Changelog

What's new in each release.

Initial public release of AppStudio — a complete AI-augmented app idea engine for solo developers. Ships with 15+ AI agents, background task execution, and the Real Demand Validation engine.

Real Demand Validation New

• AI generates a full search plan: problem-space queries, target subreddits, and domain keyword list — all in one LLM call
• Subreddit-targeted search: hits reddit.com/r/{sub}/search.json?restrict_sr=1 for the communities most likely to discuss your problem
• Keyword overlap filter: post titles must contain a domain keyword — eliminates off-topic results (history posts, drama, finance) that naive keyword search returns
• 35+ noise subreddits blocked globally (gaming, entertainment, politics, memes)
• HN search via Algolia API with same keyword gate
• TARGETED badge on posts sourced from community-specific searches
• LLM synthesis: demand score (0–100), verdict, evidence summary, top pain points, key insight, recommendation, risk flags
• Runs as a background task — navigate away freely, results save automatically
• UI shows exactly what was searched: queries, targeted subs, and relevance keywords

Background Task System New

• All long-running agents now return a task_id immediately and execute in daemon threads
• Navigate away mid-analysis — the task continues running and results persist to the database
• Global task status bar (bottom-left) shows all running tasks with elapsed time and live status
• Polled every 3 seconds via a stable React interval — no interval explosion on state changes
• Background agents: Demand Validation, Roadmap, Coaching, Code Review, Marketing, Landing Page, Pricing, Stack Advice, SEO

Analysis Agent

• 7-step AI business analyst: Plan → Market → Compete → Revenue → Risk → Verdict → Mitigate
• Streams results live step by step via SSE
• Produces viability score (0–100), revenue score (0–100), and effort score (0–100)
• Final verdict: BUILD / VALIDATE / SHELVE with reasoning
• Step 7 (Mitigate) generates per-risk strategies with effort/impact ratings and residual risk scores
• Revenue model bias toward one-time tool sales and API licensing (by design for solo devs)

Coder Agent

• Code review: scans real project files, generates 3–5 prioritised suggestions (bugs, security, quality)
• Before/after unified diffs for every suggestion
• Approve to write to disk, reject to dismiss — nothing changes without explicit approval
• Feature generation: describe a feature in plain English, get multi-file implementation code
• Scaffolding: generate a complete starter project from your idea description
• Sandboxed file writes — agent can only touch files inside your linked source_path

Project Manager

• AI roadmap generation: 4 phases with 3–6 concrete tasks each, estimated days, critical path
• Task check-off with persistent state
• Coaching sessions: health status, today's focus, active blockers, next milestone, encouragement
• Status levels: on-track / at-risk / stalled / just-started

Marketing Suite Builder+

• Landing Page Copy: headline, sub-heads, feature bullets, CTA, FAQ — copy-paste ready
• SEO Pack: meta title, meta description, target keywords, long-tail keywords, content calendar
• Press Kit: one-sentence pitch, press release, founder bio, product description, talking points
• Launch Campaign: day-by-day pre-launch, launch day, and post-launch sequence
• Community Posts: Reddit and Hacker News drafts (Show HN format)
• Competitor Intel: pricing, positioning, weaknesses, your attack vector

Security Scanner Indie+

• Pentest simulation: scans for OWASP Top 10, injection, auth issues, exposed secrets
• Findings include severity (Critical/High/Medium/Low), CWE reference, affected file and line
• Compliance check: OWASP ASVS, GDPR basics, API security best practices
• Auto-fix: generates patches via the Coder Agent for supported finding types

Platform and infrastructure

• Multi-platform: Windows (x64), macOS (Apple Silicon), Linux (x64)
• Multi-provider LLM support: Ollama, LM Studio, Jan, and any OpenAI-compatible local server
• Auto model detection and selection from installed Ollama models
• Model priority order: qwen3:14b → qwen3:8b → qwen2.5:14b → qwen2.5:7b → llama3.2
• SQLite database — single file, zero infrastructure, fully offline
• FastAPI backend + React/Vite frontend

License system

• Four tiers: Indie ($197), Builder ($397), Team ($997), Enterprise ($7,500+/yr)
• One-time purchase, no recurring checks after activation
• Works fully offline after initial activation
• 14-day free trial with Builder-level access (no credit card required)
• Machine-level activation with self-service deactivation for machine transfers

UX and capture

• Ctrl+Space global quick-capture shortcut — works from any app
• Idea vault with status pipeline: raw → researched → validated → building → live → shelved
• Idea detail view with tabbed layout: Analysis, Build, Manage, Marketing, Security
• Status auto-advances from raw → researched after first analysis run