Flagship (repository here) is YAOSFFS (Yet Another Open-Source Feature Flags Service). The twist - full gRPC communication loop by v1.0.
It is not a unique solution, but will be a great case study for myself and hopefully an interesting digest into your feed.
The Motivation:
As I recently shared with you - Feature Flags is not a universally adopted tool. My humble opinion, about heavy missing out without it in production, is nice to have. But it is not a reason to build it from scratch.
The real reason? Bloat. Most enterprise flag services require massive architectural buy-in. I wanted a zero-friction, lightweight utility that I can quickly spin up the moment I join a new team that is missing the tool.
This brings us to
The requirements:
Enclosed ecosystem which can be scaled all the way up to a SaaS solution if needed, but decoupled enough to be easily stripped down to core functionality and plugged into existing product As-Is without too much dancing around it.
The solution:
High throughput core service, written in Golang, supported with JWT Authority and Rate Limiter sidecars, which can be whatever, but as part of Flagship - implemented with Python/FastAPI (leveraging Python's velocity for operational tooling, and Go's raw concurrency for execution).
The Diagram:
[ PUBLIC TRAFFIC ]
│
▼
[ Edge Ingress / Nginx:80 ]
/ │ \
/api/v1/flags/ /api/v1/auth/ /api/v1/limiter/
/ │ \ (Restricted)
▼ ▼ ▼
[Flagship Core] [JWT Authority] [Rate Limiter]
(Go:8080) (Python:3000) (Python:8000)
│ │ │
│ └─────────(Intranet)──────────┤
▼ ▼
[PostgreSQL] ───────────────────────► [Redis]
(Write-Through) (Sliding Window)
Very simple and straightforward.
All but Flagship Core is optional.
The v0.1:
The current release is a full ecosystem, ready to be spun up with a single docker-compose command.
For this version:
- Full HTTP is acceptable for this release. But intranet between services is mandatory.
- State fully handled by Redis is acceptable for this release.
- JWT Authority can have a single high-entropy master cryptography key. It exists as a secret. On cloud, it lives in a single secret manager.
- Rate Limiter supports Token based rates.
- Rate Limiter bottleneck is known and is acceptable for this release.
The Roadmap:
| Phase | Milestone | Core Focus | Architecture Impact | Status |
|---|---|---|---|---|
| v0.1 | Container Monolith Base | Orchestrated REST Core | Edge Ingress routing, multi-language container isolation | Released |
| v0.2 | Source of Truth & Hydration | Reliable State Handling | Write-Through pattern to PostgreSQL; ultra-fast reads from Redis | In Progress |
| v0.3 | Sidecar Latency Optimization | gRPC Communication Layer | Migrate inter-service checks to gRPC to eliminate HTTP/1.1 HoL blocking | Planned |
| v0.4 | High-Perf Edge Layer | Global gRPC Ingress | Transition public edge routing to utilize Protocol Buffers & HTTP/2 streaming | Planned |
| v0.5 | Advanced Targeting Engine | Contextual Flag Evaluation | Abstract rule-matching evaluation engines beyond simple primitives | Planned |
| v0.6 | Distributed Event Bus | Real-Time Cache Sync | Integrate RabbitMQ/Kafka to synchronize cache states across cross-region nodes | Planned |
| v0.7 | Cloud-Native Scale | Infrastructure as Code | Multi-AZ AWS ECS Fargate deployment automated via Terraform | Planned |
I sure will have fun building it.
If you will have fun watching me do it - that’s up to you.
Cheers!