VibeSolve
Built for developers and researchers to quickly prototype optimisation solvers.
Free forever
Open source.
- Interactive user validation
- Self-healing pipeline
- Instant containerisation
VibeSolve is an experimental AI tool that turns plain-English problem descriptions into optimisation projects, exploring where LLMs can add real value to the optimisation workflow. Built on people-first, developer-friendly principles.
Review the extracted problem specification in plain English, correct any misunderstandings in real time and ensure the system completely understands your requirements before any code is generated.
The AI pipeline utilises Docker to automatically collect and fix compile errors in the generated code.
Launch the generated solver instantly with a single command and interact with the full Timefold solver through REST endpoints.
Assign people to time segments while managing availability, skills, contracted hours, balanced workload, required rest between shifts, and more.
employee shifts school timetables sport tournaments
Plan efficient vehicle routes to minimise travel time and/or cost while respecting capacity constraints, time windows and driver limits.
delivery services waste collection fuel transportation
Match the right resources to the right task. This is often paired with other problem types, such as scheduling, in more complex cases. Hospital rostering is a common use case for this combination.
student dorms seat assignment airport gate allocation
Aside from the popular use cases mentioned here, VibeSolve is particularly well suited to combinatorial optimisation problems based on complex, real-world business rules.
Built for developers and researchers to quickly prototype optimisation solvers.
Free forever
Open source.
Need optimisation expertise, consulting and solutions ready for production?
Talk to us
Contact us for a customised offer tailored to your needs.
Dots & Lines is a deep-tech company specialised in mathematical optimisation and AI. Based in science, we model business problems, create digital twins and use sophisticated algorithms to increase profits, customer satisfaction and staff happiness.
We believe the best optimisation improves outcomes both for businesses and for the people who keep them running.
Need consulting or a custom solution? Tell us what you’re working on.
Optimisation is the process of searching for the best solution to a problem. Which nurse works which shift? Which vehicle takes which route? Which exam goes in which room? For many optimisation problems, there is no known algorithm that can find optimal solutions in a reasonable time.
VibeSolve targets constraint-based combinatorial problems such as employee scheduling, timetabling, vehicle routing and resource allocation. It is suited to problems that require a quick, near-optimal solution. If your problem involves assigning items to other items under a set of rules, VibeSolve is likely to be a good fit.
Anyone with an OpenAI or Claude API key can use it. Whether you’re an AI enthusiast who is new to optimisation or an optimisation expert, VibeSolve aims to lower the barrier and make creating optimisation projects more accessible.
General-purpose AI tools can write code, but they lack deep knowledge of solver architecture, constraint modelling and validation. VibeSolve is built by experts with PhDs in optimisation and a proven track record of solving real-world problems outside academia. We understand how and why an optimisation project can fail. We know how to make an optimisation project a success.
VibeSolve uses a specialised multi-agent pipeline with embedded optimisation expertise. This gives you excellent output quality as well as good results with smaller LLM models and lower token costs.
Yes. VibeSolve is self-hosted and uses your own OpenAI or Claude API key, so you stay in control of costs and data.
VibeSolve currently supports OpenAI and Claude models.
Token usage depends on the selected LLM model and the number of iterations required in the automated fixer loop. VibeSolve supports multiple OpenAI and Claude models and allows you to configure the maximum number of fixer-loop iterations, so you’re in control of token consumption and cost.
For typical usage, see the Benchmarking section, which includes expected cost when using VibeSolve’s default configuration: GPT-5 Mini with a maximum of 10 fixer-loop iterations.
VibeSolve generates projects built on Timefold Solver (v1.31.0), an open-source, production-grade constraint solver for Java. Support for other solvers is not yet available, but ideas and contributions are welcome.
VibeSolve automatically validates every generated project by compiling and running it in Docker. That said, automated tests can’t replace real software developers and business process experts’ domain expertise. If you want to feel more confident about your production use case, talk to our optimisation specialists for an expert review of your solver.
Absolutely! VibeSolve is open source and actively developed. Bug reports, new problem descriptions, prompt improvements, and feature PRs are all welcome. See CONTRIBUTING.md for further contribution guidelines.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。