diff --git a/lib/llm/bedrock-response.js b/lib/llm/bedrock-response.js
index 4c32c09e25..90ca94907c 100644
--- a/lib/llm/bedrock-response.js
+++ b/lib/llm/bedrock-response.js
@@ -136,7 +136,7 @@ class BedrockResponse {
    * @returns {string}
    */
   get requestId() {
-    return this.#innerOutput.requestId
+    return this.headers?.['x-amzn-requestid']
   }
 
   /**
diff --git a/lib/llm/event.js b/lib/llm/event.js
index 1ef4ecea93..afb9385121 100644
--- a/lib/llm/event.js
+++ b/lib/llm/event.js
@@ -64,7 +64,7 @@ class LlmEvent {
     this.span_id = segment.id
     this.transaction_id = segment.transaction.id
     this.trace_id = segment.transaction.traceId
-    this.request_id = this.bedrockResponse.headers['x-amzn-requestid']
+    this.request_id = this.bedrockResponse.requestId
 
     this['response.model'] = this.bedrockCommand.modelId
     this['request.model'] = this.bedrockCommand.modelId
diff --git a/lib/llm/index.js b/lib/llm/index.js
index 39f8c52a28..7dd577fae2 100644
--- a/lib/llm/index.js
+++ b/lib/llm/index.js
@@ -11,5 +11,6 @@ module.exports = {
   LlmChatCompletionMessage: require('./chat-completion-message'),
   LlmChatCompletionSummary: require('./chat-completion-summary'),
   LlmEmbedding: require('./embedding'),
-  LlmEvent: require('./event')
+  LlmEvent: require('./event'),
+  LlmTrackedIds: require('./tracked-ids')
 }
diff --git a/lib/llm/tracked-ids.js b/lib/llm/tracked-ids.js
new file mode 100644
index 0000000000..77496790f7
--- /dev/null
+++ b/lib/llm/tracked-ids.js
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2024 New Relic Corporation. All rights reserved.
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+'use strict'
+
+/**
+ * Represents the identifiers created and collected when creating LLM
+ * chat completions.
+ *
+ * @property {string[]} [message_ids=[]] Internal identifiers for each message
+ * sent in the conversation.
+ * @property {string} [coversation_id=""] User defined identifier of the chat
+ * completion conversation.
+ * @property {string} request_id Identifier of the request from the remote
+ * service for a chat completion. Populated by the `x-request-id` header
+ * in the request.
+ * @public
+ */
+module.exports = class LlmTrackedIds {
+  constructor({ messageIds, conversationId = '', requestId } = {}) {
+    this.message_ids = messageIds ?? []
+    this.conversation_id = conversationId
+    this.request_id = requestId
+  }
+}
diff --git a/lib/v3/bedrock.js b/lib/v3/bedrock.js
index 7801a4cbde..e53d54af4a 100644
--- a/lib/v3/bedrock.js
+++ b/lib/v3/bedrock.js
@@ -8,6 +8,7 @@ const {
   LlmChatCompletionMessage,
   LlmChatCompletionSummary,
   LlmEmbedding,
+  LlmTrackedIds,
   BedrockCommand,
   BedrockResponse
 } = require('../llm')
@@ -36,6 +37,28 @@ function recordEvent({ agent, type, msg }) {
   agent.customEventAggregator.add([{ type, timestamp: Date.now() }, msg])
 }
 
+/**
+ * Assigns requestId, conversationId and messageIds for a given
+ * chat completion response on the active transaction.
+ * This is used for generating LlmFeedbackEvent via `api.recordLlmFeedbackEvent`
+ *
+ * @param {object} params input params
+ * @param {Transaction} params.tx active transaction
+ * @param {LlmChatCompletionMessage} params.completionMsg chat completion message
+ * @param {string} params.responseId id of response
+ */
+function assignIdsToTx({ tx, msg, responseId }) {
+  const tracker = tx.llm.responses
+  const trackedIds =
+    tracker.get(responseId) ??
+    new LlmTrackedIds({
+      requestId: msg.request_id,
+      conversationId: msg.conversation_id
+    })
+  trackedIds.message_ids.push(msg.id)
+  tracker.set(responseId, trackedIds)
+}
+
 /**
  * Registers the specification for instrumentation bedrock calls
  *
@@ -76,6 +99,7 @@ function getBedrockSpec({ config, commandName }, _shim, _original, _name, args)
       const bedrockResponse = new BedrockResponse({ bedrockCommand, response })
 
