Skip to main content
5 min

9 OSS Alternatives Evaluated, 3 Adopted

We evaluated 9 open-source replacements for our SaaS stack. Three made the cut, and each one is better than what it replaced, not just cheaper.

OSSSaaSInfrastructureCost Optimization

Process note: Human-directed evaluation and writing, AI-assisted editing.

Our SaaS bill was creeping up. Not dramatically, maybe $200 a month, but for tools we use three times a week the per-use cost was absurd. So we evaluated 9 open-source replacements across three categories. Three made the cut.

What we evaluated

| Tool | SaaS alternative | OSS replacement | Verdict | |------|------------------|------------------|---------| | Image generation | Midjourney ($10/mo) | FLUX via fal.ai | Adopted | | Document parsing | Notion API ($8/mo) | Local markdown parser | Adopted | | Vector search | Pinecone ($50/mo) | ChromaDB (local) | Adopted | | Code review | GitHub Teams ($21/mo) | GitLab CE (self-hosted) | Too much ops | | CI/CD | GitHub Actions (overage) | Woodpecker CI | Not ready | | Monitoring | Datadog ($15/mo) | Prometheus + Grafana | Needs more glue | | Design systems | Figma ($15/mo) | Excalidraw + tldraw | Missing features | | Email | SendGrid ($29/mo) | Local SMTP relay | Deliverability risk | | Testing | BrowserStack ($99/mo) | Playwright (local) | Not a fair comparison |

The three that won

1. FLUX via fal.ai

Midjourney is good. FLUX is better for anything with text, logos or UI. The fal.ai endpoint costs per use, so we only pay when we generate. We replaced a $10 a month subscription with a pay-per-image model that is cheaper in practice.

2. Local markdown parser

We were paying Notion API rates to extract content from markdown files. The local parser does the same job in about 200 lines of Python. No API calls, no rate limits, no bill.

3. ChromaDB

Pinecone was eating $50 a month for a vector store we barely queried. ChromaDB runs locally, integrates with our existing embeddings pipeline, and has zero per-query cost. The tradeoff is that we manage the persistence layer ourselves, which is worth it at our size.

What we learned

  • Not every SaaS has an OSS equivalent. CI/CD and monitoring are still hard self-hosted.
  • The hidden cost is ops. Woodpecker CI would save money on paper, but the maintenance burden is not worth it at our volume.
  • Adopt when the replacement is better, not just cheaper. FLUX beats Midjourney for our use case. The local parser is faster than the API. ChromaDB is more predictable.

The short version

Three tools replaced, about $73 a month saved, and nothing we cared about was lost in the swap.

Related Posts