select_ai documentation

select_ai is a Python module which enables integrating DBMS_CLOUD_AI PL/SQL package into Python workflows. It bridges the gap between PL/SQL package’s AI capabilities and Python’s rich ecosystem.

Getting Started

• 1. Introduction to Select AI for Python
  • 1.1. What you can build
  • 1.2. Core concepts
  • 1.3. Synchronous and asynchronous APIs
  • 1.4. Supported Python versions
• 2. Installing select_ai
  • 2.1. Installation requirements
  • 2.2. select_ai installation
  • 2.3. Connection smoke test
  • 2.4. Install documentation dependencies
• 3. Connecting to Oracle Database
  • 3.1. Synchronous connection
  • 3.2. Asynchronous connection
  • 3.3. Connection Pool
  • 3.4. Connection health
  • 3.5. Wallet connections
• 4. Run your first SQL query
  • 4.1. Install and configure Python
  • 4.2. Prepare the database account
  • 4.3. Run the example
  • 4.4. Waiting and timeouts
• 5. Compatibility and requirements
• 6. What’s new in 1.5.0
  • 6.1. New and expanded APIs
  • 6.2. Upgrade considerations

Actions

• 1. Supported Actions
  • 1.1. Action methods
  • 1.2. Choosing an action
  • 1.3. Examples
  • 1.4. Streaming

Privileges

• 1. Grant privilege
• 2. Revoke privilege
• 3. Grant network access
• 4. Revoke network access

Provider

• 1. Provider
  • 1.1. Examples
  • 1.2. Provider
    • Provider
  • 1.3. AnthropicProvider
    • AnthropicProvider
  • 1.4. AzureProvider
    • AzureProvider
  • 1.5. AWSProvider
    • AWSProvider
  • 1.6. CohereProvider
    • CohereProvider
  • 1.7. OpenAIProvider
    • OpenAIProvider
  • 1.8. OCIGenAIProvider
    • OCIGenAIProvider
  • 1.9. GoogleProvider
    • GoogleProvider
  • 1.10. HuggingFaceProvider
    • HuggingFaceProvider
  • 1.11. Enable AI service provider
    • 1.11.1. Enable using Sync API
    • 1.11.2. Enable using Async API
  • 1.12. Cloud provider profile samples
    • 1.12.1. AWS Bedrock
    • 1.12.2. Azure OpenAI
    • 1.12.3. Google Gemini (GCP)
  • 1.13. Disable AI service provider
    • 1.13.1. Disable using Sync API
    • 1.13.2. Disable using Async API

Credential

• 1. Create credential
  • 1.1. Create Sync API
  • 1.2. Create Async API
• 2. Delete credential
  • 2.1. Delete Sync API
  • 2.2. Delete Async API
• 3. Credential access

Sharing and ownership

• 1. Sharing and ownership
  • 1.1. Connect before running an example
  • 1.2. Profiles
  • 1.3. Vector indexes
  • 1.4. Agent teams
  • 1.5. Credentials
  • 1.6. Runnable sharing scripts

Profile Attributes

• 1. ProfileAttributes
  • 1.1. Common required attributes
  • 1.2. Attribute groups
  • 1.3. Object list examples
  • 1.4. Generation controls
  • 1.5. Additional instructions
    • ProfileAttributes

Profile

• 1. Profile lifecycle
• 2. Profile actions
  • 2.1. Profile Object Model
  • 2.2. Base Profile API
    • BaseProfile
  • 2.3. Profile API
    • Profile
  • 2.4. Create Profile
  • 2.5. Reuse Profile
  • 2.6. Enable and disable Profile
  • 2.7. Update Profile
  • 2.8. Delete Profile
  • 2.9. Narrate
  • 2.10. Show SQL
  • 2.11. Explain SQL
  • 2.12. Show Prompt
  • 2.13. Run SQL
  • 2.14. Chat
  • 2.15. Streaming chat
  • 2.16. Request-level profile attribute overrides
  • 2.17. Summarize
  • 2.18. Translate
  • 2.19. Share a profile
  • 2.20. Give feedback on generated SQL
  • 2.21. List profiles
    • 2.21.1. Connect and list your profiles
    • 2.21.2. Filter by name
    • 2.21.3. Find and reuse a shared profile

