[Search] [Playground ]handle error LLM edge cases (#189509)

## Summary

Fixes:
- sending the error message as part of the conversation. The LLM picks
up on it and reduces the effectiveness of the answer
- clear chat is now enabled when error has occured
- better error messages from retriever and LLM

### Checklist

Delete any items that are not applicable to this PR.

- [ ] Any text added follows [EUI's writing
guidelines](https://elastic.github.io/eui/#/guidelines/writing), uses
sentence case text and includes [i18n
support](https://github.com/elastic/kibana/blob/main/packages/kbn-i18n/README.md)
- [ ]
[Documentation](https://www.elastic.co/guide/en/kibana/master/development-documentation.html)
was added for features that require explanation or tutorials
- [x] [Unit or functional
tests](https://www.elastic.co/guide/en/kibana/master/development-tests.html)
were updated or added to match the most common scenarios
- [ ] [Flaky Test
Runner](https://ci-stats.kibana.dev/trigger_flaky_test_runner/1) was
used on any tests changed
- [ ] Any UI touched in this PR is usable by keyboard only (learn more
about [keyboard accessibility](https://webaim.org/techniques/keyboard/))
- [ ] Any UI touched in this PR does not create any new axe failures
(run axe in browser:
[FF](https://addons.mozilla.org/en-US/firefox/addon/axe-devtools/),
[Chrome](https://chrome.google.com/webstore/detail/axe-web-accessibility-tes/lhdoppojpmngadmnindnejefpokejbdd?hl=en-US))
- [ ] If a plugin configuration key changed, check if it needs to be
allowlisted in the cloud and added to the [docker
list](https://github.com/elastic/kibana/blob/main/src/dev/build/tasks/os_packages/docker_generator/resources/base/bin/kibana-docker)
- [ ] This renders correctly on smaller devices using a responsive
layout. (You can test this [in your
browser](https://www.browserstack.com/guide/responsive-testing-on-local-server))
- [ ] This was checked for [cross-browser
compatibility](https://www.elastic.co/support/matrix#matrix_browsers)


### Risk Matrix

Delete this section if it is not applicable to this PR.

Before closing this PR, invite QA, stakeholders, and other developers to
identify risks that should be tested prior to the change/feature
release.

When forming the risk matrix, consider some of the following examples
and how they may potentially impact the change:

| Risk | Probability | Severity | Mitigation/Notes |

|---------------------------|-------------|----------|-------------------------|
| Multiple Spaces—unexpected behavior in non-default Kibana Space.
| Low | High | Integration tests will verify that all features are still
supported in non-default Kibana Space and when user switches between
spaces. |
| Multiple nodes—Elasticsearch polling might have race conditions
when multiple Kibana nodes are polling for the same tasks. | High | Low
| Tasks are idempotent, so executing them multiple times will not result
in logical error, but will degrade performance. To test for this case we
add plenty of unit tests around this logic and document manual testing
procedure. |
| Code should gracefully handle cases when feature X or plugin Y are
disabled. | Medium | High | Unit tests will verify that any feature flag
or plugin combination still results in our service operational. |
| [See more potential risk
examples](https://github.com/elastic/kibana/blob/main/RISK_MATRIX.mdx) |


### For maintainers

- [ ] This was checked for breaking API changes and was [labeled
appropriately](https://www.elastic.co/guide/en/kibana/master/contributing.html#kibana-release-notes-process)

x-pack/plugins/search_playground/public/components/chat.tsx

@@ -56,7 +56,7 @@ export const Chat = () => {
     handleSubmit,
     getValues,
   } = useFormContext<ChatForm>();
-  const { messages, append, stop: stopRequest, setMessages, reload, error } = useChat();
+  const { messages, append, stop: stopRequest, setMessages, reload } = useChat();
   const messagesRef = useAutoBottomScroll();
   const [isRegenerating, setIsRegenerating] = useState<boolean>(false);
   const usageTracker = useUsageTracker();
@@ -88,8 +88,8 @@ export const Chat = () => {
   );
 
   const isToolBarActionsDisabled = useMemo(
-    () => chatMessages.length <= 1 || !!error || isRegenerating || isSubmitting,
-    [chatMessages, error, isSubmitting, isRegenerating]
+    () => chatMessages.length <= 1 || isRegenerating || isSubmitting,
+    [chatMessages, isSubmitting, isRegenerating]
   );
 
   const regenerateMessages = async () => {

x-pack/plugins/search_playground/server/lib/conversational_chain.test.ts

@@ -305,6 +305,45 @@ describe('conversational chain', () => {
     });
   }, 10000);
 
