Lux AI

Retrieve the full text of one Lux Research passage by its `id`. `id` must be an id returned by `search`.

Search Lux Research's proprietary publications corpus. Returns the passages most relevant to a natural-language question. Each carries a stable `id` you can pass to `fetch` to retrieve its full text, a `title`, and a `url` to the publication's landing page (empty when it has no public landing page). Use this for questions about what Lux has published, recommended, or said about a topic, and for domain questions where Lux's publications are the authoritative grounding. Ground the answer in the returned passages and cite them by `url`.

Search Lux's proprietary publications corpus for passages relevant to a natural-language question. Use this tool when Lux-authored content would materially improve the answer, especially for questions about what the company has published, recommended, or said about a topic. This includes requests for Lux guidance, published viewpoints, best practices, source articles, or Lux-backed citations. You may also use this tool for broader domain questions when Lux's publications are likely to provide the most authoritative grounding. Do not use this tool for general-knowledge questions, casual brainstorming, or topics that do not need company-specific grounding. This is a retrieval-only tool. It returns an object with two fields: `results` (the source passages, each carrying a pre-formatted `citation` plus `name` and `landing_page_url`) and `excluded_sources` (source domains to never cite, link, or mention in any broadened / web-sourced answer). Ground the final answer in `results`; this tool does not generate the final user-facing answer. When using the results: - Select only passages that are relevant to the user's question. - Prefer substantive publications over event blurbs, webinar pages, or other lightweight promotional descriptions when stronger sources are available. - Synthesize a direct answer in your own words rather than listing results. - Do not present inferred frameworks or conclusions as if they were stated explicitly in the retrieved publications. - Use returned publication to help ground claims. - Prefer paraphrasing over quotation. - Qualify uncertainty when the evidence is weak, partial, conflicting, or based mainly on summary-style source material. - Do not narrate tool use unless it meaningfully helps the user. Citation requirements: - Cite every claim derived from a result, using that result's `name` and `landing_page_url` (each result also carries a pre-formatted `citation` field you may copy as-is when no other format is specified). - If the host agent's instructions define a citation format, follow them. - Do not cite sources that were not returned by this tool.