bearwatch 0.1.2

BearWatch Python SDK - Job monitoring with heartbeat-based detection

bearwatch

Official BearWatch SDK for Python - Job monitoring and alerting for indie developers.

Table of Contents

• Installation
• Requirements
• Quick Start
• Usage
  • ping
  • wrap
• Async Support
  • Async Context Manager
  • ping_async
  • wrap_async
  • Async Error Handling
• Configuration
• Context Manager
• API Reference
  • BearWatch
  • Types
  • Errors
• Retry Policy
• Error Handling
• Common Patterns
• FAQ
• License

Installation

pip install bearwatch

Requirements

• Python 3.9 or higher
• httpx >= 0.25.0 (installed automatically)

Quick Start

1. Get API Key

Go to BearWatch Dashboard → Project Settings → Create API Key (e.g., bw_kI6t8QA21on0DKeRDlen8r2hzucVNL3WdAfaZgQdetY).

2. Create a Job

Create a job in the dashboard. You'll get a job ID (24-character hex string, e.g., 507f1f77bcf86cd799439011).

3. Install and Use

Let's assume you have a daily backup job that runs at 2:00 AM:

from apscheduler.schedulers.blocking import BlockingScheduler
from bearwatch import BearWatch

bw = BearWatch(api_key="your-api-key")

def backup_job():
    bw.wrap("507f1f77bcf86cd799439011", lambda: backup())

scheduler = BlockingScheduler()
scheduler.add_job(backup_job, "cron", hour=2)
scheduler.start()

Usage

ping - Manual Status Reporting

Use ping when you need fine-grained control over status reporting:

def backup_job():
    try:
        backup()
        bw.ping("507f1f77bcf86cd799439011", status="SUCCESS")
    except Exception as e:
        bw.ping("507f1f77bcf86cd799439011", status="FAILED", error=str(e))

Include output and metadata:

def backup_job():
    bytes_written = backup()
    bw.ping(
        "507f1f77bcf86cd799439011",
        status="SUCCESS",
        output=f"Backup completed: {bytes_written} bytes",
        metadata={
            "server": "backup-01",
            "region": "ap-northeast-2",
            "version": "1.2.0",
        },
    )

Return value:

response = bw.ping("507f1f77bcf86cd799439011", status="SUCCESS")
# response:
# {
#     'runId': '684a1b2c3d4e5f6789012345',
#     'jobId': '507f1f77bcf86cd799439011',
#     'status': 'SUCCESS',
#     'receivedAt': '2024-01-15T09:30:00.123Z'
# }

PingOptions

Option	Type	Default	Description
status	RequestStatus	"SUCCESS"	"RUNNING", "SUCCESS", or "FAILED"
output	str	-	Output message (max 10KB)
error	str	-	Error message for FAILED status (max 10KB)
started_at	datetime | str	current time	Job start time (ISO 8601 if string)
completed_at	datetime | str	current time	Job completion time (ISO 8601 if string)
metadata	dict[str, Any]	-	Additional key-value pairs (max 10KB)
retry	bool	True	Set False to disable retry for this call

Note: TIMEOUT and MISSED are server-detected states and cannot be set in requests.

Size Limit: output, error, and metadata fields have a 10KB size limit. If exceeded, the server automatically truncates the data (no error is returned). For output and error, the string is truncated. For metadata, the entire field is set to null if it exceeds the limit.

wrap - Automatic Status Reporting

Wraps a function and automatically:

• Measures started_at and completed_at
• Reports SUCCESS or FAILED based on whether the function completes or throws

# Pass function directly (no arguments)
bw.wrap("507f1f77bcf86cd799439011", backup)

# Use lambda for functions with arguments
bw.wrap("507f1f77bcf86cd799439011", lambda: backup(path="/data"))

Include output and metadata:

def backup_job():
    bw.wrap(
        "507f1f77bcf86cd799439011",
        lambda: backup(),
        output="Daily backup completed",
        metadata={
            "server": "backup-01",
            "region": "ap-northeast-2",
        },
    )

