From 710f120e38ee16ec39e17661cde2abe0aa89ec99 Mon Sep 17 00:00:00 2001
From: jerryjzhang <jerryjzhang@tencent.com>
Date: Fri, 17 May 2024 15:53:10 +0800
Subject: [PATCH] (improvement)(assembly)Improve build scripts

---
 assembly/bin/supersonic-build.sh                  | 11 ++++++-----
 .../headless/core/chat/parser/SemanticParser.java |  6 +++---
 .../core/chat/parser/llm/LLMSqlParser.java        | 15 ++++++++++-----
 3 files changed, 19 insertions(+), 13 deletions(-)

diff --git a/assembly/bin/supersonic-build.sh b/assembly/bin/supersonic-build.sh
index 4aeae14e2..39fd75a7e 100755
--- a/assembly/bin/supersonic-build.sh
+++ b/assembly/bin/supersonic-build.sh
@@ -48,7 +48,7 @@ function packageRelease {
   # package webapp
   tar xvf supersonic-webapp.tar.gz
   mv supersonic-webapp webapp
-  json='{"env": "'$model_name'"}'
+  json='{"env": "''"}'
   echo $json > webapp/supersonic.config.json
   mv webapp $release_dir/
   # package java service
@@ -57,9 +57,8 @@ function packageRelease {
   # generate zip file
   zip -r $release_dir.zip $release_dir
   # delete intermediate files
-  rm -rf supersonic-webapp.tar.gz
-  rm -rf $service_name-bin.tar.gz
-  rm -rf $service_name
+  rm supersonic-webapp.tar.gz $service_name-bin.tar.gz
+  rm -rf webapp $service_name $release_dir
   echo "finished packaging supersonic release"
 }
 
@@ -71,7 +70,9 @@ if [ "$service" == $PYLLM_SERVICE ]; then
   echo "install python modules success"
 elif [ "$service" == "webapp" ]; then
   buildWebapp
-  cp -fr webapp $projectDir/launchers/$STANDALONE_SERVICE/target/classes
+  target_path=$projectDir/launchers/$STANDALONE_SERVICE/target/classes
+  tar xvf $projectDir/webapp/supersonic-webapp.tar.gz -C $target_path
+  mv $target_path/supersonic_webapp $target_path/webapp
 else
   buildJavaService $service
   buildWebapp
diff --git a/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/SemanticParser.java b/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/SemanticParser.java
index 8e57cb973..93554ebe1 100644
--- a/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/SemanticParser.java
+++ b/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/SemanticParser.java
@@ -4,9 +4,9 @@ import com.tencent.supersonic.headless.core.pojo.ChatContext;
 import com.tencent.supersonic.headless.core.pojo.QueryContext;
 
 /**
- * A semantic parser understands user queries and extracts semantic information.
- * It could leverage either rule-based or LLM-based approach to identify query intent
- * and extract related semantic items from the query.
+ * A semantic parser understands user queries and generates semantic query statement.
+ * SuperSonic leverages a combination of rule-based and LLM-based parsers,
+ * each of which deals with specific scenarios.
  */
 public interface SemanticParser {
 
diff --git a/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/llm/LLMSqlParser.java b/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/llm/LLMSqlParser.java
index a813f8cc9..d84a65712 100644
--- a/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/llm/LLMSqlParser.java
+++ b/headless/core/src/main/java/com/tencent/supersonic/headless/core/chat/parser/llm/LLMSqlParser.java
@@ -17,6 +17,10 @@ import java.util.List;
 import java.util.Map;
 import java.util.Objects;
 
+/**
+ * LLMSqlParser uses large language model to understand query semantics and
+ * generate S2SQL statements to be executed by the semantic query engine.
+ */
 @Slf4j
 public class LLMSqlParser implements SemanticParser {
 
@@ -30,20 +34,21 @@ public class LLMSqlParser implements SemanticParser {
         try {
             //2.get dataSetId from queryCtx and chatCtx.
             Long dataSetId = requestService.getDataSetId(queryCtx);
-            log.info("dataSetId:{}", dataSetId);
             if (dataSetId == null) {
                 return;
             }
-            //3.construct a request, call the API for the large model, and retrieve the results.
+            log.info("Generate query statement for dataSetId:{}", dataSetId);
+
+            //3.invoke LLM service to do parsing.
             List<LLMReq.ElementValue> linkingValues = requestService.getValueList(queryCtx, dataSetId);
             SemanticSchema semanticSchema = queryCtx.getSemanticSchema();
             LLMReq llmReq = requestService.getLlmReq(queryCtx, dataSetId, semanticSchema, linkingValues);
             LLMResp llmResp = requestService.requestLLM(llmReq, dataSetId);
-
             if (Objects.isNull(llmResp)) {
                 return;
             }
-            //4. deduplicate the SQL result list and build parserInfo
+
+            //4. deduplicate the S2SQL result list and build parserInfo
             LLMResponseService responseService = ContextUtils.getBean(LLMResponseService.class);
             Map<String, LLMSqlResp> deduplicationSqlResp = responseService.getDeduplicationSqlResp(llmResp);
             ParseResult parseResult = ParseResult.builder()
@@ -66,7 +71,7 @@ public class LLMSqlParser implements SemanticParser {
             }
 
         } catch (Exception e) {
-            log.error("parse", e);
+            log.error("Failed to parse query:", e);
         }
     }