Knowledge Base Workflow
It is designed to enhance MCPs and AI Agents by enabling document-level retrieval and smart memory capabilities. This section provides a high-level overview of the two primary workflows involved in setting up and using the knowledge base effectively.
Document Indexing
This workflow focuses on building structured, searchable knowledge from uploaded documents, making them available for intelligent retrieval by agents and MCPs.
Steps:
-
Upload Documents: Import knowledge materials in supported formats (e.g., PDF, DOCS, Markdown) into designated knowledge folders.
-
Configure Extraction: Set up how content is extracted from each document. Define metadata, parsing strategies, and whether to split by sections or pages.
-
Build Search Index: Generate a semantic search index from the extracted content, enabling fast and accurate retrieval by AI systems.
-
Test Search Capabilities: Validate the search index by querying sample questions and reviewing the quality and relevance of returned results.
Knowledge Integration
Once documents are indexed, the next step is to bind that knowledge to the intelligent systems that need it. This workflow defines how knowledge is retrieved, interpreted, and cached by agents.
Steps:
-
Select Knowledge Source: Choose the specific knowledge folders or indices you want to connect to an agent or MCP instance.
-
Configure Retrieval Settings: Set the parameters for how information should be retrieved, including maximum number of documents, score thresholds, and filters.
-
Define Query Handling: Determine how the agent processes user questions. Specify fallback behavior, answer formatting, and use of citation or summarization.
-
Set Up Caching Strategy: Optimize performance and reduce latency by configuring memory caching for frequently asked queries or static content.