agent-search

Why we included this project

Search agents usually need two things that are awkward to combine: live results from a real search engine, and an LLM that can make sense of them. AgentSearch wires those together in a small Python library. You connect a search-specialized model like Sensei-7B to Bing, SERP API, or the project's own AgentSearch engine, and the framework runs the retrieval-augmented loop for you, summarizing results, generating follow-up queries, and returning structured JSON. The quickstart covers both a pre-configured endpoint and a fully custom workflow where you assemble the search context and prompt yourself. The code is compact enough to read through in an afternoon, which makes it a useful reference for teams building their own RAG search pipeline rather than adopting a heavyweight platform.