Oracle AI Architect

Oracle AI Architect - Extended Reference

Purpose

Comprehensive implementation reference for Oracle AI Architects building enterprise AI solutions. Covers OCI Generative AI APIs, Database 26ai Vector Search, Select AI, NVIDIA NIM deployment, and multi-agent patterns with Oracle ADK.

This document provides extended reference material for the Oracle AI Architect skill.

OCI Generative AI API Examples

Chat Completion

import oci
from oci.generative_ai_inference import GenerativeAiInferenceClient
from oci.generative_ai_inference.models import (
    ChatDetails,
    CohereChatRequest,
    OnDemandServingMode
)

config = oci.config.from_file()
client = GenerativeAiInferenceClient(config)

chat_request = CohereChatRequest(
    message="Explain Oracle AI Vector Search",
    max_tokens=500,
    temperature=0.7
)

response = client.chat(
    chat_details=ChatDetails(
        compartment_id=compartment_id,
        serving_mode=OnDemandServingMode(model_id="cohere.command-a"),
        chat_request=chat_request
    )
)

print(response.data.chat_response.text)

Embeddings

from oci.generative_ai_inference.models import (
    EmbedTextDetails,
    OnDemandServingMode
)

embed_request = EmbedTextDetails(
    inputs=["Document text to embed"],
    serving_mode=OnDemandServingMode(model_id="cohere.embed-english-v3.0"),
    compartment_id=compartment_id
)

response = client.embed_text(embed_request)
embeddings = response.data.embeddings

Database 26ai AI Vector Search Examples

Creating Vector Tables

-- Basic vector column
CREATE TABLE docs (
    id NUMBER GENERATED ALWAYS AS IDENTITY,
    content CLOB,
    embedding VECTOR(1536, FLOAT32),
    CONSTRAINT docs_pk PRIMARY KEY (id)
);

-- With automatic embedding generation
CREATE TABLE smart_docs (
    id NUMBER GENERATED ALWAYS AS IDENTITY,
    content CLOB,
    embedding VECTOR GENERATED ALWAYS AS (
        VECTOR_EMBEDDING(content USING 'doc-embed-model')
    ) VIRTUAL
);