fix #42
Overview
This PR adds a wait_for_database() helper function to simplify polling for Knowledge Base database status during creation. Previously, users had to manually write polling loops to wait for the database to become ready, which was error-prone and required handling multiple edge cases.

Problem
When creating a Knowledge Base, the database deployment can take several minutes. Users previously had to write manual polling loops like:

from gradient import Gradient
import time

client = Gradient()

while True:
knowledge_base = client.knowledge_bases.retrieve("uuid")
if knowledge_base.database_status == "ONLINE":
break
# Manual error handling, timeout tracking, etc.
time.sleep(5)
This approach has several issues:

No timeout handling
No proper error handling for failed states
Verbose and repetitive code
Easy to miss edge cases
Solution
Added a wait_for_database() helper method that:

Automatically polls the database status until it becomes ONLINE
Handles failed states (DECOMMISSIONED, UNHEALTHY) with a custom KnowledgeBaseDatabaseError
Supports configurable timeout (default: 600 seconds) and poll interval (default: 5 seconds)
Raises APITimeoutError if the database doesn't become ready within the timeout
Works for both synchronous and asynchronous clients
Usage
from gradient import Gradient
from gradient.resources.knowledge_bases import KnowledgeBaseDatabaseError
from gradient._exceptions import APITimeoutError

client = Gradient()

Create a knowledge base

kb = client.knowledge_bases.create(
name="My Knowledge Base",
region="nyc1",
embedding_model_uuid="your-embedding-model-uuid",
)

kb_uuid = kb.knowledge_base.uuid

try:
# Wait for the database to be ready (default: 10 minute timeout, 5 second polling)
result = client.knowledge_bases.wait_for_database(kb_uuid)
print(f"Database is ready: {result.database_status}")

# Or with custom parameters
result = client.knowledge_bases.wait_for_database(
 kb_uuid,
 timeout=900.0, # 15 minutes
 poll_interval=10.0 # Check every 10 seconds
)

except KnowledgeBaseDatabaseError as e:
# Database entered a failed state (DECOMMISSIONED or UNHEALTHY)
print(f"Database failed: {e}")

except APITimeoutError:
# Database did not become ready within the timeout period
print("Timeout: Database did not become ready in time")
Async Usage
from gradient import AsyncGradient

async_client = AsyncGradient()

result = await async_client.knowledge_bases.wait_for_database(kb_uuid)
Changes
New method: KnowledgeBasesResource.wait_for_database() (synchronous)
New method: AsyncKnowledgeBasesResource.wait_for_database() (asynchronous)
New exception: KnowledgeBaseDatabaseError for failed database states
Tests: 20 comprehensive unit tests covering success scenarios, failed states, timeouts, and parameter validation
Documentation: Updated README with usage examples and error handling patterns
API Reference
def wait_for_database(
uuid: str,
*,
timeout: float = 600.0,
poll_interval: float = 5.0,
extra_headers: Headers | None = None,
extra_query: Query | None = None,
extra_body: Body | None = None,
) -> KnowledgeBaseRetrieveResponse
Parameters:

uuid - Knowledge base UUID to poll (required)
timeout - Maximum wait time in seconds (default: 600)
poll_interval - Time between status checks in seconds (default: 5)
Returns: KnowledgeBaseRetrieveResponse when database status is ONLINE

Raises:

KnowledgeBaseDatabaseError - Database entered a failed state (DECOMMISSIONED or UNHEALTHY)
APITimeoutError - Timeout exceeded before database became ONLINE
ValueError - Invalid UUID parameter
Testing
All tests pass successfully:

pytest tests/api_resources/test_knowledge_bases.py -k "wait_for_database" -v

20 passed (10 sync + 10 async tests)

Tests cover:

Successful database creation and polling
Failed database states (UNHEALTHY, DECOMMISSIONED)
Timeout scenarios
Parameter validation
Both loose and strict client modes
Backward Compatibility