Step 2: Connect Google Sheets

These nodes read source video IDs and upsert the sentiment results back into your spreadsheet.

1. Open Retrieve Sheet Rows and choose your Document and Sheet from the picker.
2. Open Upsert Sheet Row and select the same Document and target Sheet for results.
3. In Upsert Sheet Row, keep Operation set to appendOrUpdate and Matching Columns set to comment_id.
4. Verify the column mappings use expressions like {{ $('Flatten Comment Data').item.json.comment_id }} and {{ $json.sentiment }} exactly as configured.
5. Credential Required: Connect your Google Sheets credentials (required for both Retrieve Sheet Rows and Upsert Sheet Row).

Step 3: Set the YouTube API Key and Batch Processing

The workflow injects the YouTube API key, then iterates across each row with a batch iterator.

1. Open Set YouTube Key and set Youtube API Key to your key value, replacing [CONFIGURE_YOUR_API_KEY].
2. Confirm Set YouTube Key connects to Batch Iterator.
3. In Batch Iterator, keep default options unless you need a specific batch size.
4. Note that Batch Iterator sends items to Idle Placeholder and then to YouTube API Call for processing.

Step 4: Fetch and Flatten YouTube Comments

Comments are fetched from the YouTube API and flattened into a row-ready format.

1. Open YouTube API Call and verify the URL is set to =https://www.googleapis.com/youtube/v3/commentThreads?part=snippet&videoId={{ $('Retrieve Sheet Rows').item.json.video_id }}&maxResults=100&key={{ $json['Youtube API Key'] }}.
2. Ensure YouTube API Call connects to Flatten Comment Data.
3. In Flatten Comment Data, keep the provided JavaScript which outputs fields like comment_id, comment_text, and like_count.

Step 5: Set Up AI Sentiment Analysis

This chain sends each comment to the LLM, enforces JSON output, and normalizes the response.

1. Open LLM Sentiment Chain and keep the Text prompt set to =Data to analyze: youtube_comment_text: "{{ $json.comment_text }}" like_count: {{ $json.like_count }} Consider like_count only as a weak signal (slightly increase |score| if like_count is high and sentiment is clear). Return the JSON only. .
2. Verify LLM Sentiment Chain uses the system message defining JSON-only output and key rules.
3. Open OpenRouter Chat Engine and set Model to google/gemini-2.0-flash-exp:free.
4. Open JSON Schema Parser and confirm the example schema matches your desired output structure.
5. Open Normalize LLM Output and keep the normalization script intact to enforce consistent sentiment, score, keywords, and language fields.
6. Credential Required: Connect your OpenRouter credentials in OpenRouter Chat Engine. The JSON Schema Parser is an AI sub-node—add credentials to OpenRouter Chat Engine, not the parser.

Step 6: Aggregate and Report Results

Sentiment data is aggregated and summarized into a Telegram report.

1. Open Aggregate Results and verify it aggregates sentiment, sentiment_score, and keywords.
2. In Summarize Metrics, keep the summarization script to produce totals, percentages, and top keywords.
3. Open Send Telegram Report and set Chat ID to your Telegram ID, replacing [YOUR_ID].
4. Keep the Text field exactly as configured to include expressions like {{ $json.total_comments }} and {{ $json.top_keywords.join(", ") }}.
5. Credential Required: Connect your Telegram credentials in Send Telegram Report.

Step 7: Test and Activate Your Workflow

Run the workflow once to confirm data flows from Sheets to YouTube, through the LLM, and back to Sheets and Telegram.

1. Click Execute Workflow to run a manual test.
2. Confirm YouTube API Call returns comment threads and Flatten Comment Data outputs individual comment items.
3. Verify Upsert Sheet Row adds or updates rows with comment_id, sentiment, sentiment_score, and keywords.
4. Check that Send Telegram Report posts a summary with totals and top keywords.
5. When satisfied, toggle the workflow to Active to enable scheduled production runs.