Abhilash Reddy Sirigi ReddyAbhilash Reddy

Work

wovv.ai: an AI writing partner that remembers the whole story

Founder, wovv.ai, 2026 to now

  • Python
  • FastAPI
  • Next.js
  • LangChain
  • RAG
  • Neo4j
  • Postgres/pgvector
  • GCP
  • Cloud Run
  • Pulumi

Context

wovv.ai is an AI-native storytelling platform for novelists and screenwriters. I founded it in 2026 and have designed, built and shipped it end to end as a solo, bootstrapped founder. It is live in public preview in 22 languages: English plus 21 Indian languages.

The problem

A long story is a world, not a sequence of sentences. wovv lets a writer branch a story and keep both versions, and once a story branches, the AI has to remember the whole world on each timeline rather than predict the next word. A partner that forgets is worse than none.

There was a second problem. I did not want a tool that writes for the writer. The writer had to stay the author.

And a third: I was one person building a three-service product, so quality, governance and cost had to be designed in rather than staffed.

What I built

  • A three-service platform from scratch. A Next.js web app, a FastAPI backend with RAG and knowledge-graph pipelines, and a GPU media-generation service.
  • Story memory. A RAG and knowledge-graph memory layer on Postgres and Neo4j that keeps persistent narrative memory across branching story timelines.
  • Tara, the in-product writing partner. She lays out options and asks questions. Nothing reaches the page until the writer chooses.
  • An engineering organisation run as a multi-agent system. AI councils plan, build, review and govern the work under formal decision records and architecture-review gates.
  • A release quality system. Golden-set evals and canary invariants validate LLM output before every release.
  • Infrastructure as code on GCP. Everything is defined in Pulumi and runs on Cloud Run, including GPU workloads, with scale-to-zero so idle services scale down rather than sit running.

The product itself covers story analysis (overall health, pacing, themes, readability, dialogue), character analysis (screen time, emotional arc, relationships), a beat sheet, rewinding to any earlier point, and one story in several shapes: prose, screenplay, shot script and stage play.

Outcome

wovv.ai is live in public preview in 22 languages. The whole platform runs for under $700 a month at full operating capacity, and ran for under $500 a month during the build. I pitched it at Draper India. It is bootstrapped, with zero paid marketing.

You can try it at wovv.ai.

All work