RightFlow Autopilot

• 

  rightflow-architecture

• 

  rightflow-devpost-thumbnail

Inspiration

People describe life situations, not policy rules. RightFlow helps turn incomplete social-rights stories into structured and reviewable cases.

What it does

RightFlow Autopilot uses Qwen Cloud to extract explicit facts, identify missing information, and coordinate bounded tools.

Before any eligibility assessment runs, a human must verify the extracted facts. Qwen understands and orchestrates; deterministic rules decide.

How we built it

We used Next.js, React, TypeScript, Qwen Cloud, and Alibaba Cloud.

The system includes strict output schemas, prompt-injection protection, timeouts, human approval gates, privacy-aware processing, and an auditable execution timeline.

The public demo runs on Alibaba Cloud Function Compute in Singapore at rightflow.sosyalhakrehberi.com. Cloudflare provides DNS and HTTPS termination at the edge; application execution and Qwen orchestration remain on Alibaba Cloud.

Challenges

The main challenge was enabling useful AI autonomy without allowing the model to make opaque, high-impact decisions.

Accomplishments

• Qwen-powered structured intake
• Human-in-the-loop approval
• Deterministic eligibility boundary
• Explainable audit timeline
• Safe fallback mode
• Alibaba Cloud deployment architecture
• 229 passing automated tests

What we learned

Production agents are more trustworthy when their autonomy is bounded, observable, and reversible.

What's next

We plan to add multilingual voice intake, verified policy sources, consent-based memory, caseworker queues, and more social-support programs.

Built With

• ai-agents
• alibaba-cloud
• docker
• next.js
• qwen-cloud
• react
• typescript