RAG Search Augmentation and Internal Chatbot

Design an internal assistant or search augmentation layer over employee logs, Confluence, Jira, GDrive, or product data. The system detects fact-seeking/task-seeking queries, retrieves relevant documents, reasons over them, and returns an answer alongside normal search results. Requirements Detect whether a user query is fact-seeking or task-seeking. Extract entities and intent, such as manager, time range, task name, product, or document source. Retrieve relevant data from internal services, document stores, org graph, or activity logs. Use a retrieval layer such as keyword search plus ANN/vector search. Generate or extract an answer and present it above or alongside traditional results. Track quality with retrieval metrics and end-to-end answer metrics, including precision/recall, MRR, latency, faithfulness, answer quality, and user satisfaction. Notes A successful design can include orchestration, retrieval, optional memory, reasoning/planning, and response generation layers. For internal employee-log navigation, examples include resolving my manager, finding a manager's activity during a time range, and retrieving task-specific events. For Jira/Confluence chatbot variants, focus on permission-aware retrieval and source grounding. Preparation Prepare a RAG architecture with ingestion, indexing, permission filtering, retrieval, ranking, answer generation, and evaluation. Practice explaining when to use exact keyword search, vector retrieval, and hybrid reranking.