LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
Client Overview
Hexuan Cai is an independent systems architect exploring the infrastructure layer above agent execution frameworks. With a background in protocol design and enterprise systems, he identified a gap in how multi-agent systems make coordination decisions — and wanted a rigorous reference implementation to demonstrate what that layer could look like.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
The Results
The Results
The Results
Delivered a production-grade reference implementation in two weeks — on spec, on budget, and fully documented. The system handles all three coordination decisions deterministically, with zero ML dependencies and complete explainability on every output.
3
3
3
Endpoints Delivered
2 weeks
2 weeks
2 weeks
From Spec to Live Deployment
100%
100%
100%
Deterministic — Every Decision Reproducible
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
Client Overview
Hexuan Cai is an independent systems architect exploring the infrastructure layer above agent execution frameworks. With a background in protocol design and enterprise systems, he identified a gap in how multi-agent systems make coordination decisions — and wanted a rigorous reference implementation to demonstrate what that layer could look like.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
The Results
The Results
The Results
Delivered a production-grade reference implementation in two weeks — on spec, on budget, and fully documented. The system handles all three coordination decisions deterministically, with zero ML dependencies and complete explainability on every output.
3
3
3
Endpoints Delivered
2 weeks
2 weeks
2 weeks
From Spec to Live Deployment
100%
100%
100%
Deterministic — Every Decision Reproducible
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
LTV Protocol — Deterministic Decision Layer for Agentic Systems
Client Overview
Hexuan Cai is an independent systems architect exploring the infrastructure layer above agent execution frameworks. With a background in protocol design and enterprise systems, he identified a gap in how multi-agent systems make coordination decisions — and wanted a rigorous reference implementation to demonstrate what that layer could look like.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Challenge
01
As multi-agent systems scale, execution becomes cheap but coordination becomes the bottleneck. There was no standard way to answer: which tasks should run, in what order, and with how much budget? Most frameworks run everything and hope for the best. Hexuan needed a clean, structured decision interface — one that was fully deterministic, explainable, and configurable without any ML or probabilistic inference, so it could serve as a credible protocol reference rather than just another AI tool.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
The Solution
02
Genta built the LTV Protocol — a stateless, config-driven decision API that sits above any agent execution framework. The system exposes three endpoints: /evaluate scores a single task and returns a decision with a full reasoning trail, /prioritize ranks a list of competing tasks deterministically, and /allocate distributes a budget pool using a greedy priority strategy. All scoring logic is driven by a single YAML config file — weights, thresholds, and reason code rules are all editable without touching code. Every output includes structured reason codes derived from explicit thresholds, making every decision fully auditable. The system was deployed on Railway and delivered within two weeks.
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
Technologies Used
03
• Python • FastAPI • Pydantic • PyYAML • Uvicorn • pytest • Railway (deployment)
The Results
The Results
The Results
Delivered a production-grade reference implementation in two weeks — on spec, on budget, and fully documented. The system handles all three coordination decisions deterministically, with zero ML dependencies and complete explainability on every output.
3
3
3
Endpoints Delivered
2 weeks
2 weeks
2 weeks
From Spec to Live Deployment
100%
100%
100%
Deterministic — Every Decision Reproducible
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.
We’re Here to Help
Ready to transform your operations? We're here to help. Contact us today to learn more about our innovative solutions and expert services.