Three ways to engage¶
Explore. Everything described in this paper is reachable today without
an account. Search and browse the graph at
kb.terpedia.com. Ask the specialist agents and the
Tersonae a question at chat.terpedia.com. Look
up a molecule’s measured presence in a real product at
mondays
curl -s https://terpedia-knowledge-nanrsdlaoa-uc.a.run.app/v1/cyc/limonene | head -c 600Or run the live examples in this paper in your browser and change the keywords.
Integrate. Teams that want programmatic access request an API key for the tabular, semantic, and entity endpoints. The knowledge API is documented with request and response examples; the chat backend exposes OpenAI-compatible endpoints so that existing tools, notebooks, and agent frameworks can point at Terpedia with a configuration change; and an MCP tool server exposes knowledge-base operations to agent runtimes that speak the Model Context Protocol. Integration engagements typically begin with a scoped question (“normalize our incoming COAs”, “attach evidence to our product catalog”) and a two-week proof of value.
Partner. Deeper collaborations include co-branded product catalogs on the MONDAYS model, Terproduct as a certificate-of-analysis system of record for a brand or a laboratory, direct BigQuery and SPARQL access for analytics teams, custom Terports on a product class or question, and joint research on questions where a partner has data and Terpedia has method. Research collaborations are published under the same evidence rules as everything else.
What an engagement looks like¶
Scoping (week 0). Identify the decisions the data must support, the sources already in use, and the regulatory context.
Identity pass (weeks 1–2). Resolve the partner’s compound, ingredient, and product vocabularies to structural identities; report unresolved items rather than forcing matches.
Evidence attachment (weeks 2–4). Join the resolved identities to occurrence, property, regulatory, assay, and literature records with provenance and evidence type intact.
Surface (weeks 3–6). Deliver through whichever surface fits the team: API, BigQuery, agent, embedded pages, or Terproduct.
Review and publish. Optional Terport or joint publication; optional inclusion in the public graph where the partner’s data is open.
Roadmap¶
The roadmap below reflects the current public plans in Terpedia’s repositories. Dates are directional.
Knowledge base (2026 H2). Complete the PubChem RDF shard set; promote MeSH and MONDO into the semantic store; implement the release-aware UniProt adapter; complete LOTUS occurrence metadata promotion; extend the open resolver across the full census union so that every candidate identity has an entity page.
Product (2026 H2 – 2027 H1). Move terpedia.com to a feed-first experience in which Tersonae surface literature, news, and product updates for the terpenes a user follows; open Tersona chat to every entity in the graph; expand Terproduct’s field-scanning and certificate ingestion; publish the Rx (recipes) module for shareable, forkable formulations linked to the encyclopedia and the product catalog.
Research (rolling). Submit the first wave of manuscripts (Census, Claims, Terpotype and its audit, Absinthe, Artemisia); finalize the Classifier and TerpNet methods papers; advance NAuTILUS from design to a validated target-panel docking pipeline with redocking and experimental anchors as gates.
Platform. Continue to publish counting and coverage reports on every source refresh; add per-source data-quality dashboards to the public site; expose license filtering directly in the API.
Contact¶
Terpedia, LLC, Monument, Colorado. terpedia.com · dan@terpedia.com · github.com/Terpedia
This white paper is maintained as a living document at its published URL. The Markdown source, the executable examples, and the bibliography are open; the examples are re-executed against the live service on every build, so the numbers on the live-examples page are as current as the last publish date on the site.