Expand description
patent — a prior-art search for your code ideas.
Takes a plain-English dev-tool idea and searches the open-source ecosystem — crates.io, npm, PyPI, GitHub, Go, Maven, NuGet, RubyGems, Docker Hub, the VS Code Marketplace, and Hacker News — for prior art, then gives an honest, scoped verdict on whether it’s already been built. The exact set searched is chosen per query; whichever sources actually responded are always surfaced.
Integrity principle: this tool can prove something exists, but never that it doesn’t — it only searched some sources. All output is scoped to “what was found in the sources checked.”
§Install
cargo install patent§Usage
patent "interactive cli to kill whatever's on a port" # interactive TUI
patent "react component for infinite scroll" --json # structured output
patent "kubernetes log viewer" --fast # skip the LLM verdict
patent "vector database" --api-base https://api.openai.com/v1 --model gpt-4o-mini§Using the library
patent is primarily the engine behind the CLI of the same name, but the
core is reusable: sources::search_all fans out to the registries,
rank orders matches by semantic similarity, and verdict::assess
turns them into an integrity-scoped Verdict via any Llm backend
(local Ollama or an OpenAI-compatible API).
Re-exports§
pub use llm::Llm;pub use model::Match;pub use model::Query;pub use model::Saturation;pub use model::Source;pub use model::Verdict;
Modules§
- llm
- Text-generation backend used to write the verdict.
- model
- Core domain types shared across the pipeline.
- ollama
- Minimal Ollama client (
localhost:11434/api/generate). - openai
- OpenAI-compatible chat-completions backend.
- rank
- Semantic ranking.
- sources
- Source registry: one implementor per ecosystem, fanned out concurrently.
- tui
- TUI state machine.
- verdict
- Verdict generation.
Enums§
- Error
- Library-level error type. The binary maps these to
anyhowwith context.
Type Aliases§
- Result
- Crate result alias.