Prompting and Context Engineering

Aider Repo Map is Aider's automatic repository summarization system that gives an LLM a compact, token-efficient map of relevant files, symbols, and relationships in a codebase.

Cursor Rules and Codebase Indexing are Cursor IDE features for steering AI behavior with project-specific instructions and giving the AI semantic awareness of a repository.

Axolotl is an open-source framework for fine-tuning large language models using configurable training recipes for LoRA, QLoRA, full fine-tuning, preference tuning, and related workflows.

OpenAI Responses API is OpenAI's unified API for building model-powered applications that combine text generation, multimodal inputs, tool use, structured outputs, and conversational context management.

Prompt caching is a context-engineering technique that reuses previously processed prompt prefixes to reduce latency and cost for repeated or long-context AI requests.

Ragas is an open-source Python framework for evaluating, monitoring, and improving retrieval-augmented generation and other LLM applications using automated metrics and test datasets.

DSPy is a Python framework for programming and automatically optimizing language-model pipelines using declarative modules, signatures, and metric-driven compilation instead of hand-written prompts.

Microsoft GraphRAG is an open-source framework for building retrieval-augmented generation systems that use knowledge graphs and graph-based indexing to improve LLM answers over complex private datasets.

LlamaIndex is a data framework for building LLM applications that connect private, structured, and unstructured data to language models through indexing, retrieval, agents, and workflow orchestration.

Retrieval-augmented generation techniques adapted for code repositories. It enables AI tools to understand large codebases through semantic search, chunking strategies, and context management.

The practice of designing effective prompts and instructions for AI coding tools to maximize output quality, reduce hallucinations, and improve task completion rates across different model types.