# How to Vibe Code Your Own Wholana (and Stop Paying for It) > The Content System for TikTok, built for MENA - Site: https://wholana.com - Category: SaaS - Verdict: **Serious undertaking** (45/100 vibecodeable) - Estimated effort: 3-4 weeks of serious part-time work ## Verdict Build a solo web app subset with Next.js, Turso, and Vercel, but expect weeks of work on the hybrid search and scraping pipelines. At $5/mo, paying for Wholana is financially rational if you create content in MENA. Replicating it solo is a serious multi-week engineering project because you have to build or maintain resilient TikTok scraping infrastructure, parse Arabic/Egyptian transcripts reliably, implement Reciprocal Rank Fusion across multiple search lanes, and construct a real-time collaborative writing canvas with Liveblocks. An AI coding agent will scaffold the UI and CRUD routes in hours, but you will spend days debugging edge cases in vector search weighting, multi-user CRDT synchronization, and API rate limits on upstream video platforms. ### What you can't replicate - The pre-compiled historical Egyptian TikTok corpus and daily automated sweep pipelines - Established creator baseline benchmarks and proprietary taxonomic data - Existing community trust and workflow integrations ## What it does Content research, data analytics, and script-writing system for TikTok creators and marketers with a deep localization focus on the MENA region. ### Core features - Filterable TikTok explore feed with multi-facet faceted search - Hybrid semantic search (vector embeddings + full-text + token exact-match via RRF) - Automated craft decoding (hook, retention, stakes, engagement, format) - In-app AI assistant (Ask Wholana) with bounded tool calling and confirmation-gated writes - Model Context Protocol (MCP) server for Claude and ChatGPT integration - Non-destructive swipe file and collection management - Collaborative script editor with live cursors, reference rail, and automated script linter - Creator benchmark dashboard tracking outlier performance ## The business ### Pricing - Personal: $5/mo — For a solo creator getting started. - Plus: $20/mo — For a creator going deeper. - Team: $20/seat/mo — For a workspace creating together (2-seat minimum). Founded 2026. Team size: Small operator team. ## The hard parts - Maintaining stable large-scale TikTok ingestion scrapers bypassing anti-bot measures and rate limits - Processing and transcribing Arabic audio accurately with regional dialect support - Implementing Reciprocal Rank Fusion (RRF) across three distinct search lanes (vector, full-text, exact-token) - Real-time multiplayer collaborative document editing with presence awareness - Prompt-engineering consistent 5-axis JSON structuring from video transcripts ## How to vibe code Wholana ### Prerequisites - Node.js (free): Runtime environment for Next.js and TypeScript toolchains. - GitHub (free): Version control and deployment pipeline source. - OpenAI API Key (Pay-as-you-go (~$10-20/mo)): Used for Whisper transcriptions and structured GPT video craft extractions. ### Recommended AI tools - Claude Code: Best-in-class agentic CLI coding tool for scaffolding complex full-stack features, database schemas, and multi-file refactors. - Cursor: Ideal for iterative UI work on the dashboard, explore grid, and script editor components with precise diff review. ### Stack - Frontend: Next.js with Tailwind CSS and shadcn/ui components - Backend: Next.js API routes / Server Actions with Vercel AI SDK - Database: Turso (libSQL/SQLite at the edge with vector support) - Auth: better-auth - Payments: Skipped (personal use clone) - Other: Liveblocks (real-time multiplayer collaborative script editing), OpenAI API (Whisper STT and GPT-4o analysis), Resend (transactional email alerts) ### Hosting - Vercel (Next.js frontend and serverless API deployment): $0-20/mo - Turso (Serverless edge database with vector capabilities for transcript semantic search): $0/mo (free tier) ### Build guide 1. **Project Scaffolding and Database Schema** — Initialize the Next.js project with Tailwind CSS, configure Turso database client, and set up schema tables for videos, transcripts, craft metadata, swipe collections, and scripts. ``` Create a new Next.js TypeScript project using App Router, Tailwind CSS, and shadcn/ui. Configure Turso (libSQL) as the database provider with tables for: videos (id, title, creator_handle, creator_followers, views, likes, comments, shares, outlier_score, created_at), transcripts (video_id, chunk_text, vector_embedding), craft_decodes (video_id, hook_verbatim, hook_archetype, narrative_structure, stake_type, share_driver, format, beat_sheet_json), collections (id, title, workspace_id), swipe_items (id, collection_id, video_id, notes), and scripts (id, title, content, workspace_id). Implement the database connection utility and migration runner using Drizzle ORM. ``` 2. **Ingestion & Craft Decoding Pipeline** — Build an admin ingest utility that downloads video metadata and audio, transcribes via OpenAI Whisper, and parses structured JSON craft breakdowns using the OpenAI API. ``` Build an ingestion utility in Next.js server actions that accepts a TikTok video URL or raw metadata payload. Download the audio track, send it to OpenAI Whisper for timestamped transcription, and pass the transcript to an LLM structured output call (using Zod) to extract 5 craft axes: Hook (verbatim, archetype, specificity), Retention (narrative structure, beat sheet), Stakes (gain, type), Engagement (comment bait, CTA, share driver), and Format. Store the resulting record in Turso with proper vector embeddings generated via OpenAI text-embedding-3-small. ``` 3. **Explore Feed & Hybrid Search Engine** — Implement the filterable explore grid with facet filtering and hybrid search combining vector similarity, full-text prose matching, and exact-token caption matching fused via Reciprocal Rank Fusion (RRF). ``` Build the Explore feed page with a multi-facet sidebar filtering by subject taxonomy, creator follower tiers, craft attributes (hook archetypes, formats), and statistics (outlier score, views > X). Implement a hybrid search backend API that queries three lanes: vector similarity on transcript chunks, full-text search on English craft summaries, and exact-token search on captions. Combine the ranked result sets using Reciprocal Rank Fusion (RRF) and display the resulting video cards in a responsive grid with creator folding support. ``` 4. **Swipe File & Collections** — Create the non-destructive bookmarking system allowing videos to be saved into multiple workspace collections with individual contextual notes. ``` Build the Swipe File interface and backend mutations. Allow users to organize saved videos into named collections stored in Turso. Each swipe item must support a non-destructive user note and be viewable across grid and list layouts within the workspace. Implement drag-and-drop or modal assignment for adding explored videos to multiple collections. ``` 5. **Real-time Collaborative Script Editor & Linter** — Integrate Liveblocks for multiplayer collaborative text editing with live cursors, presence avatars, a pinned reference rail, and an automated script linter. ``` Integrate Liveblocks into the Script Editor page to support multi-user real-time co-editing with live cursors and presence facepiles at the top of the canvas. Build a pinned reference rail on the right side that pulls pinned videos from swipe collections and matches them against the draft. Add a 'Script Check' button that sends the script content and pinned reference metadata to an LLM endpoint, returning structured feedback on transformation strength, structure, stakes, and actionable improvements. ``` 6. **Ask Wholana AI Assistant & Dashboard** — Build the side-panel AI assistant with bounded tool calling and confirmation-gated actions, alongside the creator benchmark performance dashboard. ``` Build the Ask Wholana chat assistant in a resizable bottom-right slide-over panel. Implement a tool-calling agent loop bounded at 6 steps that can query the corpus, filter the feed, or propose actions (create script, save videos, save Lens). Ensure all write actions render as confirmation cards requiring explicit user click before execution. Finally, build the personal dashboard page aggregating user-posted video performance metrics against niche medians to diagnose pipeline weaknesses. ``` ### Cost vs paying **Starting costs (one-time):** - OpenAI API initial development credits: $15.00 - Total: ~$15 one-time **Ongoing costs (monthly):** - Turso database & Vercel hosting: $0/mo (free tier) - OpenAI API usage (Whisper & embeddings/LLM calls): ~$10-25/mo - Liveblocks collaborative editing: $0/mo (free tier) - Total: ~$10-25/mo - Paying for the SaaS instead: $5 - $20/mo - Build time: 35-50 hours - AI tool credits: $20/mo (Claude Pro / Cursor) - Break-even: Never (built for learning and custom control) ## Sources - [Wholana Official Website & Documentation](https://wholana.com)