[8.x] [Security Assistant] Abort signal fix (#203041) (#203187)

# Backport

This will backport the following commits from `main` to `8.x`:
- [[Security Assistant] Abort signal fix
(#203041)](#203041)

<!--- Backport version: 9.4.3 -->

### Questions ?
Please refer to the [Backport tool
documentation](https://github.com/sqren/backport)

<!--BACKPORT [{"author":{"name":"Steph
Milovic","email":"[email protected]"},"sourceCommit":{"committedDate":"2024-12-05T21:27:29Z","message":"[Security
Assistant] Abort signal fix
(#203041)","sha":"b3b2c1745ac846b377d8e234526aa965196e4de9","branchLabelMapping":{"^v9.0.0$":"main","^v8.18.0$":"8.x","^v(\\d+).(\\d+).\\d+$":"$1.$2"}},"sourcePullRequest":{"labels":["release_note:skip","v9.0.0","Team:
SecuritySolution","backport:prev-major","Team:Security Generative
AI"],"title":"[Security Assistant] Abort signal
fix","number":203041,"url":"https://github.com/elastic/kibana/pull/203041","mergeCommit":{"message":"[Security
Assistant] Abort signal fix
(#203041)","sha":"b3b2c1745ac846b377d8e234526aa965196e4de9"}},"sourceBranch":"main","suggestedTargetBranches":[],"targetPullRequestStates":[{"branch":"main","label":"v9.0.0","branchLabelMappingKey":"^v9.0.0$","isSourceBranch":true,"state":"MERGED","url":"https://github.com/elastic/kibana/pull/203041","number":203041,"mergeCommit":{"message":"[Security
Assistant] Abort signal fix
(#203041)","sha":"b3b2c1745ac846b377d8e234526aa965196e4de9"}}]}]
BACKPORT-->

Co-authored-by: Steph Milovic <[email protected]>

x-pack/packages/kbn-langchain/server/language_models/chat_vertex/chat_vertex.ts

@@ -93,6 +93,7 @@ export class ActionsClientChatVertexAI extends ChatVertexAI {
             tools: data?.tools,
             temperature: this.temperature,
             ...systemInstruction,
+            signal: options?.signal,
           },
         },
       };

x-pack/packages/kbn-langchain/server/language_models/chat_vertex/connection.ts

@@ -82,6 +82,7 @@ export class ActionsClientChatConnection<Auth> extends ChatConnection<Auth> {
               tools: data?.tools,
               temperature: this.temperature,
               ...systemInstruction,
+              signal: options?.signal,
             },
           },
         };

x-pack/plugins/elastic_assistant/server/lib/langchain/graphs/default_assistant_graph/graph.ts

@@ -34,6 +34,7 @@ export interface GetDefaultAssistantGraphParams {
   dataClients?: AssistantDataClients;
   createLlmInstance: () => BaseChatModel;
   logger: Logger;
+  signal?: AbortSignal;
   tools: StructuredTool[];
   replacements: Replacements;
 }
@@ -45,6 +46,8 @@ export const getDefaultAssistantGraph = ({
   dataClients,
   createLlmInstance,
   logger,
+  // some chat models (bedrock) require a signal to be passed on agent invoke rather than the signal passed to the chat model
+  signal,
   tools,
   replacements,
 }: GetDefaultAssistantGraphParams) => {
@@ -137,11 +140,19 @@ export const getDefaultAssistantGraph = ({
         })
       )
       .addNode(NodeType.AGENT, (state: AgentState) =>
-        runAgent({ ...nodeParams, state, agentRunnable, kbDataClient: dataClients?.kbDataClient })
+        runAgent({
+          ...nodeParams,
+          config: { signal },
+          state,
+          agentRunnable,
+          kbDataClient: dataClients?.kbDataClient,
+        })
+      )
+      .addNode(NodeType.TOOLS, (state: AgentState) =>
+        executeTools({ ...nodeParams, config: { signal }, state, tools })
       )
-      .addNode(NodeType.TOOLS, (state: AgentState) => executeTools({ ...nodeParams, state, tools }))
       .addNode(NodeType.RESPOND, (state: AgentState) =>
-        respond({ ...nodeParams, state, model: createLlmInstance() })
+        respond({ ...nodeParams, config: { signal }, state, model: createLlmInstance() })
       )
       .addNode(NodeType.MODEL_INPUT, (state: AgentState) => modelInput({ ...nodeParams, state }))
       .addEdge(START, NodeType.MODEL_INPUT)

x-pack/plugins/elastic_assistant/server/lib/langchain/graphs/default_assistant_graph/helpers.ts