Return value (returns the wrapped function's result):

result = bw.wrap("507f1f77bcf86cd799439011", lambda: "done")
# result: 'done'

count = bw.wrap("507f1f77bcf86cd799439011", lambda: len(records))
# count: 42

WrapOptions

Option	Type	Default	Description
output	str	-	Output message (max 10KB)
metadata	dict[str, Any]	-	Additional key-value pairs (max 10KB)
retry	bool	True	Set False to disable retry for this call

Size Limit: output and metadata fields have a 10KB size limit. If exceeded, the server automatically truncates the data (no error is returned). For output, the string is truncated. For metadata, the entire field is set to null if it exceeds the limit.

Error handling behavior:

• On success: reports SUCCESS with execution duration
• On error: reports FAILED with error message, then re-raises the original exception

def backup_job():
    try:
        bw.wrap("507f1f77bcf86cd799439011", lambda: backup())
    except Exception as e:
        # BearWatch already reported FAILED status
        # You can add additional error handling here
        logger.error(e)

Tip: Use wrap for most cases. Use ping when you need more control (e.g., reporting RUNNING status for long jobs).

Async Support

The SDK provides async versions of all methods for use with asyncio, FastAPI, and other async frameworks.

Async Context Manager

Use async with for automatic resource cleanup:

async with BearWatch(api_key="your-api-key") as bw:
    await bw.ping_async("507f1f77bcf86cd799439011")
    # Resources automatically cleaned up on exit

Without context manager (remember to close):

bw = BearWatch(api_key="your-api-key")
try:
    await bw.ping_async("507f1f77bcf86cd799439011")
finally:
    await bw.aclose()

ping_async

async with BearWatch(api_key="your-api-key") as bw:
    # Simple success
    await bw.ping_async("507f1f77bcf86cd799439011")

    # With options
    await bw.ping_async(
        "507f1f77bcf86cd799439011",
        status="FAILED",
        error="Connection timeout",
        metadata={"attempt": 3},
    )

wrap_async

Wraps an async function and automatically:

• Measures started_at and completed_at
• Reports SUCCESS or FAILED based on whether the function completes or raises

async def fetch_data():
    async with aiohttp.ClientSession() as session:
        async with session.get("https://api.example.com/data") as resp:
            return await resp.json()

async with BearWatch(api_key="your-api-key") as bw:
    # Basic usage
    result = await bw.wrap_async("507f1f77bcf86cd799439011", fetch_data)

    # With output and metadata
    result = await bw.wrap_async(
        "507f1f77bcf86cd799439011",
        fetch_data,
        output="Fetched 1000 records",
        metadata={"source": "api.example.com"},
    )

Async Error Handling

For ping_async: raises BearWatchError on failure.

from bearwatch import BearWatch, BearWatchError

async with BearWatch(api_key="your-api-key") as bw:
    try:
        await bw.ping_async("507f1f77bcf86cd799439011")
    except BearWatchError as e:
        print(f"Failed to report: {e.code}")

For wrap_async: reports failure to BearWatch, then re-raises the original exception.

async with BearWatch(api_key="your-api-key") as bw:
    try:
        await bw.wrap_async("507f1f77bcf86cd799439011", async_backup)
    except BackupError as e:
        # BearWatch already reported FAILED status
        # Handle your application error here
        logger.error(f"Backup failed: {e}")

Note: If wrap_async fails to report to BearWatch (network error, etc.), the original exception is still raised. The SDK silently ignores reporting errors to preserve the original exception.