       if (modelType === 'completion') {
+        const tx = segment.transaction
         const summary = new LlmChatCompletionSummary({
           credentials,
           agent,
@@ -92,6 +116,7 @@ function getBedrockSpec({ config, commandName }, _shim, _original, _name, args)
           index: 0,
           completionId: summary.id
         })
+        assignIdsToTx({ tx, responseId: bedrockResponse.requestId, msg })
         recordEvent({ agent, type: 'LlmChatCompletionMessage', msg })
 
         bedrockResponse.completions.forEach((content, index) => {
@@ -105,6 +130,7 @@ function getBedrockSpec({ config, commandName }, _shim, _original, _name, args)
             content,
             completionId: summary.id
           })
+          assignIdsToTx({ tx, responseId: bedrockResponse.requestId, msg: chatCompletionMessage })
           recordEvent({ agent, type: 'LlmChatCompletionMessage', msg: chatCompletionMessage })
         })
 
diff --git a/tests/unit/llm/bedrock-response.tap.js b/tests/unit/llm/bedrock-response.tap.js
index 973087f07e..f08e91c743 100644
--- a/tests/unit/llm/bedrock-response.tap.js
+++ b/tests/unit/llm/bedrock-response.tap.js
@@ -57,13 +57,13 @@ tap.beforeEach((t) => {
     response: {
       statusCode: 200,
       headers: {
+        'x-amzn-requestid': 'aws-request-1',
         'x-foo': 'foo',
         ['x-amzn-bedrock-input-token-count']: 25,
         ['x-amzn-bedrock-output-token-count']: 25
       }
     },
     output: {
-      requestId: 'aws-request-1',
       body: new TextEncoder().encode('{"foo":"foo"}')
     }
   }
@@ -100,7 +100,7 @@ tap.test('non-conforming response is handled gracefully', async (t) => {
   t.equal(res.id, undefined)
   t.equal(res.inputTokenCount, 0)
   t.equal(res.outputTokenCount, 0)
-  t.equal(res.requestId, 'aws-request-1')
+  t.equal(res.requestId, undefined)
   t.equal(res.statusCode, 200)
 })
 
diff --git a/tests/unit/llm/event.tap.js b/tests/unit/llm/event.tap.js
index 7ac4f32562..c4a19d855e 100644
--- a/tests/unit/llm/event.tap.js
+++ b/tests/unit/llm/event.tap.js
@@ -46,9 +46,7 @@ tap.beforeEach((t) => {
   }
 
   t.context.bedrockResponse = {
-    headers: {
-      'x-amzn-requestid': 'request-1'
-    }
+    requestId: 'request-1'
   }
 
   t.context.bedrockCommand = {
diff --git a/tests/versioned/common.js b/tests/versioned/common.js
index 3c3295662f..9a80bbe1fe 100644
--- a/tests/versioned/common.js
+++ b/tests/versioned/common.js
@@ -117,6 +117,7 @@ function assertChatCompletionSummary({ tx, modelId, chatSummary, tokenUsage, err
     'id': /[\w]{8}-[\w]{4}-[\w]{4}-[\w]{4}-[\w]{12}/,
     'appName': 'New Relic for Node.js tests',
     'request_id': 'eda0760a-c3f0-4fc1-9a1e-75559d642866',
+    'conversation_id': 'convo-id',
     'trace_id': tx.traceId,
     'span_id': tx.trace.root.children[0].id,
     'transaction_id': tx.id,
@@ -128,6 +129,8 @@ function assertChatCompletionSummary({ tx, modelId, chatSummary, tokenUsage, err
     'api_key_last_four_digits': 'E ID',
     'response.number_of_messages': 2,
     'response.choices.finish_reason': 'endoftext',
+    'request.temperature': 0.5,
+    'request.max_tokens': 100,
     'error': error
   }
 