Async Profile

• 1. Async profile lifecycle
• 2. Async profile actions
  • 2.1. AsyncProfile API
    • AsyncProfile
  • 2.2. Async Profile creation
  • 2.3. Reuse Async Profile
  • 2.4. Enable and disable AsyncProfile
  • 2.5. Update Async Profile
  • 2.6. Delete Async Profile
  • 2.7. Async explain SQL
  • 2.8. Async run SQL
  • 2.9. Async show SQL
  • 2.10. Async show prompt
  • 2.11. Async concurrent SQL
  • 2.12. Async chat
  • 2.13. Async streaming chat
  • 2.14. Async request-level profile attribute overrides
  • 2.15. Summarize
  • 2.16. Translate
  • 2.17. Async pipeline
    • 2.17.1. Pipeline compatibility and results
  • 2.18. Share a profile asynchronously
  • 2.19. Give async feedback on generated SQL
  • 2.20. List profiles asynchronously
    • 2.20.1. Connect and list your profiles
    • 2.20.2. Filter by name
    • 2.20.3. Find and reuse a shared profile

Conversation

• 1. Conversation Object model
• 2. ConversationAttributes
  • ConversationAttributes
    • ConversationAttributes.conversation_length
    • ConversationAttributes.description
    • ConversationAttributes.dict()
    • ConversationAttributes.json()
    • ConversationAttributes.keys()
    • ConversationAttributes.retention_days
    • ConversationAttributes.title
• 3. Conversation API
  • Conversation
    • Conversation.add_tag()
    • Conversation.create()
    • Conversation.delete()
    • Conversation.delete_prompt()
    • Conversation.fetch()
    • Conversation.get_attributes()
    • Conversation.list()
    • Conversation.list_prompts()
    • Conversation.remove_tag()
    • Conversation.set_attributes()
  • 3.1. Create conversation
  • 3.2. Chat session
  • 3.3. Prompt history and tags
    • ConversationPrompt
  • 3.4. List conversations
  • 3.5. Delete conversation
• 4. AsyncConversation API
  • AsyncConversation
    • AsyncConversation.add_tag()
    • AsyncConversation.create()
    • AsyncConversation.delete()
    • AsyncConversation.delete_prompt()
    • AsyncConversation.fetch()
    • AsyncConversation.get_attributes()
    • AsyncConversation.list()
    • AsyncConversation.list_prompts()
    • AsyncConversation.remove_tag()
    • AsyncConversation.set_attributes()
  • 4.1. Async chat session
  • 4.2. Async prompt history and tags
    • 4.2.1. Create a conversation and save a prompt
    • 4.2.2. Tag and inspect the conversation
    • 4.2.3. Remove a stored prompt and its tag
    • 4.2.4. Run the complete example
  • 4.3. Async list conversations

