flutterforge_ai 0.3.0

AI-Ready Flutter template with built-in DB console, API inspector, state viewer, log viewer, and AI debug snapshots. Build observable apps that AI can debug with full context.

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.3.0 - 2026-04-18

Changed — dependency refresh (fixes pub.dev "up-to-date" score)

All direct dependencies now accept current-major versions:

Package	Was	Now (constraint)
flutter_riverpod	^2.5.1	>=2.5.1 <4.0.0 (3.x)
device_info_plus	^10.1.2	>=10.1.2 <14.0.0
flutter_dotenv	^5.1.0	>=5.1.0 <7.0.0
package_info_plus	^8.0.2	>=8.0.2 <11.0.0
sensors_plus	^6.0.1	>=6.0.1 <8.0.0
share_plus	^10.0.2	>=10.0.2 <14.0.0

⚠ Breaking — Riverpod 3 only

• FFStateObserver is rewritten for the Riverpod 3.x observer API (ProviderObserverContext replaces the 2.x (ProviderBase, …, ProviderContainer) signature). The class is now final to satisfy Riverpod's base supertype modifier.
• The class's public surface (FFStateObserver({store, trackingPrefixes, maxValueLength}), .storeForTesting) is unchanged — most users won't notice.
• Upgrade path for apps pinned to Riverpod 2.x: stay on flutterforge_ai: ^0.2.x (still on pub.dev) until you bump to Riverpod 3, then move to ^0.3.0.

Fixed

• share_plus 13.x deprecations — snapshot share now uses SharePlus.instance.share(ShareParams(...)) instead of the removed Share.shareXFiles / Share.share static helpers.

0.2.0 - 2026-04-17

Added — "Diagnose with AI" (the killer moment)

• New src/ai/ module wiring a bring-your-own-key LLM call directly into the snapshot preview:
  • FFAiProvider enum — anthropic (Claude Messages API) + openai (Chat Completions, also works with Azure OpenAI / Ollama proxies via the baseUrl override).
  • FFAiConfig — immutable config with effectiveModel / effectiveBaseUrl fallbacks. toJsonSansKey() guarantees the API key is never serialised alongside snapshots.
  • FFAiClient.forConfig(config) factory dispatching to the right provider implementation; tests inject a mock Dio.
  • FFAnthropicClient + FFOpenAiClient — concrete impls, both handle non-2xx responses as FFAiException (with status code + raw provider body for easier debugging).
  • FFAiSettingsStore — SharedPreferences-backed persistence under the flutterforge.ai.* namespace.
  • FFAiPrompt — opinionated system prompt + user-prompt builder that embeds the snapshot JSON and the problem statement.
• New UI:
  • AiSettingsScreen — provider picker, key/model/base-URL inputs, show/hide toggle, clear-key action.
  • DiagnoseResultScreen — fires the diagnosis, shows a spinner, renders the response as selectable text with FFAiResponse metadata chips (provider, model, token counts, finish reason), retry button, settings shortcut.
  • Snapshot preview screen now has a "Diagnose with AI" primary button (plus a settings gear in the app bar); the old "Copy AI prompt" stays as a secondary action so the manual workflow remains available.
• Tests: FFAiConfig behaviour, FFAnthropicClient happy path + auth-error path (mocked Dio), FFAiProviderX defaults.

Changed

• Version bumped 0.1.2 → 0.2.0 (new public API surface).
• Barrel re-exports the new ai/ modules and the two new screens.

0.1.2 - 2026-04-17

Added

• FlutterForgeAI.generateSnapshot(problem: '…') — thin top-level facade over FFSnapshotGenerator.generate(...). Both continue to work; prefer the facade at call sites that already reference FlutterForgeAI.

Changed

• README: new "What's actually in the box" capability table right after the 30-second workflow, so reviewers can see the shipped feature set (DevDashboard / snapshot / interceptor / masking / etc.) without scrolling. Roadmap section rewritten with an honest shipped/next split — CLI marked ✅, cloud parked, "Diagnose with AI" flagged as the real next move.
• ECOSYSTEM.md: status table now labels the VS Code extension as v0.2 in-repo (not just a scaffold), the CLI with its current limits, and the cloud as "parked post-v1".

0.1.1 - 2026-04-17

Changed

• Docs: README rewritten with sharper one-line positioning, a 30-second workflow diagram, a comparison table vs Alice / Talker / pretty_dio_logger, explicit dependency vs dev_dependency guidance, and a screenshot slot reference list.
• Docs: Added doc/WORKFLOW.md — step-by-step recording script and asset-layout guide for the doc/images/ screenshots + hero GIF.
• Version bumped so the published pub.dev page picks up the new README.

Fixed

• FFSensitiveDataMasker.maskUrl — Uri.replace URL-encodes the *** mask as %2A%2A%2A; the masker now decodes it back so the devtools UI shows token=*** instead of token=%2A%2A%2A.

0.1.0 - 2026-04-17

Added

• Database console with table browser, row viewer, and raw SQL runner via FFDbHelper.
• Optional sqflite_dev web workbench enabler (no-ops if the dev package is absent).
• In-app API Inspector (FFApiStore) powered by a Dio interceptor with cURL export.
• Riverpod state viewer (FFStateObserver) tracking provider adds, updates, disposals, and failures.
• Log viewer (FFLogger + FFLogStore) with level filtering, search, and full stack traces.
• AI Debug Snapshot generator (FFSnapshotGenerator) capturing DB schema, API calls, state, logs, device, and app info as structured JSON for pasting to AI assistants.
• Clipboard / share / save-to-file actions for snapshots.
• Floating access buttons (DevTools FAB + AI FAB), draggable bubble overlay, shake-to-open (mobile), and Alt+F12 keyboard shortcut (desktop).
• Automatic sensitive-data masking for headers, body keys, and URL query params.
• Release-mode safety: every devtool, snapshot generator, and trigger is automatically disabled when kReleaseMode is true.
• Cross-platform support: Android, iOS, macOS, Windows, Linux, Web.
• 80%+ unit-test coverage across ring buffer, masker, logger, API interceptor, snapshot generator, and prompt formatter.
• Fully working example app demonstrating CRUD, real HTTP calls, Riverpod state, error capture, and snapshot generation.