Inspiration
One film. One book. A reason to come together.
An independent cinema and a bookshop may share a neighborhood but still struggle to plan an evening together. A thematic connection is only the beginning: someone must check budget, screening time, book stock, availability, and who will host the discussion.
What it does
Common Bill turns that coordination problem into a shared planning desk. It filters infeasible film–book combinations, uses Qloo evidence to ground cultural selection, and prepares a running order, host brief, and announcement draft. Reject a pairing or lose the bookseller, and the planning workflow revises the proposal or explains why no feasible plan remains. Before/after differences make the revision visible.
The redesigned interface uses a film-and-book programme identity, a simple three-step introduction, clearer language, and inspectable evidence. Both partners can record local review notes and export the plan. These notes are not authenticated approvals; nothing is automatically booked, purchased, or promoted.
Try the public live demo without signing in or supplying a key. It uses real Qloo and Groq calls. Businesses, costs, stock, permissions and staffing are explicitly fictional demonstration assumptions. Results may be reused for one hour with original timestamps.
Why Qloo matters
An LLM can invent a persuasive connection between almost any film and book. Common Bill asks for evidence alongside that interpretation. It resolves exact works, constrains Qloo Insights to the supplied catalog, and checks film-to-book and book-to-film affinity separately. A Qloo-mode proposal needs both directions; missing evidence stays unknown.
The receipts retain entity IDs, query parameters, vendor scores, candidate ranks and timestamps. Affinity and relative ranks are not attendance predictions or measured audience overlap. Removing Qloo removes the cultural evidence gate and provenance, while the model-only baseline can still make an editorial proposal.
Agent workflow and technical implementation
The planning workflow combines model decisions with deterministic safeguards. The model chooses a bounded evaluation scope and an editorial selection; application code checks operational feasibility and rejects choices outside the allowed set. Feedback triggers another constrained planning run.
TypeScript, React, Next.js, Tailwind CSS, Radix UI and Framer Motion power the workbench. Server-side Qloo adapters handle search and Insights. Groq GPT-OSS 120B returns structured decisions. The public deployment uses Sites, Vinext and Cloudflare Workers. Source, setup instructions, tests and an MIT license are on GitHub. Secrets remain server-side.
Test the difference
The Compare results workflow uses matching model IDs and operational inputs, randomizes proposal order and hides methods until a reviewer locks a preference. Reviewers can record their role, explanation and willingness to try the plan. These are local self-reports, not verified operator endorsements or proof that Qloo always wins.
Companion: Common Ground
The optional /common-ground page applies taste-grounded negotiation to a shared movie night. A bounded tool-using agent chooses whether to query Qloo, ask a participant to clarify a favorite, or propose an eligible film. It shows recorded actions and receipts, preserves vetoes, and requires explicit acceptance from everyone. Its catalog is deliberately limited to six taste starters and four candidate films. Common Bill remains the primary project demonstrated here.
Challenges and what we learned
Exact identity matters when titles have multiple editions. Affinity is directional. A single returned result provides no comparative ranking advantage. Rate limits require bounded requests, retries and transparent reuse. Above all, an appealing cultural idea must still fit practical requirements.
What we verified
On October 30, 2026, all 36 automated tests, TypeScript checking and the hosted production build passed. Mobile layout was inspected; complete desktop interaction testing remains outstanding. Dated integration checks cover planning, staffing repair, rejected-pair exclusion, infeasible budgets and the model-only baseline. A real Common Ground request also returned Qloo receipts and an agent clarification. Engineering checks are not customer trials.
Potential impact and next steps
The audience is independent cinema programmers and booksellers coordinating joint events. Our hypothesis is that a shared evidence-backed draft reduces coordination effort and catches practical disagreements early. We have not measured time savings, attendance, revenue or adoption and claim no organizer endorsements.
The next step is a consenting operator pilot with verified catalogs, task completion, time-to-agreed-draft and blind proposal preference. Authenticated partner workspaces and durable review history would follow evidence of usefulness.
Try it
- Click Try the live demo.
- Open Inspect the evidence and check both Qloo directions.
- Try Bookseller unavailable, then inspect the revision or blockers.
- Try Budget drops to £100.
- Reset and open Compare results.
- Export the event brief and decision record.
Live app: https://common-bill.ananyadn678.chatgpt.site/ Public source: https://github.com/Ananya6Daitkar/common-bill
Built With
- qloo
- typescript
