Loading…
Internal Canarlo platform · In active development
A documentation platform that reads code and emits prose a non-engineer can act on. Used internally on every Canarlo build.
Outcome
v0.4
internal release
Capabilities
The lede
Software development teams repeatedly hit the same bottleneck: vague project ideas turn into incomplete documentation, misaligned expectations, and costly delays. This internal platform transforms high-level project concepts into complete documentation suites.
Objectives
Approach
AI-first, systems-driven engineering — repeatability, correctness, long-term scalability. Rather than building a static form-based tool, we designed a conversational discovery system that actively identifies ambiguity, missing requirements, and contradictions before documentation is generated. Documentation comes from validated inputs, not assumptions.
→ Teams start projects with clear, validated scope, reduce miscommunication and rework, maintain consistent documentation quality across teams, and scale project intake without increasing overhead. It serves as a foundation for long-term scope management and iteration throughout the development lifecycle.
Engineering highlights
AI orchestration
Multi-model system separating real-time conversation from deep analysis and generation
Discovery system
Adaptive, multi-step conversational workflow capturing business + technical + operational requirements
Document generation
Parallel, dependency-aware generation of multiple specialist documents
Type safety
End-to-end TypeScript with schema-validated AI outputs
Security
Multi-tenant architecture with strict data isolation and server-side validation
Performance
Real-time AI streaming with sub-second response times
Outcome
v0.4
internal release
Reflection
AI orchestration, structured validation, and modern SaaS architecture can turn a fragile, manual process into a scalable system. By focusing on conversational discovery and correctness from day one, we built a platform that captures complex requirements through natural dialogue and produces production-ready documentation in minutes.
Further reading
SaaS Platform
A product-led business needed a SaaS that could support multiple frontends, partner integrations, and future expansion without repeatedly reworking core logic.
Read
Modernisation
A working platform that had outgrown its original architecture. Auth, data, and observability rebuilt without a service interruption — the customers never noticed.
Read
AI Automation
Categorisation, summarisation, and triage automated across an operations team's daily workload. Graceful degradation when the model is unavailable — never hard-fail.
Read