clari-rate-limits
Handle Clari API rate limits with backoff and export job scheduling. Use when hitting 429 errors, optimizing export frequency, or scheduling bulk forecast exports. Trigger with phrases like "clari rate limit", "clari 429", "clari throttle", "clari api limits".
Allowed Tools
Provided by Plugin
clari-pack
Claude Code skill pack for Clari (18 skills)
Installation
This skill is included in the clari-pack plugin:
/plugin install clari-pack@claude-code-plugins-plus
Click to copy
Instructions
Clari Rate Limits
Overview
The Clari API enforces rate limits per API key. Export jobs are asynchronous and queued server-side, so the primary concern is polling frequency and concurrent export requests.
Prerequisites
- A scoped API token and approved forecast export scope
- Persistent job/attempt tracking for retry and idempotency decisions
- A scheduler that can defer work without spawning duplicate workers
- Monitoring for queue depth, response class, and terminal failures
Rate Limit Behavior
| Aspect | Value |
|---|---|
| Scope | Per API key |
| Response on limit | HTTP 429 |
| Export job queue | Server-managed, async |
| Recommended polling | 5-10 second intervals |
Instructions
Exponential Backoff for Export Polling
import time
import requests
def poll_with_backoff(
job_id: str,
api_key: str,
max_attempts: int = 60,
base_delay: float = 5.0,
max_delay: float = 60.0,
) -> dict:
for attempt in range(max_attempts):
resp = requests.get(
f"https://api.clari.com/v4/export/jobs/{job_id}",
headers={"apikey": api_key},
)
if resp.status_code == 429:
retry_after = int(resp.headers.get("Retry-After", base_delay))
time.sleep(retry_after)
continue
resp.raise_for_status()
status = resp.json()
if status["status"] in ("COMPLETED", "FAILED"):
return status
delay = min(base_delay * (1.5 ** attempt), max_delay)
time.sleep(delay)
raise TimeoutError(f"Job {job_id} did not complete in {max_attempts} attempts")
Sequential Export Scheduler
def export_all_periods(
client,
forecast_name: str,
periods: list[str],
delay_between: float = 10.0,
) -> list[dict]:
results = []
for period in periods:
print(f"Exporting {period}...")
job = client.export_forecast(forecast_name, period)
result = poll_with_backoff(job["jobId"], client.config.api_key)
results.append(result)
time.sleep(delay_between) # Avoid hitting rate limits
return results
Error Handling
| Scenario | Detection | Response |
|---|---|---|
| 429 with Retry-After | Check header | Wait exact duration |
| 429 without header | Status code only | Backoff from 5s |
| Job queue full | Multiple pending jobs | Wait for completion before new exports |
Output
Return a per-job state record with export period, provider job ID, attempts, applied delay, response class, and terminal decision. Do not report tokens or raw forecast records; callers must receive a bounded retryable failure rather than silently starting a parallel export.
Examples
Submit one export for a staging period and persist its job ID before polling. On a 429 with Retry-After, reschedule the same job after that interval; if the attempt budget expires, mark the run unavailable and alert the scheduler instead of issuing another export request.
Resources
Next Steps
For security configuration, see clari-security-basics.