Generative AI and AI Agents
Who this is for:
Architecture / Concept Overview: Generative AI and AI Agents
The Databricks Generative AI stack connects foundation models, retrieval, agent orchestration, and governance into a unified platform.
%%{init: {"theme":"base","themeVariables":{"background":"#0B0E14","primaryTextColor":"#E0E6ED","lineColor":"#5D6470","darkMode":true,"primaryColor":"#2E4A4A","secondaryColor":"#374151","secondaryTextColor":"#E0E6ED","tertiaryColor":"#111827","tertiaryTextColor":"#E0E6ED","edgeLabelBackground":"#1f2937"}}}%%
flowchart LR
classDef source fill:#3F4B59,stroke:#9CA3AF,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef ingestion fill:#5A4B36,stroke:#C9A86B,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef processing fill:#535072,stroke:#8E82B4,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef storage fill:#2E4A4A,stroke:#5FAFA8,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef serving fill:#3D5550,stroke:#6BB7AA,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef governance fill:#5A3F52,stroke:#C28BB0,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
DOCS[Enterprise Documents] -->|Chunk & Embed| VS[Vector Search Index]
VS -->|Retrieve| RAG[RAG Chain]
FM[Foundation Models] -->|Generate| RAG
RAG -->|Orchestrate| AGENT[AI Agent]
AGENT -->|Use| TOOLS[Tools & APIs]
AGENT -->|Deploy| EP[Serving Endpoint]
EP -->|Monitor| EVAL[Agent Evaluation]
EP -->|Govern| GW[AI Gateway]
DOCS:::source
VS:::storage
FM:::processing
RAG:::processing
AGENT:::serving
TOOLS:::ingestion
EP:::serving
EVAL:::governance
GW:::governance
*Generative AI stack: documents feed vector search, foundation models power RAG chains, agents orchestrate tools, and governance wraps the entire surface.*
%%{init: {"theme":"base","themeVariables":{"background":"#0B0E14","primaryTextColor":"#E0E6ED","lineColor":"#5D6470","darkMode":true,"primaryColor":"#2E4A4A","secondaryColor":"#374151","secondaryTextColor":"#E0E6ED","tertiaryColor":"#111827","tertiaryTextColor":"#E0E6ED","edgeLabelBackground":"#1f2937"}}}%%
graph TD
classDef source fill:#3F4B59,stroke:#9CA3AF,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef ingestion fill:#5A4B36,stroke:#C9A86B,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef processing fill:#535072,stroke:#8E82B4,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef storage fill:#2E4A4A,stroke:#5FAFA8,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef serving fill:#3D5550,stroke:#6BB7AA,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
classDef governance fill:#5A3F52,stroke:#C28BB0,stroke-width:2px,rx:8,ry:8,color:#E0E6ED
GENAI[Generative AI Capabilities] --> MODELS[Foundation Models]
GENAI --> RAG_SYS[RAG Systems]
GENAI --> AGENTS[AI Agents]
GENAI --> FINETUNE[Fine-Tuning]
MODELS --> LLAMA[Meta Llama]
MODELS --> GPT[GPT]
MODELS --> CLAUDE[Claude]
MODELS --> DBRX[DBRX]
RAG_SYS --> VECTSEARCH[Vector Search]
RAG_SYS --> EMBED[Embedding Models]
AGENTS --> FRAMEWORK[Agent Framework]
AGENTS --> BRICKS[Agent Bricks]
AGENTS --> MULTI[Multi-Agent Systems]
FINETUNE --> SFT[Supervised Fine-Tuning]
FINETUNE --> RLHF[RLHF]
GENAI:::governance
MODELS:::processing
RAG_SYS:::storage
AGENTS:::serving
FINETUNE:::ingestion
LLAMA:::source
GPT:::source
CLAUDE:::source
DBRX:::source
VECTSEARCH:::storage
EMBED:::storage
FRAMEWORK:::serving
BRICKS:::serving
MULTI:::serving
SFT:::ingestion
RLHF:::ingestion
*Generative AI capability tree: foundation models, RAG, agents, and fine-tuning.*
Key Terms
Prerequisites and Setup
- Databricks workspace (Premium or Enterprise) with Unity Catalog.
- Foundation Model APIs enabled (available in supported regions).
- Vector Search endpoints configured for RAG workloads.
- Python SDK:
databricks-agents,mlflow.
Step-by-Step Implementation
Configuration Reference
| Parameter | Default | Description |
|---|---|---|
max_tokens | 256 | Maximum tokens in LLM response |
temperature | 1.0 | Sampling temperature (lower = more deterministic) |
num_results (Vector Search) | 10 | Number of nearest neighbours to retrieve |
index_type | DELTA_SYNC | Vector index type: DELTA_SYNC or DIRECT_ACCESS |
pipeline_type | TRIGGERED | Sync mode: TRIGGERED or CONTINUOUS |
embedding_model_endpoint_name | — | Model endpoint used for automatic embedding |