Configuration

bw = BearWatch(
    api_key="your-api-key",

    # Optional (defaults shown)
    timeout=30.0,       # 30 seconds
    max_retries=3,
    retry_delay=0.5,    # 500ms base delay
)

Option	Type	Required	Default	Description
api_key	str	Yes	-	API key for authentication
timeout	float	No	30.0	Request timeout (seconds)
max_retries	int	No	3	Max retry attempts
retry_delay	float	No	0.5	Initial retry delay (seconds)

Context Manager

Use context managers for automatic resource cleanup:

# Sync
with BearWatch(api_key="your-api-key") as bw:
    bw.ping("507f1f77bcf86cd799439011")

# Async
async with BearWatch(api_key="your-api-key") as bw:
    await bw.ping_async("507f1f77bcf86cd799439011")

API Reference

BearWatch

class BearWatch:
    def __init__(
        self,
        api_key: str,
        *,
        timeout: float = 30.0,
        max_retries: int = 3,
        retry_delay: float = 0.5,
    ) -> None: ...

    @classmethod
    def create(cls, config: BearWatchConfig) -> BearWatch: ...

    def ping(
        self,
        job_id: str,
        *,
        status: RequestStatus = "SUCCESS",
        output: str | None = None,
        error: str | None = None,
        started_at: datetime | str | None = None,
        completed_at: datetime | str | None = None,
        metadata: dict[str, Any] | None = None,
        retry: bool = True,
    ) -> HeartbeatResponse: ...

    def wrap(
        self,
        job_id: str,
        fn: Callable[[], T],
        *,
        output: str | None = None,
        metadata: dict[str, Any] | None = None,
        retry: bool = True,
    ) -> T: ...

    async def ping_async(
        self,
        job_id: str,
        *,
        status: RequestStatus = "SUCCESS",
        output: str | None = None,
        error: str | None = None,
        started_at: datetime | str | None = None,
        completed_at: datetime | str | None = None,
        metadata: dict[str, Any] | None = None,
        retry: bool = True,
    ) -> HeartbeatResponse: ...

    async def wrap_async(
        self,
        job_id: str,
        fn: Callable[[], Awaitable[T]],
        *,
        output: str | None = None,
        metadata: dict[str, Any] | None = None,
        retry: bool = True,
    ) -> T: ...

    def close(self) -> None: ...
    async def aclose(self) -> None: ...

Types

Status Types

# Status that SDK can send to the server
RequestStatus = Literal["RUNNING", "SUCCESS", "FAILED"]

# Status that server can return (includes server-detected states)
ResponseStatus = Literal["RUNNING", "SUCCESS", "FAILED", "TIMEOUT", "MISSED"]

HeartbeatResponse

Return value of ping() and wrap() methods:

class HeartbeatResponse(TypedDict):
    runId: str           # Generated run ID
    jobId: str           # Job ID
    status: ResponseStatus
    receivedAt: str      # Server received time (ISO 8601)

BearWatchConfig

Configuration for BearWatch.create() method:

@dataclass
class BearWatchConfig:
    api_key: str              # Required - API authentication key
    timeout: float = 30.0     # Request timeout in seconds
    max_retries: int = 3      # Maximum retry attempts
    retry_delay: float = 0.5  # Retry interval in seconds (exponential backoff)

Errors

BearWatchError

class BearWatchError(Exception):
    code: ErrorCode           # Error code
    status_code: int | None   # HTTP status code
    context: ErrorContext | None
    response_body: str | None # Response body for debugging

ErrorCode

ErrorCode = Literal[
    "INVALID_API_KEY",   # 401 - Invalid API key
    "JOB_NOT_FOUND",     # 404 - Job not found
    "RATE_LIMITED",      # 429 - Rate limit exceeded
    "SERVER_ERROR",      # 5xx - Server error
    "INVALID_RESPONSE",  # Unexpected response format
    "NETWORK_ERROR",     # Network failure
    "TIMEOUT",           # Request timeout
]

ErrorContext

@dataclass
class ErrorContext:
    job_id: str | None = None
    run_id: str | None = None
    operation: str | None = None  # "ping", "wrap", etc.

