diff --git a/README.md b/README.md
index b412eea5..8d12019a 100644
--- a/README.md
+++ b/README.md
@@ -84,6 +84,8 @@ Example:
 
 ![PR Conversation](./docs/images/openai-review-conversation.png)
 
+![PR Conversation Request](./docs/images/openai-review-request.png)
+
 #### Environment variables
 
 - `GITHUB_TOKEN`: This should already be available to the GitHub Action
diff --git a/action.yml b/action.yml
index eef76a8c..3f42f5ee 100644
--- a/action.yml
+++ b/action.yml
@@ -70,13 +70,13 @@ inputs:
     required: false
     description: 'System message to be sent to OpenAI'
     default: |
-      You are a highly experienced software engineer with a strong ability to 
-      review code changes thoroughly.You can uncover issues such as logic 
-      errors, syntax errors, out of bound errors, potential data races, 
+      You are `@openai`, a highly experienced software engineer with a strong 
+      ability to review code changes thoroughly.You can uncover issues such as 
+      logic errors, syntax errors, out of bound errors, potential data races, 
       livelocks, starvation, suspension, order violation, atomicity violation, 
       consistency, complexity, error handling, and more.
 
-      We will be conducting code reviews today.
+      You will be conducting code reviews today and write code if asked to do so.
   summarize_beginning:
     required: false
     description: 'The prompt for the whole pull request'
@@ -156,7 +156,7 @@ inputs:
       $summary
       ```
 
-      Reply "OK" to confirm that you are ready for further instructions.
+      Reply "OK" to confirm.
   review_file:
     required: false
     description: 'The prompt for each file'
@@ -167,6 +167,8 @@ inputs:
       ```
       $file_content
       ```
+
+      Reply "OK" to confirm.
   review_file_diff:
     required: false
     description: 'The prompt for each file diff'
@@ -177,6 +179,8 @@ inputs:
       ```diff
       $file_diff
       ```
+
+      Reply "OK" to confirm.
   review_patch_begin:
     required: false
     description: 'The prompt for each file diff'
@@ -187,7 +191,9 @@ inputs:
       "LGTM!" with a short comment. 
 
       Your responses will be recorded as review comments on the pull request. 
-      Markdown format is preferred for your responses. Reply "OK" to confirm.
+      Markdown format is preferred for your responses. 
+
+      Reply "OK" to confirm.
   review_patch:
     required: false
     description: 'The prompt for each chunks/patches'
@@ -211,11 +217,13 @@ inputs:
       $system_message
 
       A comment was made on a review for a diff patch on file 
-      `$filename`. You will be replying directly to that. If possible, I 
-      will provide you the file and the entire diff to help 
-      provide overall context.
+      `$filename`. I would like you to follow the instructions 
+      in that comment. 
 
-      For context, the pull request has the title "$title" and the following 
+      If possible, I will provide you the file and the entire diff 
+      to help provide overall context for your response.
+
+      The pull request has the title "$title" and the following 
       description:
 
       ```
@@ -238,6 +246,8 @@ inputs:
       ```
       $file_content
       ```
+
+      Reply "OK" to confirm.
   comment_file_diff:
     required: false
     description: 'Prompt for file diff'
@@ -247,12 +257,14 @@ inputs:
       ```diff
       $file_diff
       ```
+
+      Reply "OK" to confirm.
   comment:
     required: false
     description: 'Prompt for comment'
     default: |
-      I would like you to reply to the new comment made on 
-      a conversation chain on a code review diff.
+      I would like you to follow the instructions in the new 
+      comment made on a conversation chain on a code review diff.
 
       Diff being commented on:
 
@@ -269,19 +281,21 @@ inputs:
       $comment_chain
       ```
 
-      Please reply directly to the new comment in the conversation 
-      chain (instead of suggesting a reply) and your reply will be 
-      posted as-is.
+      Please reply directly to the new comment (instead of suggesting 
+      a reply) and your reply will be posted as-is.
+
+      If the comment contains instructions/request for you, please comply. 
+      For example, if the comment is asking you to generate documentation 
+      comments on the code, in your reply please generate the required code.
 
