add some simple alerts when a model does not exist (#229)

* add some simple alerts when a model does not exist

* add nav buttons

* switch to using suspense query for amp config

* better error handling for streaming

* bumping crewai configs to be more lenient

* bump max iter time to 120 seconds

* fixed merge conflicts

* bump back to 120

* use the proper model name for the crew planner agent

* changed to runtime error

* unquote crew ai errors

* added more info in tips

* remove unused import

* Update llm-service/app/services/query/agents/crewai_querier.py

Co-authored-by: mliu-cloudera <mliu@cloudera.com>

* remove unquote

---------

Co-authored-by: Baasit Sharief <baasitsharief@gmail.com>
Co-authored-by: jwatson <jkwatson@gmail.com>
Co-authored-by: mliu-cloudera <mliu@cloudera.com>

Author: 4 people

llm-service/app/services/query/agents/crewai_querier.py

@@ -151,8 +151,10 @@ def assemble_crew(
         step_callback=lambda output: step_callback(
             output, "Tool Result", crew_events_queue
         ),
-        max_execution_time=30,
-        max_iter=5,
+        max_execution_time=120,
+        max_iter=15,
+        max_rpm=10,
+        max_retry_limit=5,
     )
 
     # Define tasks for the researcher agents
@@ -236,7 +238,7 @@ def assemble_crew(
             output, "Research Complete", crew_events_queue
         ),
         tools=mcp_tools,
-        max_retries=3,
+        max_retries=5,
         guardrail=validate_with_context,
     )
 
@@ -256,8 +258,10 @@ def assemble_crew(
             output, "Response Computed", crew_events_queue
         ),
         verbose=True,
-        max_execution_time=30,
-        max_iter=5,
+        max_execution_time=120,
+        max_iter=15,
+        max_rpm=10,
+        max_retry_limit=5,
     )
 
     response_context = [research_task, calculation_task]
@@ -309,31 +313,35 @@ def launch_crew(
     query_str: str,
 ) -> Tuple[str, list[Tuple[str, float]]]:
     # Run the crew to get the enhanced response
-    crew_result: CrewOutput = crew.kickoff()
-
-    source_node_ids_w_score = extract_node_ids_from_crew_result(crew_result)
-
-    # Create an enhanced query that includes the CrewAI insights
-    return (
-        f"""
-Original query: {query_str}
-
-Research insights: {crew_result}
-
-Please provide a response to the original query, incorporating the insights from research with in-line citations. \
-
-Adhere to the following guidelines:
-* If you cannot find relevant information in the research insights, answer the question directly and indicate that \
-you don't have enough information. 
-* If citations from the research insights are used, use the in-line links and \
-citations from the research insights as is. Keep markdown formatted links as is i.e. [<text>](<web_link>). \
-Keep the in-line citations of format `<a class='rag_citation' href='node_id'>node_id</a>` as is. 
-* Do not make up any links or citations of the form `<a class='rag_citation' href='node_id'>node_id</a>` \
-that are not present in the research insights. Do not make up any markdown links as well. Only use the \
-links and citations from the research insights. 
-""",
-        source_node_ids_w_score,
-    )
+    try:
+        crew_result: CrewOutput = crew.kickoff()
+
+        source_node_ids_w_score = extract_node_ids_from_crew_result(crew_result)
+
+        # Create an enhanced query that includes the CrewAI insights
+        return (
+            f"""
+    Original query: {query_str}
+    
+    Research insights: {crew_result}
+    
+    Please provide a response to the original query, incorporating the insights from research with in-line citations. \
+    
+    Adhere to the following guidelines:
+    * If you cannot find relevant information in the research insights, answer the question directly and indicate that \
+    you don't have enough information. 
+    * If citations from the research insights are used, use the in-line links and \
+    citations from the research insights as is. Keep markdown formatted links as is i.e. [<text>](<web_link>). \
+    Keep the in-line citations of format `<a class='rag_citation' href='node_id'>node_id</a>` as is. 
+    * Do not make up any links or citations of the form `<a class='rag_citation' href='node_id'>node_id</a>` \
+    that are not present in the research insights. Do not make up any markdown links as well. Only use the \
+    links and citations from the research insights. 
+    """,
+            source_node_ids_w_score,
+        )
+    except Exception as e:
+        logger.exception("Error running CrewAI crew")
+        raise RuntimeError("Error running CrewAI crew: %s" % str(e)) from e
 
