AI-News-Agent

AI News Agent is a production-grade agent that ingests AI and machine learning news from multiple sources, deduplicates and summarizes it with an LLM, and serves it through an API and a dashboard you can query in natural language. The ingestion pipeline and the agent are separate systems: the pipeline runs on a schedule and builds a curated corpus, while the agent decides at query time which tools to use and whether it has enough evidence to answer. The project automates tracking AI/ML developments by pulling from RSS feeds, APIs, and scraped pages on a schedule, detecting when several articles describe the same story, summarizing each story once, and letting you ask questions about the result. Design priorities include cost control, failure isolation, and observability. Embeddings run locally and cost nothing, LLM calls happen once per deduplicated story, and a daily budget guard halts enrichment before it overspends. A broken feed marks itself degraded and the run continues, and every run records what it attempted, what succeeded, what it cost, and what failed.