Reading
Recombine Engine is a TypeScript library designed specifically for building agent workflows for conversational AI. It supports multi-step workflow orchestration, conditional execution, structured responses, file-based prompt management, and message history management, and can be integrated with the Recombine Bosun prompt engineering IDE.
Section 01
Recombine Engine is a TypeScript library designed specifically for building agent workflows for conversational AI. It supports multi-step workflow orchestration, conditional execution, structured responses, file-based prompt management, and message history management, and can be integrated with the Recombine Bosun prompt engineering IDE.
Section 02
In the development practice of conversational AI, simple single-turn Q&A often cannot meet the needs of complex business scenarios. Users expect AI to perform multi-step tasks, make decisions based on intermediate results, and coordinate multiple sub-agents to work together. Recombine Engine is a TypeScript library born to address these complex orchestration needs.
The project's design philosophy emphasizes declarative workflow definition and type safety. Developers can define each step of the workflow through clear configuration objects, use TypeScript's type system to catch potential errors at compile time, while maintaining code readability and maintainability. This design concept makes complex agent workflows easy to understand, test, and iterate.
Section 03
Recombine Engine provides a complete set of features needed to build conversational AI workflows:
Section 04
A workflow consists of a series of interconnected steps (Step). Each step can access the results of previous steps and determine the execution path of the next step based on business logic. This chain structure supports complex decision trees, loops, and conditional branches.
Section 05
Each step can be configured with conditional execution logic (runIf) to decide whether to skip the step. It also supports the Reviewer mode—when the output of a step does not meet expectations, it can automatically retry or roll back to the previous step for re-execution. This mechanism provides a foundation for building robust fault-tolerant systems.
Section 06
Through integration with the Zod library, developers can define strict output schemas (Schema) for each step. The responses returned by the model are automatically validated and parsed to ensure data structure consistency. This not only improves code reliability but also enables IDEs to provide better auto-completion support.
Section 07
The project supports storing prompt templates in independent text files, which can be loaded via the engine.loadFile() method. This separation allows prompt engineering to proceed in parallel with code development, and non-technical personnel can also participate in prompt optimization without modifying the code.
Section 08
The built-in message list manager (makeMessagesList) is responsible for maintaining conversation context. It supports adding user messages, system messages, and assistant messages, and can automatically handle message truncation and context window management.
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.