1. The Core Bottleneck
Commercial personal finance software forces users into an uncomfortable compromise: surrendering private banking credentials and net worth ledgers to proprietary SaaS silos. When Maybe Finance burned nearly one million dollars between 2021 and 2022 before abandoning development, the codebase was left stranded. Sure emerged as a community-maintained fork that rescues this heavy investment. By stripping away the unsustainable B2C subscription wrapper, the project opens up a production-ready balance sheet engine, cash flow visualizer, and AI interaction layer for self-hosting enthusiasts.
💡 Core Architectural Insight: Sure proves that capital-abandoned software can achieve a second life through open-source self-hosting, eliminating privacy compromises by returning data ownership directly to the engineer.
2. Architecture and Data Flow
Sure relies on a monolithic Ruby on Rails backend coupled with dynamic frontend rendering, offloading high-latency financial institution scraping and AI agent logic to Sidekiq asynchronous workers. PostgreSQL acts as the single source of truth, enforcing ACID compliance across all financial transactions.
[ Browser / Mobile / CLI ] ---> [ Rails Router / Controller ] ---> [ ActiveRecord & PostgreSQL ]
|
v
[ Sidekiq Async Workers ]
|
v
[ LLM Gateway / API ]
Incoming requests hit the Rails routing layer and controllers, interacting directly with ActiveRecord models. Heavy input/output operations, such as bank synchronization and LLM inference, are dispatched to Sidekiq background workers to maintain low latency for interactive web clients.
3. Technical Trade-offs and Benchmark
| Dimension | Sure (this project) | Traditional Stack | Commercial SaaS | Production Benefit |
|---|---|---|---|---|
| Core Framework | Ruby on Rails | Node.js / Express | Proprietary | High domain expressiveness, 50% fewer boilerplate files |
| Database | PostgreSQL | MongoDB | Managed SQL | Strict ACID guarantees for ledger accuracy and complex queries |
| Deployment | Docker / Render | Custom VMs | Closed Cloud | Containerized delivery with absolute data sovereignty |
| Async Jobs | Redis + Sidekiq | Celery / RabbitMQ | Internal Cron | Robust concurrency scheduling for slow financial syncs |
| AI Integration | pgvector + LLM | None / Hardcoded | Cloud-locked API | Seamless switching between local Ollama and cloud LLMs |
The maturity of the Ruby on Rails ecosystem gives Sure an advantage in domain-driven data modeling. Convention over configuration minimizes code rot, keeping validation logic for complex financial instruments tightly encapsulated.
4. Local Development Setup
Running Sure locally requires Ruby, PostgreSQL, and Redis installed on the host machine. Execute the following sequence to bootstrap the environment:
# Clone the community-maintained repository
git clone https://github.com/we-promise/sure.git
# Move into the working directory
cd sure
# Duplicate the sample environment configuration
cp .env.local.example .env.local
# Install gem dependencies and run database migrations
bin/setup
# Boot the Rails server and asset compilation watchers
bin/dev
# Seed optional demo data for immediate UI testing
rake demo_data:default
Navigate to http://localhost:3000 in your browser. Log in using [email protected] with the password Password1! to inspect the default cash flow and net worth dashboards.
5. Production Gotchas and Mitigations
Data-heavy applications introduce predictable scaling friction. The project maintainers maintain an open Skylight monitoring dashboard to expose latency bottlenecks, but self-hosters must plan infrastructure capacity carefully.
⚠️ Gotcha: Database Connection Pool Exhaustion: During bulk financial API synchronization or heavy background processing, Puma threads can easily saturate the default PostgreSQL connection limit. Tune the
poolparameter insidedatabase.ymland deploy PgBouncer for connection multiplexing.⚠️ Gotcha: Background Job Queue Starvation: Default Sidekiq queues can block when processing heavy historical CSV imports or recursive AI agent parsing tasks. Isolate financial syncing, vector embedding generation, and web requests into dedicated Sidekiq queues in production environments.
