fix: add cmd & test case

app/pages/Console/OutputBox/ForceGraph/index.tsx

@@ -45,8 +45,13 @@ const ForceGraphBox = (props: IProps) => {
 
   useEffect(() => {
     init();
+    const resizeObserver = new ResizeObserver(() => {
+      graphs[uuid].resize();
+    });
+    resizeObserver.observe(grapfDomRef.current);
     return () => {
       clearGraph(uuid);
+      resizeObserver.disconnect();
     };
   }, []);
 

app/pages/Console/OutputBox/index.module.less

@@ -145,7 +145,6 @@
               padding-left: 0;
               position: relative;
               background-color: @lightBlue;
-              max-height: 720px;
               overflow: auto;
               height: 100%;
             }

app/pages/LLMBot/chat.tsx

@@ -97,7 +97,7 @@ function Chat() {
         </div>
       );
     }
-    const gqls = message.content.split(/```([^`]+)```/);
+    const gqls = message.content.split(/```\w*\n([^`]+)```/);
     return gqls.map((item, index) => {
       if (index % 2 === 0) {
         return <p key={index}>{item}</p>;

app/stores/llm.ts

@@ -45,6 +45,27 @@ Return the results directly, without explain and comment. The results should be
 The name of the nodes should be an actual object and a noun.
 Result:
 `;
+
+export const docFinderPrompt = `Assume your are doc finder,from the following graph database book categories:
+"{category_string}"
+user current space is: {space_name}
+find top two useful categories to solve the question:"{query_str}",
+don't explain, if you can't find, return "Sorry".
+just return the two combined categories, separated by ',' is:`;
+
+export const text2queryPrompt = `Assume you are a NebulaGraph AI chat asistant to help user write NGQL.
+You have access to the following information:
+the user space schema is:
+----
+{schema}
+ ----
+the doc is: \n
+----
+{doc}
+----
+you need use markdown to reply short and clearly.  add \`\`\` for markdown code block to write the ngql. one ngql need be one line.
+please use user's language to answer the question: {query_str}`;
+
 export const AgentTask = `Assume you are a NebulaGraph AI chat asistant. You need to help the user to write NGQL or solve other question. 
 You have access to the following information:
 1. The user's console NGQL context is: {current_ngql}
@@ -261,11 +282,10 @@ class LLM {
           messages: [
             {
               role: 'user',
-              content: `Assume your are doc finder,from the following graph database book categories:
-               "${ngqlDoc.NGQLCategoryString}"
-              find top two useful categories to solve the question:"${text}",
-              don't explain, if you can't find, return "Sorry".
-              just return the two combined categories, separated by ',' is:`,
+              content: docFinderPrompt
+                .replace('{category_string}', ngqlDoc.NGQLCategoryString)
+                .replace('{query_str}', text)
+                .replace('{space_name}', rootStore.console.currentSpace),
             },
           ],
         },
@@ -289,19 +309,12 @@ class LLM {
           doc = doc.replaceAll(/\n\n+/g, '');
           if (doc.length) {
             console.log('docString:', doc);
-            prompt = `learn the below doc, and use it to help user ,the user space schema is "{schema}" the doc is: \n${doc.slice(
-              0,
-              this.config.maxContextLength,
-            )} the question is "{query_str}"`;
+            prompt = text2queryPrompt.replace('{doc}', doc.slice(0, this.config.maxContextLength));
           }
         }
       }
     }
     prompt = prompt.replace('{query_str}', text);
-    prompt = `you need use markdown to reply short and clearly. 
-    add \`\`\` for markdown code block to write the ngql.
-    one ngql need be one line. use user question language to reply.
-    ${prompt}`;
 
     const pathname = window.location.pathname;
     const space = pathname.indexOf('schema') > -1 ? rootStore.schema.currentSpace : rootStore.console.currentSpace;

app/stores/twoGraph.ts

@@ -84,7 +84,7 @@ class TwoGraph {
     Graph.d3Force('x', forceX().strength(0.0085));
     Graph.d3Force('y', forceY().strength(0.0085));
     Graph.d3Force('charge')!.strength(-100);
-    Graph.width(rect.width).height(700);
+    Graph.width(rect.width).height(rect.height);
     Graph.onZoom((v) => {
       this.setTransform(v);
       graph.setPointer({

app/utils/ngql.ts

@@ -5,9 +5,10 @@ const urlTransformerMap = {
   MATCH: 'MatchOrCypherOrPatternOrScan',
   GO: 'GOFromVID',
   FIND: 'FindPath',
+  'Schema-related functions': 'type_src_dst_rank_edge-functions',
 };
 
