Awesome Context Engineering

Report Abuse

Basic Information

This repository is a comprehensive, curated survey and resource hub for Context Engineering — the practice of assembling, formatting and optimizing the complete information payload given to large language models at inference time. It collects and organizes academic papers, surveys, benchmarks, theoretical frameworks and implementation notes that contrast static prompt engineering with dynamic, structured context assembly. The README defines context engineering formally, presents a Bayesian and optimization-based view, and breaks the field into components such as retrieval-augmented generation, memory systems, agent communication, tool/function calling, context scaling and multimodal integration. It also links to published survey work from the authors and to community channels, and provides contribution, citation and licensing information to support ongoing research and collaborative updates.

Links

Categorization

App Details

Features
Extensive, well-structured bibliography and annotated links to surveys, arXiv papers and benchmarks across context engineering topics. Formal definitions and mathematical models that reframe context as an assembly of instructions, knowledge, tools, memory, state and query. Topic sections covering components and architectures: RAG, memory systems, long-context techniques, graph-based and multimodal methods, self-generated context and tool/function-calling. Evaluation and benchmarking coverage including long-context and multimodal benchmarks. Implementation and production considerations for enterprise systems, a list of open-source toolkits and leaderboards, contribution guidelines and community links for discussion. MIT license and citation information for a published survey paper.
Use Cases
The repository serves as a single reference point for researchers, engineers and decision makers who need to understand or build context-aware LLM systems. It consolidates foundational theory, practical architectures and evaluation paradigms so readers can locate key papers, compare methods, and follow recommended best practices for retrieval, memory and context assembly. Practitioners can use the benchmarks and implementation notes to guide system design and trade-offs for latency, token budgets and reliability. Academics gain a curated literature map and citation for the authors" survey paper. The community links and contribution guide make it easy to stay current and to add new research or tools as the field evolves.

Please fill the required fields*