@@ -160,10 +160,7 @@ export const streamGraph = async ({
               finalMessage += msg.content;
             }
           } else if (event.event === 'on_llm_end' && !didEnd) {
-            const generations = event.data.output?.generations[0];
-            if (generations && generations[0]?.generationInfo.finish_reason === 'stop') {
-              handleStreamEnd(generations[0]?.text ?? finalMessage);
-            }
+            handleStreamEnd(event.data.output?.generations[0][0]?.text ?? finalMessage);
           }
         }
       }

x-pack/plugins/elastic_assistant/server/lib/langchain/graphs/default_assistant_graph/index.ts

@@ -173,6 +173,7 @@ export const callAssistantGraph: AgentExecutor<true | false> = async ({
     // we need to pass it like this or streaming does not work for bedrock
     createLlmInstance,
     logger,
+    signal: abortSignal,
     tools,
     replacements,
   });

x-pack/plugins/elastic_assistant/server/lib/langchain/graphs/default_assistant_graph/nodes/respond.ts

@@ -7,16 +7,19 @@
 
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { StringWithAutocomplete } from '@langchain/core/dist/utils/types';
+import { RunnableConfig } from '@langchain/core/runnables';
 import { AGENT_NODE_TAG } from './run_agent';
 import { AgentState, NodeParamsBase } from '../types';
 import { NodeType } from '../constants';
 
 export interface RespondParams extends NodeParamsBase {
   state: AgentState;
   model: BaseChatModel;
+  config?: RunnableConfig;
 }
 
 export async function respond({
+  config,
   logger,
   state,
   model,
@@ -34,7 +37,7 @@ export async function respond({
 
     const responseMessage = await model
       // use AGENT_NODE_TAG to identify as agent node for stream parsing
-      .withConfig({ runName: 'Summarizer', tags: [AGENT_NODE_TAG] })
+      .withConfig({ runName: 'Summarizer', tags: [AGENT_NODE_TAG], signal: config?.signal })
       .invoke([userMessage]);
 
     return {

x-pack/plugins/elastic_assistant/server/lib/langchain/graphs/default_assistant_graph/nodes/run_agent.ts

@@ -43,21 +43,22 @@ export async function runAgent({
   logger.debug(() => `${NodeType.AGENT}: Node state:\n${JSON.stringify(state, null, 2)}`);
 
   const knowledgeHistory = await kbDataClient?.getRequiredKnowledgeBaseDocumentEntries();
-
-  const agentOutcome = await agentRunnable.withConfig({ tags: [AGENT_NODE_TAG] }).invoke(
-    {
-      ...state,
-      knowledge_history: `${KNOWLEDGE_HISTORY_PREFIX}\n${
-        knowledgeHistory?.length
-          ? JSON.stringify(knowledgeHistory.map((e) => e.text))
-          : NO_KNOWLEDGE_HISTORY
-      }`,
-      // prepend any user prompt (gemini)
-      input: formatLatestUserMessage(state.input, state.llmType),
-      chat_history: state.messages, // TODO: Message de-dupe with ...state spread
-    },
-    config
-  );
+  const agentOutcome = await agentRunnable
+    .withConfig({ tags: [AGENT_NODE_TAG], signal: config?.signal })
+    .invoke(
+      {
+        ...state,
+        knowledge_history: `${KNOWLEDGE_HISTORY_PREFIX}\n${
+          knowledgeHistory?.length
+            ? JSON.stringify(knowledgeHistory.map((e) => e.text))
+            : NO_KNOWLEDGE_HISTORY
+        }`,
+        // prepend any user prompt (gemini)
+        input: formatLatestUserMessage(state.input, state.llmType),
+        chat_history: state.messages, // TODO: Message de-dupe with ...state spread
+      },
+      config
+    );
 
   return {
     agentOutcome,