Vector Index

• 1. Vector Index
• 2. VectorIndex Object Model
• 3. VectorIndexAttributes
  • VectorIndexAttributes
    • VectorIndexAttributes.chunk_overlap
    • VectorIndexAttributes.chunk_size
    • VectorIndexAttributes.create()
    • VectorIndexAttributes.dict()
    • VectorIndexAttributes.enable_sources
    • VectorIndexAttributes.json()
    • VectorIndexAttributes.keys()
    • VectorIndexAttributes.location
    • VectorIndexAttributes.match_limit
    • VectorIndexAttributes.object_storage_credential_name
    • VectorIndexAttributes.pipeline_name
    • VectorIndexAttributes.profile_name
    • VectorIndexAttributes.refresh_rate
    • VectorIndexAttributes.similarity_threshold
    • VectorIndexAttributes.vector_db_credential_name
    • VectorIndexAttributes.vector_db_endpoint
    • VectorIndexAttributes.vector_db_provider
    • VectorIndexAttributes.vector_dimension
    • VectorIndexAttributes.vector_distance_metric
    • VectorIndexAttributes.vector_table_name
  • 3.1. OracleVectorIndexAttributes
    • OracleVectorIndexAttributes
• 4. VectorIndex API
  • VectorIndex
    • VectorIndex.create()
    • VectorIndex.delete()
    • VectorIndex.delete_index()
    • VectorIndex.disable()
    • VectorIndex.enable()
    • VectorIndex.fetch()
    • VectorIndex.get_attributes()
    • VectorIndex.get_next_refresh_timestamp()
    • VectorIndex.get_profile()
    • VectorIndex.grant_access()
    • VectorIndex.list()
    • VectorIndex.qualified_name
    • VectorIndex.revoke_access()
    • VectorIndex.set_attribute()
    • VectorIndex.set_attributes()
  • 4.1. Create vector index
  • 4.2. List vector index
  • 4.3. Fetch vector index
  • 4.4. Update vector index attributes
  • 4.5. RAG using vector index
  • 4.6. Delete vector index
• 5. AsyncVectorIndex API
  • AsyncVectorIndex
    • AsyncVectorIndex.create()
    • AsyncVectorIndex.delete()
    • AsyncVectorIndex.delete_index()
    • AsyncVectorIndex.disable()
    • AsyncVectorIndex.enable()
    • AsyncVectorIndex.fetch()
    • AsyncVectorIndex.get_attributes()
    • AsyncVectorIndex.get_next_refresh_timestamp()
    • AsyncVectorIndex.get_profile()
    • AsyncVectorIndex.grant_access()
    • AsyncVectorIndex.list()
    • AsyncVectorIndex.qualified_name
    • AsyncVectorIndex.revoke_access()
    • AsyncVectorIndex.set_attribute()
    • AsyncVectorIndex.set_attributes()
  • 5.1. Async create vector index
  • 5.2. Async list vector index
  • 5.3. Async fetch vector index
  • 5.4. Async update vector index attributes
  • 5.5. Async RAG using vector index
    • 5.5.1. Share a vector index

Synthetic Data

• 1. Generation modes
• 2. Generation parameters
• 3. Sync and async APIs
  • 3.1. SyntheticDataAttributes
    • SyntheticDataAttributes
  • 3.2. SyntheticDataParams
    • SyntheticDataParams
  • 3.3. Single table synthetic data
    • 3.3.1. Single Table Sync API
    • 3.3.2. Single Table Async API
  • 3.4. Multi table synthetic data
    • 3.4.1. Multi table Sync API
    • 3.4.2. Multi table Async API

Summary

• 1. Inline content
• 2. Content from a URI
• 3. Content from object storage
• 4. Summary parameters
• 5. Async summary
• 6. Validation
  • 6.1. SummaryParams
    • SummaryParams
  • 6.2. ChunkProcessingMethod
    • ChunkProcessingMethod
  • 6.3. ExtractivenessLevel
    • ExtractivenessLevel
  • 6.4. SummaryStyle
    • Style