-      In your reply, please make sure to begin the reply by
-      tagging the user with "@user".
+      In your reply, please make sure to begin the reply by tagging the user 
+      with "@user".
 
-      The comment that you need to directly reply to:
+      The comment/request that you need to directly reply to:
 
       ```
       $comment
       ```
-
 runs:
   using: 'node16'
   main: 'dist/index.js'
diff --git a/dist/index.js b/dist/index.js
index aa1c400e..4a5d950f 100644
--- a/dist/index.js
+++ b/dist/index.js
@@ -29273,6 +29273,7 @@ const handleReviewComment = async (bot, prompts) => {
         if (comment_chain.includes(_commenter_js__WEBPACK_IMPORTED_MODULE_2__/* .COMMENT_TAG */ .Rs) ||
             comment_chain.includes(_commenter_js__WEBPACK_IMPORTED_MODULE_2__/* .COMMENT_REPLY_TAG */ .aD) ||
             comment.body.startsWith(ASK_BOT)) {
+            let file_content = '';
             try {
                 const contents = await octokit.repos.getContent({
                     owner: repo.owner,
@@ -29283,7 +29284,7 @@ const handleReviewComment = async (bot, prompts) => {
                 if (contents.data) {
                     if (!Array.isArray(contents.data)) {
                         if (contents.data.type === 'file' && contents.data.content) {
-                            inputs.file_content = Buffer.from(contents.data.content, 'base64').toString();
+                            file_content = Buffer.from(contents.data.content, 'base64').toString();
                         }
                     }
                 }
@@ -29291,6 +29292,7 @@ const handleReviewComment = async (bot, prompts) => {
             catch (error) {
                 _actions_core__WEBPACK_IMPORTED_MODULE_0__.warning(`Failed to get file contents: ${error}, skipping.`);
             }
+            let file_diff = '';
             try {
                 // get diff for this file by comparing the base and head commits
                 const diffAll = await octokit.repos.compareCommits({
@@ -29304,7 +29306,7 @@ const handleReviewComment = async (bot, prompts) => {
                     if (files) {
                         const file = files.find(f => f.filename === comment.path);
                         if (file && file.patch) {
-                            inputs.file_diff = file.patch;
+                            file_diff = file.patch;
                         }
                     }
                 }
@@ -29320,8 +29322,9 @@ const handleReviewComment = async (bot, prompts) => {
             // begin comment generation
             const [, comment_begin_ids] = await bot.chat(prompts.render_comment_beginning(inputs), {});
             let next_comment_ids = comment_begin_ids;
-            if (inputs.file_content.length > 0) {
-                const file_content_tokens = _tokenizer_js__WEBPACK_IMPORTED_MODULE_4__/* .get_token_count */ .u(inputs.file_content);
+            if (file_content.length > 0) {
+                inputs.file_content = file_content;
+                const file_content_tokens = _tokenizer_js__WEBPACK_IMPORTED_MODULE_4__/* .get_token_count */ .u(file_content);
                 if (file_content_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
                     const [file_content_resp, file_content_ids] = await bot.chat(prompts.render_comment_file(inputs), next_comment_ids);
                     if (file_content_resp) {
@@ -29329,8 +29332,13 @@ const handleReviewComment = async (bot, prompts) => {
                     }
                 }
             }
-            if (inputs.file_diff.length > 0) {
-                const file_diff_tokens = _tokenizer_js__WEBPACK_IMPORTED_MODULE_4__/* .get_token_count */ .u(inputs.file_diff);
+            if (file_diff.length > 0) {
+                inputs.file_diff = file_diff;
+                // use file diff if no diff was found in the comment
+                if (inputs.diff.length === 0) {
+                    inputs.diff = file_diff;
+                }
+                const file_diff_tokens = _tokenizer_js__WEBPACK_IMPORTED_MODULE_4__/* .get_token_count */ .u(file_diff);
                 if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
                     const [file_diff_resp, file_diff_ids] = await bot.chat(prompts.render_comment_file_diff(inputs), next_comment_ids);
                     if (file_diff_resp) {
@@ -29367,6 +29375,18 @@ ${_commenter_js__WEBPACK_IMPORTED_MODULE_2__/* .COMMENT_REPLY_TAG */ .aD}
                 }
                 catch (error) {
                     _actions_core__WEBPACK_IMPORTED_MODULE_0__.warning(`Failed to reply to the top-level comment`);
+                    try {
+                        await octokit.pulls.createReplyForReviewComment({
+                            owner: repo.owner,
+                            repo: repo.repo,
+                            pull_number,
+                            body: `Could not post the reply to the top-level comment due to the following error: ${error}`,
+                            comment_id: topLevelCommentId
+                        });
+                    }
+                    catch (error) {
+                        _actions_core__WEBPACK_IMPORTED_MODULE_0__.warning(`Failed to reply to the top-level comment`);
+                    }
                 }
             }
             else {
@@ -29657,9 +29677,11 @@ const codeReview = async (bot, options, prompts) => {
         const generateSummary = async (filename, file_content, file_diff) => {
             const ins = inputs.clone();
             ins.filename = filename;
-            ins.file_content = file_content;
-            ins.file_diff = file_diff;
+            if (file_content.length > 0) {
+                ins.file_content = file_content;
+            }
             if (file_diff.length > 0) {
+                ins.file_diff = file_diff;
                 const file_diff_tokens = tokenizer/* get_token_count */.u(file_diff);
                 if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
                     // summarize diff
@@ -29705,8 +29727,8 @@ ${filename}: ${summary}
 ---
 
