RETRIEVAL-AUGMENTED GENERATION (RAG)

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS [START OF PROMPT] Review this retrieval step's results for relevance quality. Given this query and these retrieved chunks, tell me which chunks should and should not have been retrieved, and why: query = [PASTE QUERY, e.g. "resumes matching senior backend engineer posting"], retrieved chunks = [PASTE LAB RETRIEVED JOB POSTING / RESUME CHUNKS HERE]. Output format: a relevance table with the columns Chunk, Should Retrieve (Y/N), and Reason, plus a suggested query or indexing fix if recall looks poor. [END OF PROMPT] [START OF PROMPT] Recommend a chunking strategy (chunk size and overlap) for this document set, given its structure: document description = [PASTE LAB JOB POSTINGS / RESUME DOCUMENT STRUCTURE HERE, e.g. headings, bullet lists, tables], retrieval goal = [GOAL]. Walk through the trade-off of too-small versus too-large chunks for this specific content and give concrete size and overlap numbers. Output format: a recommendation with specific numbers and a one-paragraph justification tied to the document structure provided. [END OF PROMPT]