-const extralPaths = ['graph-modeling', 'ngql-guide'];
+const extralPaths = ['graph-modeling', 'ngql-guide', 'use-importer'];
 export const ngqlDoc = (ngqlJson as { url: string; content: string; title: string }[])
   .map((item) => {
     if (urlTransformerMap[item.title]) {

server/api/studio/Makefile

@@ -19,6 +19,9 @@ all: build
 
 gen: gen-api
 
+build-ai-importer:
+	$(GO_BUILD) -ldflags '$(LDFLAGS)' -o bin/ai-importer ./cmd/ai_importer.go
+
 gen-api: $(GOBIN)/goctl
 	rm -rf internal/handler
 	goctl api format -dir ./restapi

server/api/studio/cmd/ai_importer.go

@@ -0,0 +1,189 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"log"
+	"path"
+	"path/filepath"
+	"time"
+
+	nebula_go "github.com/vesoft-inc/nebula-go/v3"
+	"github.com/vesoft-inc/nebula-studio/server/api/studio/internal/config"
+	db "github.com/vesoft-inc/nebula-studio/server/api/studio/internal/model"
+	"github.com/vesoft-inc/nebula-studio/server/api/studio/pkg/auth"
+	"github.com/vesoft-inc/nebula-studio/server/api/studio/pkg/base"
+	"github.com/vesoft-inc/nebula-studio/server/api/studio/pkg/client"
+	"github.com/vesoft-inc/nebula-studio/server/api/studio/pkg/llm"
+	"github.com/zeromicro/go-zero/core/conf"
+)
+
+type Config struct {
+	LLMJob struct {
+		Space          string
+		File           string
+		PromptTemplate string
+	}
+	Auth struct {
+		Address  string
+		Port     int
+		Username string
+		Password string
+	}
+	LLMConfig struct {
+		URL                string
+		Key                string
+		APIType            db.APIType
+		ContextLengthLimit int
+	}
+	GQLBatchSize int `json:",default=100"`
+	MaxBlockSize int `json:",default=0"`
+}
+
+func main() {
+	configFile := flag.String("config", "config.yaml", "Configuration file for the import job")
+	outputPath := flag.String("output", ".", "Output path for the import job")
+	flag.Parse()
+	if *configFile == "" {
+		log.Fatal("config file is empty")
+	}
+
+	var c Config
+	conf.MustLoad(*configFile, &c, conf.UseEnv())
+
+	job := llm.ImportJob{
+		CacheNodes: make(map[string]llm.Node),
+		CacheEdges: make(map[string]map[string]llm.Edge),
+		LLMJob: &db.LLMJob{
+			JobID:          fmt.Sprintf("%d", time.Now().UnixNano()),
+			Space:          c.LLMJob.Space,
+			File:           c.LLMJob.File,
+			PromptTemplate: c.LLMJob.PromptTemplate,
+		},
+		AuthData: &auth.AuthData{
+			Address:  c.Auth.Address,
+			Port:     c.Auth.Port,
+			Username: c.Auth.Username,
+			Password: c.Auth.Password,
+		},
+		LLMConfig: &llm.LLMConfig{
+			URL:                c.LLMConfig.URL,
+			Key:                c.LLMConfig.Key,
+			APIType:            c.LLMConfig.APIType,
+			ContextLengthLimit: c.LLMConfig.ContextLengthLimit,
+		},
+	}
+	studioConfig := config.Config{
+		LLM: struct {
+			GQLPath      string `json:",default=./data/llm"`
+			GQLBatchSize int    `json:",default=100"`
+			MaxBlockSize int    `json:",default=0"`
+		}{
+			GQLPath:      *outputPath,
+			GQLBatchSize: c.GQLBatchSize,
+			MaxBlockSize: c.MaxBlockSize,
+		},
+	}
+	studioConfig.InitConfig()
+	RunFileJob(&job)
+}
+
+func RunFileJob(llmJob *llm.ImportJob) {
+	llmJob.Process = &base.Process{
+		TotalSize:        0,
+		CurrentSize:      0,
+		Ratio:            0,
+		PromptTokens:     0,
+		CompletionTokens: 0,
+	}
+
+	err := llmJob.AddLogFile()
+	if err != nil {
+		llmJob.SetJobFailed(err)
+		return
+	}
+	defer llmJob.CloseLogFile()
+	defer func() {
+		if err := recover(); err != nil {
+			llmJob.WriteLogFile(fmt.Sprintf("panic: %v", err), "error")
+			llmJob.SetJobFailed(err)
+		}
+		if llmJob.LLMJob.Status == base.LLMStatusFailed {
+			log.Fatalf("job failed: %v", llmJob.Process.FailedReason)
+		} else {
+			log.Printf("job %s %s finished", llmJob.LLMJob.JobID, llmJob.LLMJob.Status)
+		}
+	}()
+	go func() {
+		oldRatio := float64(0)
+		for {
+			time.Sleep(3 * time.Second)
+			if oldRatio != llmJob.Process.Ratio {
+				llmJob.WriteLogFile(fmt.Sprintf("process ratio: %f", llmJob.Process.Ratio), "info")
+				oldRatio = llmJob.Process.Ratio
+			}
+		}
+	}()
+
+	llmJob.Process.Ratio = 0.01
+	connectInfo := llmJob.AuthData
+	clientInfo, err := client.NewClient(connectInfo.Address, connectInfo.Port, connectInfo.Username, connectInfo.Password, nebula_go.GetDefaultConf())
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("create client error: %v", err), "error")
+		return
+	}
+	llmJob.NSID = clientInfo.ClientID
+	llmJob.Process.Ratio = 0.03
+
+	err = llmJob.MakeSchema()
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("make schema error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	err = llmJob.GetSchemaMap()
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("get schema map error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	llmJob.Process.Ratio = 0.05
+
+	llmJob.WriteLogFile(fmt.Sprintf("start run file job, file path: %s", llmJob.LLMJob.File), "info")
+
+	filePath := path.Join(llmJob.LLMJob.File)
+	text, err := llmJob.ReadFile(filePath)
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("read file error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	blocks, err := llmJob.SplitText(text)
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("split text error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	llmJob.Process.Ratio = 0.07
+	err = llmJob.QueryBlocks(blocks)
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("query blocks error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	llmJob.Process.Ratio = 0.8
+
+	fileName := filepath.Base(llmJob.LLMJob.File)
+	gqlPath := filepath.Join(config.GetConfig().LLM.GQLPath, fmt.Sprintf("%s/%s.ngql", llmJob.LLMJob.JobID, fileName))
+	gqls, err := llmJob.MakeGQLFile(gqlPath)
+	if err != nil {
+		llmJob.WriteLogFile(fmt.Sprintf("make gql file error: %v", err), "error")
+		llmJob.SetJobFailed(err)
+		return
+	}
+	llmJob.Process.Ratio = 0.9
+
+	llmJob.RunGQLFile(gqls)
+	llmJob.Process.Ratio = 1
+	llmJob.LLMJob.Status = base.LLMStatusSuccess
+}