 Tips: 
-- You can reply on the review comment left by this bot to ask follow-up questions.
-- You can invite the bot into a review conversation by typing \`@openai\` in the beginning of the comment.
+- Reply on the review comment left by this bot to ask follow-up questions.
+- Invite the bot into a review conversation by typing \`@openai\` in the beginning of the comment.
 `;
             next_summarize_ids = summarize_final_response_ids;
             await commenter.comment(`${summarize_comment}`, lib_commenter/* SUMMARIZE_TAG */.Rp, 'replace');
@@ -29731,9 +29753,8 @@ Tips:
             // make a copy of inputs
             const ins = inputs.clone();
             ins.filename = filename;
-            ins.file_content = file_content;
-            ins.file_diff = file_diff;
             if (file_content.length > 0) {
+                ins.file_content = file_content;
                 const file_content_tokens = tokenizer/* get_token_count */.u(file_content);
                 if (file_content_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
                     try {
@@ -29755,6 +29776,7 @@ Tips:
                 }
             }
             if (file_diff.length > 0) {
+                ins.file_diff = file_diff;
                 const file_diff_tokens = tokenizer/* get_token_count */.u(file_diff);
                 if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
                     try {
diff --git a/docs/images/openai-review-request.png b/docs/images/openai-review-request.png
new file mode 100644
index 00000000..9847f017
Binary files /dev/null and b/docs/images/openai-review-request.png differ
diff --git a/src/review-comment.ts b/src/review-comment.ts
index 0a79315c..86753227 100644
--- a/src/review-comment.ts
+++ b/src/review-comment.ts
@@ -79,6 +79,7 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
       comment_chain.includes(COMMENT_REPLY_TAG) ||
       comment.body.startsWith(ASK_BOT)
     ) {
+      let file_content = ''
       try {
         const contents = await octokit.repos.getContent({
           owner: repo.owner,
@@ -89,7 +90,7 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
         if (contents.data) {
           if (!Array.isArray(contents.data)) {
             if (contents.data.type === 'file' && contents.data.content) {
-              inputs.file_content = Buffer.from(
+              file_content = Buffer.from(
                 contents.data.content,
                 'base64'
               ).toString()
@@ -99,6 +100,8 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
       } catch (error) {
         core.warning(`Failed to get file contents: ${error}, skipping.`)
       }
+
+      let file_diff = ''
       try {
         // get diff for this file by comparing the base and head commits
         const diffAll = await octokit.repos.compareCommits({
@@ -112,7 +115,7 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
           if (files) {
             const file = files.find(f => f.filename === comment.path)
             if (file && file.patch) {
-              inputs.file_diff = file.patch
+              file_diff = file.patch
             }
           }
         }
@@ -135,10 +138,9 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
         {}
       )
       let next_comment_ids = comment_begin_ids
-      if (inputs.file_content.length > 0) {
-        const file_content_tokens = tokenizer.get_token_count(
-          inputs.file_content
-        )
+      if (file_content.length > 0) {
+        inputs.file_content = file_content
+        const file_content_tokens = tokenizer.get_token_count(file_content)
         if (file_content_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
           const [file_content_resp, file_content_ids] = await bot.chat(
             prompts.render_comment_file(inputs),
@@ -150,8 +152,13 @@ export const handleReviewComment = async (bot: Bot, prompts: Prompts) => {
         }
       }
 
