Backend Finalization & Features Database and API Work
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted1 hour ago
I have a pre-launch adult fiction web app which writes adult type stories. Nothing is live yet. No users, no revenue, no traffic. I am looking for one engineer to take it from where it is to a working production launch.
The stack: React 18, TypeScript, Vite and Tailwind on the front end (originally built in Bolt). Supabase on the back end, meaning Postgres, auth, private storage, Edge Functions, and RLS on every table. The app takes a user prompt and generates a text story, plus audio narration.
I have a complete engineering spec in a single file, written to be built from directly. It covers the schema, the phase by phase build order, API contracts, error handling, rate limits, and what done looks like for each phase. I have already audited the repo against that spec, so the gaps are known and written down.
The work, in order:
Schema reconciliation. Stories currently store all episodes in one jsonb column. The spec calls for a normalized model with one row per version and one active version per episode slot, so continuations and revisions behave correctly. This comes first because everything else sits on top of it.
Generation pipeline. All text and audio generation is async and job based, through a single polling endpoint. Text generation runs on a specific open model served through Featherless (flat monthly rate, OpenAI compatible API). Audio narration runs through Fish Audio's TTS API, billed per character. Both API keys and endpoint details are provided. Neither is wired up yet.
Subscriptions and entitlements. Tiered allotments, atomic credit spend, trial handling, a shared add on pool, and entitlement versioning so a future pricing change does not retroactively change existing subscribers. Parts of this exist and need to be finished to spec.
Launch security. RLS verified end to end, rate limiting, account deletion, moderation logging, and an anonymous first story flow that cannot leak data across users.
What I need from you:
Real Postgres depth. Migrations, RLS policies, and RPCs you can write and defend, not just familiarity with the Supabase dashboard.
TypeScript that holds up. Typed API responses, consistent error handling, no any scattered through the services.
Comfort integrating third party LLM and TTS APIs, including async job handling and polling, not just calling OpenAI directly.
Willingness to push back when the spec is wrong. It has been through two rounds of outside review and got better both times.
Comfort working on an 18+ platform. The content is explicit adult fiction, not extreme, and there is a content classifier with hard blocks built into the design.
The stack: React 18, TypeScript, Vite and Tailwind on the front end (originally built in Bolt). Supabase on the back end, meaning Postgres, auth, private storage, Edge Functions, and RLS on every table. The app takes a user prompt and generates a text story, plus audio narration.
I have a complete engineering spec in a single file, written to be built from directly. It covers the schema, the phase by phase build order, API contracts, error handling, rate limits, and what done looks like for each phase. I have already audited the repo against that spec, so the gaps are known and written down.
The work, in order:
Schema reconciliation. Stories currently store all episodes in one jsonb column. The spec calls for a normalized model with one row per version and one active version per episode slot, so continuations and revisions behave correctly. This comes first because everything else sits on top of it.
Generation pipeline. All text and audio generation is async and job based, through a single polling endpoint. Text generation runs on a specific open model served through Featherless (flat monthly rate, OpenAI compatible API). Audio narration runs through Fish Audio's TTS API, billed per character. Both API keys and endpoint details are provided. Neither is wired up yet.
Subscriptions and entitlements. Tiered allotments, atomic credit spend, trial handling, a shared add on pool, and entitlement versioning so a future pricing change does not retroactively change existing subscribers. Parts of this exist and need to be finished to spec.
Launch security. RLS verified end to end, rate limiting, account deletion, moderation logging, and an anonymous first story flow that cannot leak data across users.
What I need from you:
Real Postgres depth. Migrations, RLS policies, and RPCs you can write and defend, not just familiarity with the Supabase dashboard.
TypeScript that holds up. Typed API responses, consistent error handling, no any scattered through the services.
Comfort integrating third party LLM and TTS APIs, including async job handling and polling, not just calling OpenAI directly.
Willingness to push back when the spec is wrong. It has been through two rounds of outside review and got better both times.
Comfort working on an 18+ platform. The content is explicit adult fiction, not extreme, and there is a content classifier with hard blocks built into the design.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.