 
 def extract_node_ids_from_crew_result(

llm-service/app/services/query/agents/planner_agent.py

@@ -121,7 +121,7 @@ def decide_retrieval_strategy(
             role="Planner",
             goal="Decide whether to use retrieval or answer directly",
             backstory="You are an expert planner who decides the most efficient way to answer a query.",
-            llm=get_crewai_llm_object_direct(self.llm, self.configuration.model_name),
+            llm=get_crewai_llm_object_direct(self.llm, getattr(self.llm, "model", "")),
             # verbose=True,
         )
 

ui/src/api/ampMetadataApi.ts

@@ -185,7 +185,7 @@ export const useGetAmpConfig = () => {
 };
 
 export const useGetPollingAmpConfig = (poll?: boolean) => {
-  return useQuery({
+  return queryOptions({
     queryKey: [QueryKeys.getPollingAmpConfig],
     queryFn: getAmpConfig,
     refetchInterval: () => {

ui/src/api/chatApi.ts

@@ -452,24 +452,32 @@ const streamChatMutation = async (
       }),
       signal: ctrl.signal,
       onmessage(msg: EventSourceMessage) {
-        const data = JSON.parse(msg.data) as ChatMutationResponse;
-
-        if (data.error) {
-          onError(data.error);
+        try {
+          const data = JSON.parse(msg.data) as ChatMutationResponse;
+
+          if (data.error) {
+            onError(data.error);
+            ctrl.abort();
+          }
+
+          if (data.text) {
+            onChunk(data.text);
+          }
+
+          if (data.event) {
+            onEvent(data.event);
+          }
+
+          if (data.response_id) {
+            responseId = data.response_id;
+          }
+        } catch (error) {
+          console.error("Error parsing message data:", error);
+          onError(
+            `An error occurred while processing the response. Original error message: ${JSON.stringify(msg)}. Error in parsing: ${JSON.stringify(error)}`,
+          );
           ctrl.abort();
         }
-
-        if (data.text) {
-          onChunk(data.text);
-        }
-
-        if (data.event) {
-          onEvent(data.event);
-        }
-
-        if (data.response_id) {
-          responseId = data.response_id;
-        }
       },
       onerror(err: unknown) {
         ctrl.abort();

ui/src/layout/Sidebar.tsx

@@ -61,6 +61,7 @@ import "./style.css";
 import AmpUpdateBanner from "src/components/AmpUpdate/AmpUpdateBanner.tsx";
 import { useGetPollingAmpConfig } from "src/api/ampMetadataApi.ts";
 import { getItem } from "./TopNav";
+import { useSuspenseQuery } from "@tanstack/react-query";
 
 const { Sider } = Layout;
 
@@ -71,7 +72,7 @@ const Sidebar: React.FC = () => {
   const matchRoute = useMatchRoute();
   const navigate = useNavigate();
   const ref = useRef<HTMLDivElement>(null);
-  const { data: config } = useGetPollingAmpConfig();
+  const { data: config } = useSuspenseQuery(useGetPollingAmpConfig());
 
   const navToRagApp = () => {
     navigate({ to: "/chats" }).catch(() => null);

ui/src/layout/TopNav.tsx

@@ -54,11 +54,12 @@ import {
   ProjectConfig,
   useGetPollingAmpConfig,
 } from "src/api/ampMetadataApi.ts";
+import { useSuspenseQuery } from "@tanstack/react-query";
 
 const TopNav: React.FC = () => {
   const matchRoute = useMatchRoute();
   const navigate = useNavigate();
-  const { data: config } = useGetPollingAmpConfig();
+  const { data: config } = useSuspenseQuery(useGetPollingAmpConfig());
 
   const navigateTo = (path: string) => () => {
     navigate({ to: path }).catch(() => null);

ui/src/pages/DataSources/DataSourcesManagement/DataSourcesForm.tsx

@@ -37,6 +37,8 @@
  ******************************************************************************/
 