AI Agent

• 1. Tool
  • 1.1. Supported Tools
  • 1.2. Tool selection
    • ToolAttributes
    • ToolParams
    • Tool
  • 1.3. Create Tool
  • 1.4. List Tools
  • 1.5. Inspect and run a tool
• 2. Task
  • TaskAttributes
    • TaskAttributes.dict()
    • TaskAttributes.enable_human_tool
    • TaskAttributes.input
    • TaskAttributes.instruction
    • TaskAttributes.json()
    • TaskAttributes.keys()
    • TaskAttributes.tools
  • Task
    • Task.create()
    • Task.delete()
    • Task.delete_task()
    • Task.disable()
    • Task.enable()
    • Task.fetch()
    • Task.list()
    • Task.set_attribute()
    • Task.set_attributes()
  • 2.1. Create Task
  • 2.2. List Tasks
• 3. Agent
  • AgentAttributes
    • AgentAttributes.dict()
    • AgentAttributes.enable_human_tool
    • AgentAttributes.json()
    • AgentAttributes.keys()
    • AgentAttributes.profile_name
    • AgentAttributes.role
    • AgentAttributes.supervisor
  • Agent
    • Agent.create()
    • Agent.delete()
    • Agent.delete_agent()
    • Agent.disable()
    • Agent.enable()
    • Agent.fetch()
    • Agent.list()
    • Agent.set_attribute()
    • Agent.set_attributes()
  • 3.1. Create Agent
• 4. Team
  • TeamAttributes
    • TeamAttributes.agents
    • TeamAttributes.dict()
    • TeamAttributes.json()
    • TeamAttributes.keys()
    • TeamAttributes.process
    • TeamAttributes.supervisor_agent
    • TeamAttributes.supervisor_task
  • 4.1. Supervised teams
    • Team
  • 4.2. Share a team
  • 4.3. Run Team
  • 4.4. Export and Import Team
  • 4.5. Inspect a team
  • 4.6. Lifecycle helpers
    • 4.6.1. Object definitions
• 5. Agent history
  • TeamHistory
    • TeamHistory.list()
  • TaskHistory
    • TaskHistory.list()
  • ToolHistory
    • ToolHistory.list()
• 6. AI agent examples
  • 6.1. Web Search Agent using OpenAI’s GPT model

Async AI Agent

• 1. AsyncTool
  • AsyncTool
    • AsyncTool.create()
    • AsyncTool.create_built_in_tool()
    • AsyncTool.create_email_notification_tool()
    • AsyncTool.create_pl_sql_tool()
    • AsyncTool.create_rag_tool()
    • AsyncTool.create_slack_notification_tool()
    • AsyncTool.create_sql_tool()
    • AsyncTool.create_websearch_tool()
    • AsyncTool.delete()
    • AsyncTool.delete_tool()
    • AsyncTool.describe_tool()
    • AsyncTool.disable()
    • AsyncTool.enable()
    • AsyncTool.fetch()
    • AsyncTool.list()
    • AsyncTool.run_tool()
    • AsyncTool.set_attribute()
    • AsyncTool.set_attributes()
  • 1.1. Create Tool
  • 1.2. List Tools
  • 1.3. Inspect and run a tool
• 2. AsyncTask
  • AsyncTask
    • AsyncTask.create()
    • AsyncTask.delete()
    • AsyncTask.delete_task()
    • AsyncTask.disable()
    • AsyncTask.enable()
    • AsyncTask.fetch()
    • AsyncTask.list()
    • AsyncTask.set_attribute()
    • AsyncTask.set_attributes()
  • 2.1. Create Task
  • 2.2. List Tasks
• 3. AsyncAgent
  • AsyncAgent
    • AsyncAgent.create()
    • AsyncAgent.delete()
    • AsyncAgent.delete_agent()
    • AsyncAgent.disable()
    • AsyncAgent.enable()
    • AsyncAgent.fetch()
    • AsyncAgent.list()
    • AsyncAgent.set_attribute()
    • AsyncAgent.set_attributes()
  • 3.1. Create Agent
  • 3.2. List Agents
• 4. AsyncTeam
  • AsyncTeam
    • AsyncTeam.create()
    • AsyncTeam.delete()
    • AsyncTeam.delete_team()
    • AsyncTeam.describe_team()
    • AsyncTeam.disable()
    • AsyncTeam.enable()
    • AsyncTeam.export()
    • AsyncTeam.export_team()
    • AsyncTeam.fetch()
    • AsyncTeam.grant_access()
    • AsyncTeam.import_team()
    • AsyncTeam.list()
    • AsyncTeam.list_tools()
    • AsyncTeam.revoke_access()
    • AsyncTeam.run()
    • AsyncTeam.set_attribute()
    • AsyncTeam.set_attributes()
  • 4.1. Share an async team
  • 4.2. Supervised teams
  • 4.3. Run Team
  • 4.4. Export and Import Team
  • 4.5. Inspect a team
  • 4.6. Lifecycle helpers
    • 4.6.1. Object definitions
• 5. Async agent history
  • AsyncTeamHistory
    • AsyncTeamHistory.list()
  • AsyncTaskHistory
    • AsyncTaskHistory.list()
  • AsyncToolHistory
    • AsyncToolHistory.list()
  • 5.1. List Teams
• 6. Async AI agent examples
  • 6.1. Web Search Agent using OpenAI’s GPT model

