Autonomous Knowledge Synthesis for Agentic Workflows

Act as an expert Systems Architect and Knowledge Engineer specializing in Autonomous Information Retrieval (AIR) systems. Your task is to design a robust, automated framework for ingestion and synthesis of technical research from diverse sources like ArXiv, research blogs, and repository ecosystems. Step 1: Define a data-adaptive ingestion strategy that utilizes rate-limited scrapers and semantic deduplication to normalize incoming technical content. Step 2: Implement a cognitive evaluation layer using a chain-of-thought LLM agent to score content based on architectural utility, RAG compatibility, and technical complexity (1-10 scale). Step 3: Outline a vector-based graph storage schema that links entities, libraries, and design patterns, ensuring that the knowledge base remains queryable for downstream agentic tools. Step 4: Create a maintenance loop that includes a decay function to automatically flag or prune 'stale' nodes that have been superseded by newer research or updated frameworks. Constraints: Focus strictly on production-grade intelligence nodes rather than broad listicles. Ensure the output is language-agnostic and structured for integration into Model Context Protocol (MCP) servers. The final output must include a Mermaid flowchart representing the data lifecycle from raw source discovery to synthesized knowledge node publication. Maintain a focus on low-noise signal extraction, ensuring the system filters out marketing fluff while retaining core logic and code-level architectural patterns. Your response should look like a comprehensive technical blueprint with clear definitions for each pipeline component, including specific criteria for the LLM evaluation prompts used to score content quality and relevance for AI engineers.