 import {
+  Alert,
+  Button,
   Collapse,
   Divider,
   Form,
@@ -51,6 +53,8 @@ import { ConnectionType, DataSourceBaseType } from "src/api/dataSourceApi";
 import { useGetEmbeddingModels, useGetLlmModels } from "src/api/modelsApi.ts";
 import { useEffect } from "react";
 import { transformModelOptions } from "src/utils/modelUtils.ts";
+import { useNavigate } from "@tanstack/react-router";
+import messageQueue from "src/utils/messageQueue.ts";
 
 export const distanceMetricOptions = [
   {
@@ -143,6 +147,7 @@ const DataSourcesForm = ({
 }: DataSourcesFormProps) => {
   const embeddingsModels = useGetEmbeddingModels();
   const llmModels = useGetLlmModels();
+  const navigate = useNavigate();
 
   useEffect(() => {
     if (initialValues.embeddingModel) {
@@ -160,6 +165,30 @@ const DataSourcesForm = ({
       style={{ width: "100%" }}
       {...layout}
     >
+      {embeddingsModels.isSuccess && embeddingsModels.data.length === 0 ? (
+        <Alert
+          type="warning"
+          showIcon={true}
+          message={
+            "One embedding model must be available to create a knowledge base"
+          }
+          style={{ marginBottom: 16 }}
+          action={
+            <Button
+              onClick={() => {
+                navigate({
+                  to: "/settings",
+                  hash: "modelConfiguration",
+                }).catch(() => {
+                  messageQueue.error("Failed to navigate to models page");
+                });
+              }}
+            >
+              Model Config
+            </Button>
+          }
+        />
+      ) : null}
       <Form.Item
         name="name"
         label="Name"

ui/src/pages/Models/ModelTips.tsx

@@ -84,6 +84,10 @@ const helpText = {
           <li>Cohere Rerank v3.5</li>
           <li>Amazon Rerank v1</li>
         </li>
+        <li>
+          For model capabilities including tool calling, refer to the Amazon
+          Bedrock User Guide.
+        </li>
       </ul>
     </Typography>
   ),

ui/src/pages/RagChatTab/ChatOutput/ChatMessages/ChatMessageController.tsx

@@ -40,11 +40,9 @@ import { useContext, useEffect, useRef } from "react";
 import { useInView } from "react-intersection-observer";
 import ChatMessage from "pages/RagChatTab/ChatOutput/ChatMessages/ChatMessage.tsx";
 import { RagChatContext } from "pages/RagChatTab/State/RagChatContext.tsx";
-import { Image, Skeleton, Typography } from "antd";
-import Images from "src/components/images/Images.ts";
+import { Skeleton } from "antd";
 import PendingRagOutputSkeleton from "pages/RagChatTab/ChatOutput/Loaders/PendingRagOutputSkeleton.tsx";
 import { ChatLoading } from "pages/RagChatTab/ChatOutput/Loaders/ChatLoading.tsx";
-import SuggestedQuestionsCards from "pages/RagChatTab/ChatOutput/Placeholders/SuggestedQuestionsCards.tsx";
 import { useSearch } from "@tanstack/react-router";
 import messageQueue from "src/utils/messageQueue.ts";
 import {
@@ -54,7 +52,7 @@ import {
   useStreamingChatMutation,
 } from "src/api/chatApi.ts";
 import { useRenameNameMutation } from "src/api/sessionApi.ts";
-import NoDataSourcesState from "pages/RagChatTab/ChatOutput/Placeholders/NoDataSourcesState.tsx";
+import EmptyChatState from "pages/RagChatTab/ChatOutput/ChatMessages/EmptyChatState.tsx";
 
 const ChatMessageController = () => {
   const {
@@ -166,21 +164,7 @@ const ChatMessageController = () => {
     if (isFetchingHistory) {
       return <ChatLoading />;
     }
-    return (
-      <>
-        <Image
-          src={Images.BrandTalking}
-          alt="Machines Chatting"
-          style={{ width: 80 }}
-          preview={false}
-        />
-        <Typography.Title level={4} style={{ fontWeight: 300, margin: 0 }}>
-          Welcome to RAG Studio
-        </Typography.Title>
-        <SuggestedQuestionsCards />
-        <NoDataSourcesState />
-      </>
-    );
+    return <EmptyChatState />;
   }
 