-      if (inputs.file_diff.length > 0) {
-        const file_diff_tokens = tokenizer.get_token_count(inputs.file_diff)
+      if (file_diff.length > 0) {
+        inputs.file_diff = file_diff
+        // use file diff if no diff was found in the comment
+        if (inputs.diff.length === 0) {
+          inputs.diff = file_diff
+        }
+        const file_diff_tokens = tokenizer.get_token_count(file_diff)
         if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
           const [file_diff_resp, file_diff_ids] = await bot.chat(
             prompts.render_comment_file_diff(inputs),
@@ -198,6 +205,17 @@ ${COMMENT_REPLY_TAG}
           })
         } catch (error) {
           core.warning(`Failed to reply to the top-level comment`)
+          try {
+            await octokit.pulls.createReplyForReviewComment({
+              owner: repo.owner,
+              repo: repo.repo,
+              pull_number,
+              body: `Could not post the reply to the top-level comment due to the following error: ${error}`,
+              comment_id: topLevelCommentId
+            })
+          } catch (error) {
+            core.warning(`Failed to reply to the top-level comment`)
+          }
         }
       } else {
         core.warning(`Failed to find the top-level comment to reply to`)
diff --git a/src/review.ts b/src/review.ts
index 3fbfc328..50396c7c 100644
--- a/src/review.ts
+++ b/src/review.ts
@@ -164,9 +164,12 @@ export const codeReview = async (
     ): Promise<[string, string] | null> => {
       const ins = inputs.clone()
       ins.filename = filename
-      ins.file_content = file_content
-      ins.file_diff = file_diff
+
+      if (file_content.length > 0) {
+        ins.file_content = file_content
+      }
       if (file_diff.length > 0) {
+        ins.file_diff = file_diff
         const file_diff_tokens = tokenizer.get_token_count(file_diff)
         if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
           // summarize diff
@@ -225,8 +228,8 @@ ${filename}: ${summary}
 ---
 
 Tips: 
-- You can reply on the review comment left by this bot to ask follow-up questions.
-- You can invite the bot into a review conversation by typing \`@openai\` in the beginning of the comment.
+- Reply on the review comment left by this bot to ask follow-up questions.
+- Invite the bot into a review conversation by typing \`@openai\` in the beginning of the comment.
 `
 
       next_summarize_ids = summarize_final_response_ids
@@ -263,10 +266,9 @@ Tips:
           const ins: Inputs = inputs.clone()
 
           ins.filename = filename
-          ins.file_content = file_content
-          ins.file_diff = file_diff
 
           if (file_content.length > 0) {
+            ins.file_content = file_content
             const file_content_tokens = tokenizer.get_token_count(file_content)
             if (file_content_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
               try {
@@ -291,6 +293,7 @@ Tips:
           }
 
           if (file_diff.length > 0) {
+            ins.file_diff = file_diff
             const file_diff_tokens = tokenizer.get_token_count(file_diff)
             if (file_diff_tokens < MAX_TOKENS_FOR_EXTRA_CONTENT) {
               try {