GraphBit: A Graph-based Agentic Framework for Non-Linear Agent Orchestration

TL;DR

GraphBit is a deterministic, graph-based framework for orchestrating agent workflows that improves reproducibility, efficiency, and accuracy over prompted methods by explicitly defining workflows as DAGs.

Contribution

It introduces a novel engine-controlled, DAG-based architecture with a three-tier memory system, outperforming existing frameworks in accuracy and reliability.

Findings

Achieves 67.6% accuracy on GAIA benchmark tasks.

Zero hallucinations observed during execution.

Lowest latency of 11.9 ms overhead.

Abstract

Agentic LLM frameworks that rely on prompted orchestration, where the model itself determines workflow transitions, often suffer from hallucinated routing, infinite loops, and non-reproducible execution. We introduce GraphBit, an engine-orchestrated framework that defines workflows explicitly and deterministically as a directed acyclic graph (DAG). Unlike prompted orchestration, agents in GraphBit operate as typed functions, while a Rust-based engine governs routing, state transitions, and tool invocation, ensuring reproducibility and auditability. The engine supports parallel branch execution, conditional control flow over structured state predicates, and configurable error recovery. A three-tier memory architecture consisting of ephemeral scratch space, structured state, and external connectors isolates context across stages, preventing cascading context bloat that degrades reasoning in long-running pipelines. Across GAIA benchmark tasks spanning zero-tool, document-augmented, and web-enabled workflows, GraphBit outperforms six existing frameworks, achieving the highest accuracy (67.6 percent), zero framework-induced hallucinations, the lowest latency (11.9 ms overhead), and the highest throughput. Ablation studies demonstrate that each memory tier contributes measurably to performance, with deterministic execution providing the greatest gains on tool-intensive tasks representative of real-world deployments.