Imports

from bearwatch import (
    BearWatch,
    BearWatchConfig,
    BearWatchError,
    ErrorCode,
    ErrorContext,
    HeartbeatResponse,
    PingOptions,
    WrapOptions,
    RequestStatus,
    ResponseStatus,
)

Retry Policy

Method	Default Retry	Reason
ping()	Enabled	Idempotent operation
ping_async()	Enabled	Idempotent operation
wrap()	Enabled	Uses ping() internally
wrap_async()	Enabled	Uses ping_async() internally

Retry Behavior

• Exponential backoff: 500ms → 1000ms → 2000ms
• 429 Rate Limit: Respects Retry-After header (rate limit: 100 requests/minute per API key)
• 5xx Server Errors: Retries with backoff
• 401/404: No retry (client errors)

Disable Retry

Use retry=False to skip retries for a specific call, regardless of the max_retries setting:

bw.ping("507f1f77bcf86cd799439011", retry=False)

Error Handling

When the SDK fails to communicate with BearWatch (network failure, server down, invalid API key, etc.), it raises a BearWatchError:

from bearwatch import BearWatch, BearWatchError

try:
    bw.ping("507f1f77bcf86cd799439011")
except BearWatchError as e:
    # SDK failed to report to BearWatch
    print(f"Code: {e.code}")
    print(f"Status: {e.status_code}")
    print(f"Context: {e.context}")

Error Codes

Code	Description	Retry
INVALID_API_KEY	401 - Invalid API key	No
JOB_NOT_FOUND	404 - Job not found	No
RATE_LIMITED	429 - Rate limit reached	Yes
SERVER_ERROR	5xx - Server error	Yes
INVALID_RESPONSE	Unexpected response format	No
NETWORK_ERROR	Network failure	Yes
TIMEOUT	Request timed out	Yes

Common Patterns

APScheduler

from apscheduler.schedulers.blocking import BlockingScheduler
from bearwatch import BearWatch

bw = BearWatch(api_key="your-api-key")

def backup_job():
    bw.wrap("6848c9e5f8a2b3d4e5f60001", lambda: backup())

scheduler = BlockingScheduler()
scheduler.add_job(backup_job, "cron", hour=3)
scheduler.start()

Celery Beat

from celery import Celery
from bearwatch import BearWatch

app = Celery("tasks")
bw = BearWatch(api_key="your-api-key")

@app.task
def backup_task():
    bw.wrap("6848c9e5f8a2b3d4e5f60002", lambda: backup())

AWS Lambda (EventBridge Scheduler)

import os
from bearwatch import BearWatch

bw = BearWatch(api_key=os.environ["BEARWATCH_API_KEY"])

def handler(event, context):
    bw.wrap("6848c9e5f8a2b3d4e5f60003", lambda: backup())

Long-Running Jobs

from datetime import datetime, timezone

def run_backup():
    job_id = "6848c9e5f8a2b3d4e5f60004"
    started_at = datetime.now(timezone.utc)

    bw.ping(job_id, status="RUNNING")

    try:
        backup()
        bw.ping(
            job_id,
            status="SUCCESS",
            started_at=started_at,
            completed_at=datetime.now(timezone.utc),
        )
    except Exception as e:
        bw.ping(
            job_id,
            status="FAILED",
            started_at=started_at,
            completed_at=datetime.now(timezone.utc),
            error=str(e),
        )
        raise

FAQ

Q: Do I need to create jobs in the dashboard first? A: Yes, create a job in the BearWatch Dashboard first to get a job ID.

Q: What's the difference between wrap and ping? A: wrap automatically measures execution time and reports SUCCESS/FAILED based on whether the function completes or raises an exception. ping gives you manual control over when and what to report.

Q: What happens if the SDK fails to report (network error)? A: By default, the SDK attempts up to 4 times total (1 initial + 3 retries) with exponential backoff. If all attempts fail, ping raises a BearWatchError. For wrap, the original function's exception takes priority and is always re-raised.

Q: Can I use this with async frameworks like FastAPI? A: Yes, use ping_async and wrap_async for async contexts.

License

MIT