Reading
Deliberate is an approval middleware specifically designed for LangGraph agents, providing a policy engine, notification system, and multi-approver workflow to help enterprises find a safe and controllable balance between AI automation and human supervision.
Section 01
Deliberate is an approval middleware specifically designed for LangGraph agents, offering a policy engine, notification system, and multi-approver workflow. It helps enterprises strike a safe and controllable balance between AI automation and human supervision, addressing the trust gap in AI decision-making within high-risk domains.
Section 02
As agent frameworks like LangGraph and AutoGPT mature, AI has evolved from simple Q&A tools to intelligent agents capable of autonomous planning and tool invocation. While efficiency has improved, this raises a core concern for enterprises: how to keep AI efficient while ensuring human supervision over critical decisions? Fully autonomous AI decisions are unacceptable in high-risk fields (finance, healthcare, law, etc.). Enterprises need a "human-AI collaboration" model—where AI performs initial judgments and humans approve key nodes. This is how the Deliberate project came into being.
Section 03
Deliberate is positioned as the "approval layer" for LangGraph agents, serving as a complete enterprise-grade approval infrastructure. Deeply integrated with LangGraph, it allows inserting approval nodes anywhere in the agent's execution path, dynamically determining whether approval is needed. During the approval wait, it maintains state persistence, and after approval, execution resumes seamlessly without disrupting the original workflow logic.
Section 04
Section 05
Section 06
Section 07
| Feature | Deliberate | Traditional Workflow Engine | Simple Approval Plugin |
|---|---|---|---|
| AI-Native Design | ✓ | ✗ | △ |
| LangGraph Integration | Deep Native | Needs Adaptation | Shallow Integration |
| Policy Engine | Intelligent Trigger | Rule-Driven | Simple Conditions |
| Context Awareness | ✓ | ✗ | ✗ |
| Multi-Approver Workflow | ✓ | ✓ | Limited Support |
Deliberate's unique value lies in being "designed for AI", understanding the semantic behavior of agents, and enabling more intelligent approval decisions.
Section 08
Future Directions: Adaptive policies (machine learning to optimize approval rules), predictive approval (preparing context in advance), multi-modal interfaces (voice/video interaction), cross-framework support (extending to other agent frameworks).
Conclusion: Deliberate addresses the pain points of enterprise AI applications, promotes the long-term "human-AI collaboration" model, serves as a practical paradigm for responsible AI, helps AI move from the lab to production environments, and provides reliable approval infrastructure for enterprise-level agents.
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
Cross-generational hardware testing based on NVIDIA A10G and H100 GPUs, comparing and analyzing the throughput, latency, and concurrent scalability of three mainstream LLM inference frameworks. SGLang achieves a 3.4x performance advantage over vLLM on H100, while Ollama faces architectural bottlenecks in high-concurrency scenarios.