Machine LearningSystem architecture

RAG LLM Agent Architecture

A retrieval-augmented agent architecture covering ingestion, retrieval, generation, tools, and feedback loops.

學科: Machine Learning
輸入方式: Text To Image
視覺風格: Technical schematic
畫布比例: 16:9

RAG LLM Agent Architecture template preview

預覽提示詞

Create a horizontal system-architecture diagram of a retrieval-augmented-generation (RAG) LLM agent with three connected lanes plus a surrounding agent loop. Top-left ingestion lane: a stack of documents flows into a "Chunker (512 tokens, 50 overlap)" box, then an "Embedding Model (text-3-large, 1536-d)" box, then a cylinder labeled "Vector DB (1.2M chunks)". Center query lane: "User Query" embedded, arrow to "Top-k Similarity Search (k=5, cosine)" reading from the vector DB, then "Prompt Template (context + question)" feeding an "LLM" box that outputs a "Grounded Answer with citations [1][2]". Wrap these in an agent loop with rounded nodes: Planner, Tool Calling (web search, calculator, code), and Short-term / Long-term Memory, joined by curved feedback arrows. Add a small inset bar chart titled "Answer Accuracy" comparing No-RAG 61% vs RAG 89%, and latency tags reading "retrieve 40 ms" and "generate 1.2 s". Use a clean technical style, labeled arrows, monospace component captions, soft blue and teal accents on white.

相關範本

返回全部範本

Genomics

CRISPR-Cas9 Editing Mechanism

A four-stage gene-editing mechanism with sgRNA targeting, Cas9 cleavage, NHEJ, and HDR outcomes.

Genomics

Single-Cell RNA-seq Workflow

A six-stage single-cell sequencing workflow from tissue dissociation through UMAP clustering.

Machine Learning

Transformer Model Architecture

An encoder architecture with attention, residual paths, feed-forward blocks, and dimensional annotations.

RAG LLM Agent Architecture 科研繪圖範本 | LabFig