Overview
This is the first resume-ready release of Serverless Cloud Platform, a local-first, Lambda-inspired function execution platform built with FastAPI, PostgreSQL, Redis Streams, Docker, and React.
The release focuses on reliable asynchronous invocation, isolated function execution, failure recovery, authentication, observability, and reproducible performance evidence.
Highlights
- Immutable Python 3.11 function versions with package hashing and configurable CPU, memory, and timeout limits.
- Transactional PostgreSQL outbox that prevents accepted invocations from being lost before Redis publication.
- Redis Streams delivery with consumer-aware pending-message recovery and at-least-once execution semantics.
- Race-safe idempotency, invocation-attempt uniqueness, and duplicate-delivery suppression.
- Bounded Worker concurrency with delayed exponential-backoff retries and total invocation deadlines.
- Docker-based function isolation with explicit success, handler-error, timeout, memory-limit, and invalid-output handling.
- Durable invocation results, errors, and user-function logs.
- bcrypt password hashing, JWT authentication, owner isolation, and atomic Redis token-bucket rate limiting.
- Operational metrics for queue depth and age, retries, throughput, error rate, workers, and p50/p95/p99 latency.
- Authenticated React Dashboard for function deployment, invocation, result inspection, logs, and metrics.
- Reproducible Docker Compose startup with health checks and automatic Alembic migrations.
Validation
GitHub Actions CI run #36 passed all four jobs:
testqualityfrontenddocker-smoke
Local and real-service validation also completed successfully:
- 115 default tests passed, with 10 opt-in tests skipped
- 2 real PostgreSQL/Redis integration tests passed
- 1 real Docker runtime test passed
- 7 complete Docker Compose E2E workflows passed
- Ruff, Mypy, compileall, frontend production build, migrations, and Compose validation passed
Release Benchmark
The audited release benchmark ran from clean commit eb421c01eaaf110e7d24f7690284e1556296a7ca against fresh PostgreSQL, Redis, storage, and Docker Compose state.
It completed nine independent runs containing 750 total cold-container invocations with a 100% success rate.
| Workload | Runs | Invocations per run | Clients | Median throughput | Median p95 latency |
|---|---|---|---|---|---|
| No-op | 3 | 100 | 10 | 2.44 invocations/s | 6617.83 ms |
| Sleep 200 ms | 3 | 100 | 10 | 1.96 invocations/s | 7407.67 ms |
CPU-bound (n=250000) |
3 | 50 | 5 | 1.53 invocations/s | 4583.24 ms |
Benchmark environment:
- Apple M5
- 10 logical CPUs
- 24 GiB RAM
- Docker 29.6.1
- One active Worker
- Total Worker concurrency: 2
- One new Docker container per invocation
The full generated report and all raw evidence are versioned in the repository:
Known Limitations
- The platform is designed for local development and systems demonstrations, not production-grade multi-tenant isolation.
- Every measured invocation uses a cold Docker container; warm-container pooling is not implemented.
- Autoscaling and Kubernetes scheduling are outside the scope of this release.
- The browser stores the access token in session storage for the local MVP.
- Published benchmark results describe one specific local machine and Worker topology and should not be interpreted as general production capacity.