diff --git a/tests/versioned/v3/bedrock-chat-completions.tap.js b/tests/versioned/v3/bedrock-chat-completions.tap.js
index 3ad453fbe1..3f39e2cc89 100644
--- a/tests/versioned/v3/bedrock-chat-completions.tap.js
+++ b/tests/versioned/v3/bedrock-chat-completions.tap.js
@@ -13,23 +13,26 @@ const createAiResponseServer = require('../aws-server-stubs/ai-server')
 const { FAKE_CREDENTIALS } = require('../aws-server-stubs')
 const requests = {
   ai21: (prompt, modelId) => ({
-    body: JSON.stringify({ prompt }),
+    body: JSON.stringify({ prompt, temperature: 0.5, maxTokens: 100 }),
     modelId
   }),
   amazon: (prompt, modelId) => ({
-    body: JSON.stringify({ inputText: prompt }),
+    body: JSON.stringify({
+      inputText: prompt,
+      textGenerationConfig: { temperature: 0.5, maxTokenCount: 100 }
+    }),
     modelId
   }),
   claude: (prompt, modelId) => ({
-    body: JSON.stringify({ prompt }),
+    body: JSON.stringify({ prompt, temperature: 0.5, max_tokens_to_sample: 100 }),
     modelId
   }),
   cohere: (prompt, modelId) => ({
-    body: JSON.stringify({ prompt }),
+    body: JSON.stringify({ prompt, temperature: 0.5, max_tokens: 100 }),
     modelId
   }),
   llama2: (prompt, modelId) => ({
-    body: JSON.stringify({ prompt }),
+    body: JSON.stringify({ prompt, max_gen_length: 100, temperature: 0.5 }),
     modelId
   })
 }
@@ -116,7 +119,9 @@ tap.afterEach(async (t) => {
       const command = new bedrock.InvokeModelCommand(input)
 
       const { agent } = helper
+      const api = helper.getAgentApi()
       helper.runInTransaction(async (tx) => {
+        api.addCustomAttribute('llm.conversation_id', 'convo-id')
         await client.send(command)
         const events = agent.customEventAggregator.events.toArray()
         t.equal(events.length, 3)
@@ -153,4 +158,54 @@ tap.afterEach(async (t) => {
       t.equal(error.message, expected.body.message)
     }
   })
+
+  tap.test('should store ids and record feedback message accordingly', (t) => {
+    const { bedrock, client, helper } = t.context
+    const conversationId = 'convo-id'
+    const prompt = `text ${resKey} ultimate question`
+    const input = requests[resKey](prompt, modelId)
+    const command = new bedrock.InvokeModelCommand(input)
+
+    const { agent } = helper
+    const api = helper.getAgentApi()
+    helper.runInTransaction(async (tx) => {
+      api.addCustomAttribute('llm.conversation_id', conversationId)
+      const response = await client.send(command)
+      const responseId = response.$metadata.requestId
+      const events = agent.customEventAggregator.events.toArray()
+      const chatMsgs = events.filter(([{ type }]) => type === 'LlmChatCompletionMessage')
+      const ids = api.getLlmMessageIds({ responseId })
+      const messageIds = chatMsgs.map((msg) => msg[1].id)
+      t.equal(ids.request_id, responseId)
+      t.equal(ids.conversation_id, conversationId)
+      // message_ids order varies over test run, sort them to assure consistency
+      t.same(ids.message_ids.sort(), messageIds.sort())
+      api.recordLlmFeedbackEvent({
+        conversationId: ids.conversation_id,
+        requestId: ids.request_id,
+        messageId: ids.message_ids[0],
+        category: 'test-event',
+        rating: '5 star',
+        message: 'You are a mathematician.',
+        metadata: { foo: 'foo' }
+      })
+      const recordedEvents = agent.customEventAggregator.getEvents()
+      const [[, feedback]] = recordedEvents.filter(([{ type }]) => type === 'LlmFeedbackMessage')
+
+      t.match(feedback, {
+        id: /[\w\d]{32}/,
+        conversation_id: ids.conversation_id,
+        request_id: ids.request_id,
+        message_id: ids.message_ids[0],
+        category: 'test-event',
+        rating: '5 star',
+        message: 'You are a mathematician.',
+        ingest_source: 'Node',
+        foo: 'foo'
+      })
+
+      tx.end()
+      t.end()
+    })
+  })
 })