Simple Typo-Tolerant Lookup Tool for LLMs

<p><strong>Typo-Tolerant Customer Lookup Tool</strong></p><p>This workflow implements a <strong>simple fuzzy lookup service</strong> that allows an LLM agent to search for customers by name, even when the query or data contains typos or spelling variations.</p><ol><li><p>The <strong>tool input (customer name)</strong> is received from the LLM agent and converted into a KNIME table row.</p></li><li><p>The <strong>reference customer data</strong> is loaded from Hub Space, providing the lookup base.</p></li><li><p>Using the <strong>Approximate String Matcher (Levenshtein distance)</strong>, the workflow compares the input name with the reference data and calculates similarity scores.</p></li><li><p>An <strong>Expression Row Filter</strong> ensures that only the best match(es) are kept.</p></li><li><p>The resulting customer record is transformed into <strong>JSON format</strong> and returned to the agent as structured <strong>tool output</strong>.</p></li></ol><p>This setup enables <strong>typo-tolerant search</strong> for customer records, making it robust for use in LLM-powered assistants or data quality tasks where exact matches are unreliable.</p>