azpaddypy 0.9.9

Comprehensive Python logger for Azure, integrating OpenTelemetry for advanced, structured, and distributed tracing.

azpaddypy

Overview

azpaddypy provides robust, production-grade tools for managing Azure resources, with a focus on multi-tenant SaaS applications. The CosmosPromptManager enables efficient, tenant-aware prompt storage and retrieval using Azure Cosmos DB.

CosmosPromptManager: Multi-Tenant Prompt Management

Key Features

• Multi-Tenancy: All prompt operations are tenant-aware. Use tenant_id for isolation and performance.
• Partition Key Logic: Uses tenant_id as the Cosmos DB partition key when provided, falling back to prompt_name for global/shared prompts.
• Batch Operations: Batch methods (get_prompts_batch, save_prompts_batch, delete_prompts_batch) are implemented as sequential single operations. For large-scale batch, consider Cosmos DB stored procedures.
• Retry Logic: All operations use exponential backoff for resilience.
• Async Support: Async methods are available for high-throughput scenarios.
• Comprehensive Logging: Standardized logging and error handling throughout.

Best Practices

• Always provide tenant_id for tenant-specific operations in multi-tenant SaaS.
• Use global prompts (no tenant_id) only for shared defaults.
• For large-scale batch operations, use Cosmos DB stored procedures for efficiency.
• For performance, ensure your Cosmos DB container uses /tenant_id as the partition key.

Example Usage

from azpaddypy.tools.cosmos_prompt_manager import CosmosPromptManager

# Initialize manager (see full example for AzureCosmosDB setup)
prompt_manager = CosmosPromptManager(
    cosmos_client=cosmos_client,
    database_name="prompts",
    container_name="prompts"
)

# Save a tenant-specific prompt
prompt_manager.save_prompt(
    prompt_name="answering_user_prompt",
    prompt_data="Your prompt template here...",
    tenant_id="tenant_abc123"
)

# Retrieve a tenant-specific prompt
prompt = prompt_manager.get_prompt(
    prompt_name="answering_user_prompt",
    tenant_id="tenant_abc123"
)

# List all prompts for a tenant
prompt_names = prompt_manager.list_prompts(tenant_id="tenant_abc123")

# Batch save prompts (sequential, not true batch)
prompts_to_save = [
    {"prompt_name": "prompt1", "prompt_data": "template1"},
    {"prompt_name": "prompt2", "prompt_data": "template2"}
]
prompt_manager.save_prompts_batch(prompts_to_save, tenant_id="tenant_abc123")

# Batch delete prompts
prompt_manager.delete_prompts_batch(["prompt1", "prompt2"], tenant_id="tenant_abc123")

Partition Key Strategy

• For best performance and scalability, use /tenant_id as the partition key in your Cosmos DB container.
• This ensures tenant data is isolated and queries are efficient.

Batch Operation Caveats

• The *_batch methods perform sequential single operations, not true Cosmos DB batch.
• For high-volume batch, use Cosmos DB stored procedures or read_many_items.

Error Handling & Logging

• All operations use retry logic with exponential backoff.
• Errors are logged with context for troubleshooting.

Async Usage

• Async methods are available for high-throughput scenarios.
• See the code for details on async context management.

Configuration Management

• The configuration system is fully multi-tenant ready.
• All settings (checkboxes, dropdowns, etc.) in the admin UI are tenant-specific when a tenant is selected.
• Configurations are saved and loaded per tenant, with fallback to global defaults.

License

MIT