server/api/studio/cmd/ai_importer_test.go

@@ -0,0 +1,15 @@
+package main
+
+import (
+	"os"
+	"testing"
+)
+
+func TestMain(t *testing.T) {
+	oldArgs := os.Args
+	defer func() { os.Args = oldArgs }()
+
+	os.Args = []string{"cmd", "-config", "../etc/ai-importer.yaml"}
+
+	main()
+}

server/api/studio/etc/ai-importer.yaml

@@ -0,0 +1,33 @@
+LLMJob:
+  Space: "" #space name
+  File: "" #file path,support pdf,txt,json,csv and other text format
+  PromptTemplate: |
+    As a knowledge graph AI importer, your task is to extract useful data from the following text:
+    ----text
+    {text}
+    ----
+
+    the knowledge graph has following schema and node name must be a real :
+    ----graph schema
+    {spaceSchema}
+    ----
+
+    Return the results directly, without explain and comment. The results should be in the following JSON format:
+    {
+      "nodes":[{ "name":string,"type":string,"props":object }],
+      "edges":[{ "src":string,"dst":string,"edgeType":string,"props":object }]
+    }
+    The name of the nodes should be an actual object and a noun.
+    Result:
+Auth:
+  Address: "127.0.0.1" # nebula graphd address
+  Port: 9669
+  Username: "root"
+  Password: "nebula"
+LLMConfig:
+  URL: "" # openai api url
+  Key: "" # openai api key
+  APIType: "openai"
+  ContextLengthLimit: 1024
+MaxBlockSize: 0 # max request block num
+GQLBatchSize: 100 # max gql batch size

server/api/studio/pkg/auth/authorize.go

@@ -29,7 +29,7 @@ type (
 		Port     int    `json:"port"`
 		Username string `json:"username"`
 		Password string `json:"password"`
-		NSID     string `json:"nsid"`
+		NSID     string `json:"nsid,optional"`
 	}
 
 	authClaims struct {