+  it('should omit the system messages in chat', async () => {
+    await createTestChain({
+      responses: ['the final answer'],
+      chat: [
+        {
+          id: '1',
+          role: MessageRole.user,
+          content: 'what is the work from home policy?',
+        },
+        {
+          id: '2',
+          role: MessageRole.system,
+          content: 'Error occurred. Please try again.',
+        },
+      ],
+      expectedFinalAnswer: 'the final answer',
+      expectedDocs: [
+        {
+          documents: [
+            { metadata: { _id: '1', _index: 'index' }, pageContent: 'value' },
+            { metadata: { _id: '1', _index: 'website' }, pageContent: 'value2' },
+          ],
+          type: 'retrieved_docs',
+        },
+      ],
+      expectedTokens: [
+        { type: 'context_token_count', count: 15 },
+        { type: 'prompt_token_count', count: 28 },
+      ],
+      expectedSearchRequest: [
+        {
+          method: 'POST',
+          path: '/index,website/_search',
+          body: { query: { match: { field: 'what is the work from home policy?' } }, size: 3 },
+        },
+      ],
+    });
+  }, 10000);
+
   it('should cope with quotes in the query', async () => {
     await createTestChain({
       responses: ['rewrite "the" question', 'the final answer'],

x-pack/plugins/search_playground/server/lib/conversational_chain.ts

@@ -18,7 +18,7 @@ import { createStreamDataTransformer, experimental_StreamData } from 'ai';
 import { BaseLanguageModel } from '@langchain/core/language_models/base';
 import { BaseMessage } from '@langchain/core/messages';
 import { HumanMessage, AIMessage } from '@langchain/core/messages';
-import { ChatMessage, MessageRole } from '../types';
+import { ChatMessage } from '../types';
 import { ElasticsearchRetriever } from './elasticsearch_retriever';
 import { renderTemplate } from '../utils/render_template';
 
@@ -49,25 +49,28 @@ interface ContextInputs {
   question: string;
 }
 
-const getSerialisedMessages = (chatHistory: ChatMessage[]) => {
+const getSerialisedMessages = (chatHistory: BaseMessage[]) => {
   const formattedDialogueTurns = chatHistory.map((message) => {
-    if (message.role === MessageRole.user) {
+    if (message instanceof HumanMessage) {
       return `Human: ${message.content}`;
-    } else if (message.role === MessageRole.assistant) {
+    } else if (message instanceof AIMessage) {
       return `Assistant: ${message.content}`;
     }
   });
   return formattedDialogueTurns.join('\n');
 };
 
-const getMessages = (chatHistory: ChatMessage[]) => {
-  return chatHistory.map((message) => {
-    if (message.role === 'human') {
-      return new HumanMessage(message.content);
-    } else {
-      return new AIMessage(message.content);
-    }
-  });
+export const getMessages = (chatHistory: ChatMessage[]) => {
+  return chatHistory
+    .map((message) => {
+      if (message.role === 'human') {
+        return new HumanMessage(message.content);
+      } else if (message.role === 'assistant') {
+        return new AIMessage(message.content);
+      }
+      return null;
+    })
+    .filter((message): message is BaseMessage => message !== null);
 };
 
 const buildContext = (docs: Document[]) => {
@@ -141,8 +144,9 @@ class ConversationalChainFn {
     const data = new experimental_StreamData();
 
     const messages = msgs ?? [];
-    const previousMessages = messages.slice(0, -1);
-    const question = messages[messages.length - 1]!.content;
+    const lcMessages = getMessages(messages);
+    const previousMessages = lcMessages.slice(0, -1);
+    const question = lcMessages[lcMessages.length - 1]!.content;
     const retrievedDocs: Document[] = [];
 
     let retrievalChain: Runnable = RunnableLambda.from(() => '');
@@ -165,7 +169,7 @@ class ConversationalChainFn {
       return input.question;
     });
 
-    if (previousMessages.length > 0) {
+    if (lcMessages.length > 1) {
       const questionRewritePromptTemplate = PromptTemplate.fromTemplate(
         this.options.questionRewritePrompt
       );
@@ -184,7 +188,6 @@ class ConversationalChainFn {
       });
     }
 
-    const lcMessages = getMessages(messages);
     const prompt = ChatPromptTemplate.fromMessages([
       SystemMessagePromptTemplate.fromTemplate(this.options.prompt),
       ...lcMessages,

x-pack/plugins/search_playground/server/routes.ts

@@ -31,7 +31,7 @@ export function createRetriever(esQuery: string) {
       const query = JSON.parse(replacedQuery);
       return query;
     } catch (e) {
-      throw Error(e);
+      throw Error("Failed to parse the Elasticsearch Query. Check Query to make sure it's valid.");
     }
   };
 }