channel_qa_bot"
+	TenantID_JavaChannelQaBot       TenantID = "java_channel_qa_bot"
+	TenantID_JavaScriptChannelQaBot TenantID = "javascript_channel_qa_bot"
+	TenantID_GeneralQaBot           TenantID = "general_qa_bot"
+	TenantID_APISpecReviewBot       TenantID = "api_spec_review_bot"
 )
 
 type Source string
@@ -29,6 +34,12 @@ const (
 	Source_AzureSDKForGo                  Source = "azure_sdk_for_go_docs"
 	Source_StaticAzureDocs                Source = "static_azure_docs"
 	Source_StaticTypeSpecToSwaggerMapping Source = "static_typespec_to_swagger_mapping"
+	Source_AzureSDKForJava                Source = "azure_sdk_for_java_docs"
+	Source_AzureSDKForJavaWiki            Source = "azure_sdk_for_java_wiki"
+	Source_AutorestJava                   Source = "autorest_java_docs"
+	Source_AzureSDKForJavaScript          Source = "azure_sdk_for_js_docs"
+	Source_AzureSDKForJavaScriptWiki      Source = "azure_sdk_for_js_wiki"
+	Source_AzureSDKForNetDocs             Source = "azure_sdk_for_net_docs"
 )
 
 type Role string
@@ -85,6 +96,7 @@ type CompletionResp struct {
 	FullContext       *string          `json:"full_context" jsonschema:"omitempty,description=The full context used to generate the answer"`
 	Intention         *IntentionResult `json:"intention" jsonschema:"omitempty,description=The intention of the question"`
 	ReasoningProgress *string          `json:"reasoning_progress,omitempty" jsonschema:"omitempty,description=The reasoning progress of generating the answer"`
+	RouteTenant       *TenantID        `json:"route_tenant,omitempty" jsonschema:"omitempty,description=The tenant ID the question is routed to"`
 }
 
 type QuestionScope string
@@ -102,3 +114,7 @@ type IntentionResult struct {
 	Scope              QuestionScope `json:"scope,omitempty" jsonschema:"omitempty,description=The scope of the question"`
 	NeedsRagProcessing bool          `json:"needs_rag_processing" jsonschema:"required,description=Whether to invoke RAG workflow"`
 }
+
+type TenantRoutingResult struct {
+	RouteTenant TenantID `json:"route_tenant" jsonschema:"required,description=The tenant ID to route the question to"`
+}

tools/sdk-ai-bots/azure-sdk-qa-bot-backend/model/search.go

@@ -4,6 +4,32 @@ import (
 	"strings"
 )
 
+// ExpansionType defines how a chunk should be expanded
+type ExpansionType int
+
+const (
+	ExpansionNone         ExpansionType = iota // No expansion needed - use chunk as-is
+	ExpansionHierarchical                      // Expand based on header hierarchy
+	ExpansionQA                                // Expand based on Q&A pairs
+	ExpansionMapping                           // Expand based on TypeSpec to Swagger mapping
+)
+
+// ChunkWithExpansion wraps a chunk with its expansion strategy
+type ChunkWithExpansion struct {
+	Chunk     Index
+	Expansion ExpansionType
+}
+
+// ChunkHierarchy represents the header level of a chunk
+type ChunkHierarchy int
+
+const (
+	HierarchyHeader3 ChunkHierarchy = iota // Full chunk with header3 (most specific)
+	HierarchyHeader2                       // Chunk with header1 and header2
+	HierarchyHeader1                       // Chunk with only header1
+	HierarchyUnknown                       // Unknown or no headers
+)
+
 type AgenticSearchRequest struct {
 	Messages                 []KnowledgeAgentMessage   `json:"messages"`
 	KnowledgeSourceParams    []KnowledgeSourceParams   `json:"knowledgeSourceParams,omitempty"`
@@ -247,11 +273,25 @@ func GetIndexLink(chunk Index) string {
 		return "https://azure.github.io/typespec-azure/docs/migrate-swagger/faq/breakingchange"
 	case Source_AzureSDKForGo:
 		return "https://github.com/Azure/azure-sdk-for-go/blob/main/documentation/" + path
+	case Source_AzureSDKForJava:
+		return "https://github.com/Azure/azure-sdk-for-java/blob/main/" + path
+	case Source_AzureSDKForJavaWiki:
+		path = TrimFileFormat(path)
+		return "https://github.com/Azure/azure-sdk-for-java/wiki/" + path
+	case Source_AutorestJava:
+		return "https://github.com/Azure/autorest.java/blob/main/" + path
 	case Source_StaticAzureDocs:
 		if chunk.Title == "Azure Versioning and Breaking Changes Policy V1.3.2" {
 			return "http://aka.ms/azbreakingchangespolicy"
 		}
 		return ""
+	case Source_AzureSDKForJavaScript:
+		return "https://github.com/Azure/azure-sdk-for-js/blob/main/" + path
+	case Source_AzureSDKForJavaScriptWiki:
+		path = TrimFileFormat(path)
+		return "https://github.com/Azure/azure-sdk-for-js/wiki/" + path
+	case Source_AzureSDKForNetDocs:
+		return "https://github.com/Azure/azure-sdk-for-net/blob/main/" + path
 	default:
 		return ""
 	}