   return (

ui/src/pages/RagChatTab/ChatOutput/ChatMessages/EmptyChatState.tsx

@@ -0,0 +1,90 @@
+/*
+ * CLOUDERA APPLIED MACHINE LEARNING PROTOTYPE (AMP)
+ * (C) Cloudera, Inc. 2025
+ * All rights reserved.
+ *
+ * Applicable Open Source License: Apache 2.0
+ *
+ * NOTE: Cloudera open source products are modular software products
+ * made up of hundreds of individual components, each of which was
+ * individually copyrighted.  Each Cloudera open source product is a
+ * collective work under U.S. Copyright Law. Your license to use the
+ * collective work is as provided in your written agreement with
+ * Cloudera.  Used apart from the collective work, this file is
+ * licensed for your use pursuant to the open source license
+ * identified above.
+ *
+ * This code is provided to you pursuant a written agreement with
+ * (i) Cloudera, Inc. or (ii) a third-party authorized to distribute
+ * this code. If you do not have a written agreement with Cloudera nor
+ * with an authorized and properly licensed third party, you do not
+ * have any rights to access nor to use this code.
+ *
+ * Absent a written agreement with Cloudera, Inc. ("Cloudera") to the
+ * contrary, A) CLOUDERA PROVIDES THIS CODE TO YOU WITHOUT WARRANTIES OF ANY
+ * KIND; (B) CLOUDERA DISCLAIMS ANY AND ALL EXPRESS AND IMPLIED
+ * WARRANTIES WITH RESPECT TO THIS CODE, INCLUDING BUT NOT LIMITED TO
+ * IMPLIED WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY AND
+ * FITNESS FOR A PARTICULAR PURPOSE; (C) CLOUDERA IS NOT LIABLE TO YOU,
+ * AND WILL NOT DEFEND, INDEMNIFY, NOR HOLD YOU HARMLESS FOR ANY CLAIMS
+ * ARISING FROM OR RELATED TO THE CODE; AND (D)WITH RESPECT TO YOUR EXERCISE
+ * OF ANY RIGHTS GRANTED TO YOU FOR THE CODE, CLOUDERA IS NOT LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE OR
+ * CONSEQUENTIAL DAMAGES INCLUDING, BUT NOT LIMITED TO, DAMAGES
+ * RELATED TO LOST REVENUE, LOST PROFITS, LOSS OF INCOME, LOSS OF
+ * BUSINESS ADVANTAGE OR UNAVAILABILITY, OR LOSS OR CORRUPTION OF
+ * DATA.
+ */
+
+import { Alert, Button, Image, Typography } from "antd";
+import Images from "src/components/images/Images.ts";
+import SuggestedQuestionsCards from "pages/RagChatTab/ChatOutput/Placeholders/SuggestedQuestionsCards.tsx";
+import NoDataSourcesState from "pages/RagChatTab/ChatOutput/Placeholders/NoDataSourcesState.tsx";
+import { useGetLlmModels } from "src/api/modelsApi.ts";
+import messageQueue from "src/utils/messageQueue.ts";
+import { useNavigate } from "@tanstack/react-router";
+
+const EmptyChatState = () => {
+  const { data: llmModels, isSuccess } = useGetLlmModels();
+  const navigate = useNavigate();
+
+  return (
+    <>
+      {isSuccess && llmModels.length === 0 ? (
+        <Alert
+          type="warning"
+          showIcon
+          message={"One inference model must be available to chat"}
+          action={
+            <Button
+              style={{ marginLeft: 12 }}
+              onClick={() => {
+                navigate({
+                  to: "/settings",
+                  hash: "modelConfiguration",
+                }).catch(() => {
+                  messageQueue.error("Failed to navigate to models page");
+                });
+              }}
+            >
+              Model Config
+            </Button>
+          }
+        />
+      ) : null}
+      <Image
+        src={Images.BrandTalking}
+        alt="Machines Chatting"
+        style={{ width: 80 }}
+        preview={false}
+      />
+      <Typography.Title level={4} style={{ fontWeight: 300, margin: 0 }}>
+        Welcome to RAG Studio
+      </Typography.Title>
+      <SuggestedQuestionsCards />
+      <NoDataSourcesState />
+    </>
+  );
+};
+
+export default EmptyChatState;