A2A Integration

• 1. Agent2Agent (A2A)
  • 1.1. Choose a topology
  • 1.2. Why clustered deployments use Consul
  • 1.3. Choose how connections are configured
  • 1.4. CLI-based deployment on any cloud
  • 1.5. Installation and prerequisites
  • 1.6. Protocol compatibility
  • 1.7. Standalone A2A server
    • 1.7.1. Fixed database connection
    • 1.7.2. Dynamic standalone sessions
    • 1.7.3. End-user OAuth and session ownership
    • 1.7.4. Agent Card discovery
    • 1.7.5. Tasks, polling, streaming, and cancellation
  • 1.8. Persistent Oracle-backed state
  • 1.9. Wallet-based database connections
  • 1.10. Private database CA certificates
  • 1.11. Database call timeout
  • 1.12. Clustered A2A gateway and worker
    • 1.12.1. Clustered deployment with a fixed connection
    • 1.12.2. A2UI database-connection form
    • 1.12.3. Session isolation, routing, and cleanup
    • 1.12.4. Public gateway HTTPS
    • 1.12.5. Gateway-to-worker HTTP and TLS
  • 1.13. Google Cloud deployment
    • 1.13.1. Standalone Google Cloud deployment
    • 1.13.2. Clustered Google Cloud deployment
  • 1.14. Troubleshooting and security notes

Command Line Interface

• 1. Command line interface
  • 1.1. Connection options
  • 1.2. Interactive chat
  • 1.3. SQL commands
  • 1.4. Profile commands
  • 1.5. A2A commands
    • 1.5.1. A2A deployment examples
    • 1.5.2. Agent Card generation
  • 1.6. Command summary

Web Frameworks

• 1. Using select_ai with Python web frameworks
  • 1.1. Framework patterns
  • 1.2. Install dependencies
  • 1.3. FastAPI synchronous endpoints
  • 1.4. FastAPI asynchronous endpoints
  • 1.5. Flask example
  • 1.6. Django example
  • 1.7. Pool sizing
  • 1.8. Pool wait behavior
  • 1.9. Request handling

Concurrent Prompt Processing

• 1. Concurrent prompt processing
  • 1.1. Recipe summary
  • 1.2. Environment variables
  • 1.3. sync_thread_pool.py
  • 1.4. sync_ordered_results.py
  • 1.5. sync_queue_workers.py
  • 1.6. async_gather.py
  • 1.7. async_as_completed.py
  • 1.8. async_pipeline.py
  • 1.9. async_queue_workers.py
  • 1.10. Pool sizing

API Reference

• Errors, sessions, and result objects
  • Error handling
    • AgentAttributesEmptyError
    • AgentNotFoundError
    • AgentTaskAttributesEmptyError
    • AgentTaskNotFoundError
    • AgentTeamAttributesEmptyError
    • AgentTeamNotFoundError
    • AgentToolAttributesEmptyError
    • AgentToolNotFoundError
    • ConversationNotFoundError
    • DatabaseNotConnectedError
    • InvalidSQLError
    • ProfileAttributesEmptyError
    • ProfileExistsError
    • ProfileNotFoundError
    • SelectAIError
    • VectorIndexNotFoundError
  • Session objects
    • Session
    • AsyncSession
  • Agent history results
    • TeamHistoryEvent
    • TaskHistoryEvent
    • ToolHistoryEvent
  • Enumerations
    • Action
    • ToolType
    • NotificationType
    • VectorDistanceMetric
    • VectorDBProvider