Generate Structured Metadata for English and Chinese

• Multi-Source Data Integration: Connects to Google Drive and potentially external scraping services (Apify).
• Automated Data Cleaning & Formatting: Parses complex JSON fields, standardizes opening hours, and cleans phone numbers.
• AI-Powered Enrichment: Utilizes OpenAI models for generating business summaries and lead scoring.
• Lead Scoring & Categorization: Assigns a lead score and quality (High, Medium, Low) based on various restaurant attributes.
• Structured Metadata Generation: Creates a clean, organized dataset suitable for analysis and further processing.
• Vector Database Integration: Embeds and stores data in Supabase for AI-powered retrieval.
• Bilingual Output Capability: While not explicitly shown in transformation steps, the AI agent's system message implies a need for comprehensive data that can be adapted for bilingual use. The overall aim is to generate structured data that can then be translated or used to generate bilingual content.
• Conversational AI Support: Integrates chat memory for maintaining context in AI interactions.

1. Google Drive Setup: Ensure your Google Drive is configured with the ReCharge folder (1dh1Rr2yrhSdoSYpiR8s1yXiSWLYxpoLJ) and the Khaisa Studio credentials are set up for Google Drive OAuth2.
2. Apify Integration (if used): Configure the Scrape Maps node with your Apify API URL and ensure the language parameter is set appropriately for your target region (currently id for Indonesian). The locationQuery should be dynamically provided.
3. Google Sheets Setup: Configure the Update Data and Append Leads nodes with your Google Sheets credentials (Sheet & Drive) and the correct Spreadsheet ID (1DKXYqpb0EWNYsbq2x8h5xqUWXy772toNDVrl8d-8nfA) and Sheet name (Restaurant).
4. OpenAI Credentials: Set up your OpenAI API credentials (n8n - Money manager Khairul) for embedding and chat model usage.
5. Supabase Credentials: Configure your Supabase credentials (REcharge) for the vector store, ensuring the documents table exists.
6. PostgreSQL Credentials: Set up your PostgreSQL credentials (ReCharge Database) for chat memory.
7. Workflow Trigger: The Google Drive Trigger is set to fileUpdated on the ReCharge folder. Alternatively, the When clicking ‘Execute workflow’ node can be used for manual execution.
8. Data Flow: When a file is updated in Google Drive, it will be searched (Search File), downloaded (Get Data), and then processed. If Apify is used, its output will be cleaned (Clean Data) and appended to Google Sheets (Append Leads). The cleaned data from Google Drive will be processed and then inserted into Supabase (Supabase Vector Store).
9. AI Agent & Chat: The AI Agent node uses an OpenAI chat model (OpenAI Chat Model) with PostgreSQL chat memory (Postgres Chat Memory) for conversational interactions. Embeddings are generated using Embeddings OpenAI and Embeddings OpenAI1.

To generate structured metadata in both English and Chinese:

• English Metadata: The Clean Data node and the AI agent generate structured data primarily in English. This is evident from the variable names and the business_summary which is built using English descriptions.
• Chinese Metadata: To achieve Chinese metadata:
  • Modify AI System Prompt: Update the AI Agent node's systemMessage to explicitly instruct the AI to provide outputs in both English and Chinese, or to generate content specifically for Chinese users.
  • Translation Steps: Introduce translation nodes (e.g., using an AI translation service) after the initial data cleaning or AI generation steps to translate relevant fields (like business_summary, description, etc.) into Chinese.
  • Data Structuring for Bilingual: Ensure that when structuring the final metadata, fields are created for both English and Chinese versions of key information (e.g., title_en, title_zh, description_en, description_zh). The current JSON structure in the Clean Data node needs to be extended to accommodate this.
  • Output Destination: Consider how the bilingual metadata will be stored or used. This might involve adding new columns to the Google Sheet or a separate output mechanism for the Chinese version.