Learn the basics of LangChain to build AI applications. From installation to your first functional pattern.
Share your comments or suggestions to improve our content.
Learn to treat your agent's context window as the finite resource it is: the four levers of context engineering (write, select, compress, isolate), compaction with structured summaries, persistent memory outside the context, just-in-time retrieval, subagents with isolated windows, and tool-result hygiene. With production-ready Python code and the mistakes that silently degrade your agent. New expansion: the context-failure taxonomy (poisoning, distraction, confusion, clash), reasoning budgets with interleaved extended thinking, the real cost of multimodal context, managed memory (Letta, Zep, mem0), effective context length per RULER, and multi-agent handoffs with structured payloads. Latest expansion: the positional anatomy of context (lost in the middle and cache-aware placement), isolating untrusted content by design with Dual-LLM and CaMeL, and generating outputs longer than the window with outlines, rolling summaries, and patch-based revision. August extension: per-section token budgets with a degradation ladder, prompt compression with LLMLingua, semantic retrieval deduplication, self-hosted KV cache (vLLM and SGLang) with cache-aware routing, and memory evaluation with LongMemEval.
Battle-tested patterns for building production-ready RAG (Retrieval-Augmented Generation) systems in 2026: semantic chunking, embedding selection, hybrid search, and cross-encoder reranking.
A practical guide to the number one vulnerability in the OWASP Top 10 for LLM Applications: why prompt injection cannot be patched the way SQL injection can, the difference between direct and indirect injection, and the lethal trifecta threat model (private data + untrusted content + external communication) as a tool for auditing what your agent can call. Covers six defense layers with production-ready code: random-id delimiting and spotlighting, real least privilege using end-user permissions and the tenant_id-as-model-argument mistake, egress control against Markdown image exfiltration with allowlists, CSP and network policy, per-action human approval with summaries rendered from validated arguments, the Dual LLM / CaMeL pattern that makes the trifecta impossible by construction, and blast radius containment with sandboxing and short-lived credentials. Includes adversarial evals in pytest with an ASR budget in CI, the canary trick, eight recurring mistakes (RAG index poisoning, third-party MCP servers, classifiers as the only barrier) and an eleven-point production checklist.