Repository navigation
[feat] Add configurable jitter strategy (full, equal, decorrelated, none) to backoff calculations #48
Description
Description:
Problem
Currently:
smooth-api-tshardcodes Equal Jitter (half + Math.random() * half).smooth-api-pyhardcodes Full Jitter (random.uniform(0, capped)).
Users cannot configure deterministic backoff for testing or choose their preferred jitter distribution algorithm.
Proposed Solution
- Add
jitter?: 'full' | 'equal' | 'decorrelated' | 'none'toBackoffConfigin both TypeScript and Python. - Implement standard AWS-style jitter algorithms:
-
none: Deterministic exponential ($base \times 2^{attempt}$ ) capped at$maxDelay$ . -
full: Random value between$0$ and$capped$ . -
equal:$half + \text{rand}(0, half)$ . -
decorrelated:$min(maxDelay, \text{rand}(baseDelay, prevDelay \times 3))$ .
-
- Default to
equalacross both packages for parity.
Acceptance Criteria
BackoffConfigupdated inpackages/smooth-api-ts/src/types.tsandpackages/smooth-api-py/smooth_api/config.py.calculateBackoff/calculate_backoffupdated in both packages.- Unit tests added verifying that calculated delays stay strictly within
[0, maxDelay]bounds for all 4 strategies.
References
-
TypeScript Implementation
Add configurable jitter strategies to backoff calculation in TS #53 -
Python Implementation
N\A
Activity
shreyansh0207 commentedon Sep 10, 2026
Hi @AryanSharma48, apologies — I forked the repo and implemented this directly instead of commenting first to ask. I've opened a PR with the TypeScript implementation: #53
Summary of what I did:
jitterfield toBackoffConfigin types.tsI focused on TypeScript only for now — happy to implement the Python side (smooth-api-py) as well if you'd like, just let me know! And let me know if you'd prefer I follow a different process next time.
AryanSharma48 commentedon Sep 10, 2026
@shreyansh0207 No worries! Just a good thing to get assigned first as it removes possibility of 2 people working on the same issue!
The implementation was good, ive left some comments regarding some nitpicks. Have a look at them and you can proceed with the python implementation as well!
[-]feat(core): Add configurable jitter strategy (`full`, `equal`, `decorrelated`, `none`) to backoff calculations[/-][+][feat] Add configurable jitter strategy (`full`, `equal`, `decorrelated`, `none`) to backoff calculations[/+]