Agents

Normalize raw sustainability source fragments into Litmus claim families, source types, evidence requirements, and missing context fields.

normalize_source_fragments

Search the Litmus sustainability canon two ways at once: exact alias matching for known phrases, and semantic similarity over the knowledge graph for natural-language questions. Semantic matches arrive with their graph context — the certificates that back a claim and the programmes it counts toward.

search_litmus_canon

Read a material's sustainability claims against a project context — a rating-system target, a use (e.g. healthcare, education), and a region. Returns the curated Litmus briefing for the closest matching context: which evidence is required by law, required for the target, merely strengthens the case, or does not apply. This explains what a context REQUIRES; it never verifies a product or decides whether a product qualifies. If no curated context matches, it returns the available contexts instead.

assess_in_context

Classify free-text sustainability language from a product page or marketing copy into Litmus claim families. Returns each detected claim with the evidence it needs, its greenwashing risk, plus how much of the text was recognised and what went unmatched. Use when the user pastes product wording rather than asking a single question.

classify_material_claims