Semantic Ports (Semports) are an umbrella term we use for one-time or ongoing automated ports, often between languages.
Learning from the Best
We have great examples of LLM-native designs in the wild: the Vercel AI SDK, the OpenAI Agents SDK, and others. These examples have already thought through the best loop shapes, the best tool-calling abstractions, and so on. We want to benefit from that work, automatically.
Dependency Through Translation
Semports deliver changes from an upstream we trust into an internal library we can use. It's a form of dependency, like a traditional library, but the dependency can also be rewritten and adapted to our specific needs, automatically, while our human team members sleep.
One-Time vs. Ongoing
One-Time Ports
Migrate a library from one language to another, then own the result
Ongoing Ports
Continuously sync changes from upstream, merging new features automatically
Adaptive Ports
Reshape APIs to match internal conventions while preserving semantics
Why Not Just Use the Original?
Sometimes the original library is in the wrong language. Sometimes it has dependencies we can't accept. Sometimes we need to integrate deeply with internal systems. Semports let us benefit from upstream thinking without being constrained by upstream choices.