feat(ai-core): provider-agnostic AI architecture with streaming by tomymaritano · Pull Request #156 · dripnex/readide

tomymaritano · 2026-03-14T17:37:44Z

Summary

New @readied/ai-core package — provider-agnostic AI architecture with LLMProvider interface, ProviderRegistry, streaming SSE parser, retry logic with exponential backoff, context builder with token budgeting, and AIService orchestrator
Full Electron integration — IPC bridge with batched streaming events, preload API, and renderer AiPanel wired to the new streaming architecture
UX fixes — unified dual AI panel entry points (Cmd+K and Sparkles button) into single panel, fixed panel height to fill container, made React DevTools dev-only via dynamic import
Monorepo fix — pinned @types/react to 18.x via pnpm overrides to resolve lucide-react type mismatch

Architecture

Renderer (AiPanel) → IPC → Main (ai-core bridge) → AIService → ProviderRegistry → AnthropicProvider → SSE stream
                   ←  batched LLMEvents (text/error/done)  ←

What's included

packages/ai-core (new)

LLMProvider interface + ProviderRegistry for pluggable providers
AnthropicProvider with native SSE streaming (no SDK dependency)
ContextBuilder with token budgeting and automatic content trimming
AIService orchestrator with retry, cancellation, and context building
LLMEvent protocol: text, error, done, tool_call, tool_result
Reusable SSE stream parser

apps/desktop

IPC bridge with 50ms batched event streaming (main → renderer)
Preload API: window.readied.ai.chat(), .onEvent(), .cancel()
Unified AI panel (single instance, both Cmd+K and Sparkles button toggle it)
Panel height fix (fills container instead of 350px cutoff)
Dynamic import for electron-devtools-installer (excluded from prod)

Monorepo

Removed deprecated @readied/ai-assistant package
Added pnpm.overrides to pin @types/react@18.3.27 across workspace

Test plan

All 16 packages pass pnpm test
pnpm typecheck passes (17/17 tasks)
pnpm build succeeds
Manual: Cmd+K opens AI panel in ask-notes mode
Manual: Sparkles button toggles same panel
Manual: AI chat streams responses correctly
Manual: Panel fills full height of sidebar

🤖 Generated with Claude Code

Summary by CodeRabbit

New Features
- Added streaming AI chat responses for real-time interaction
- Introduced provider selection (Anthropic, OpenAI, Ollama support)
- Added ability to cancel in-flight AI requests
- Implemented configuration validation for AI providers
- Enhanced chat interface with conversation modes
Refactor
- Restructured AI system for multi-provider support
- Updated settings schema with provider configuration

## Summary Release 0.9.0 — a major milestone with AI features, sync stability, and a complete website redesign. - **Website redesign** with shadcn/ui + Magic UI - **Auth UX rethink** — Enable Sync flow + middleware fix (500 → 401) - **Sync stability** — error propagation, exponential backoff, abort on logout, typed token refresh - **Sync onboarding** — prompt after 5 notes + offline queue visibility - **AI Commands (Cmd+K v1)** — command panel, settings, keybindings - **AI Knowledge (Cmd+K v2)** — RAG, ask notes, related context - **AI Extensibility** — plugin API, presets import/export - **API documentation** ## Version bumps | File | From | To | |------|------|----| | `package.json` | 0.2.0 | 0.9.0 | | `apps/desktop/package.json` | 0.8.0 | 0.9.0 | ## Checklist - [x] Version bumped in root `package.json` - [x] Version bumped in `apps/desktop/package.json` - [x] CHANGELOG.md updated - [x] Tests pass - [ ] QA smoke test on macOS - [ ] QA smoke test on Windows - [ ] Tag `v0.9.0` after merge ## Post-merge After merging to `main`: 1. Tag `v0.9.0` on main 2. Merge `main` back into `develop` 3. Delete `release/0.9.0` branch 🤖 Generated with [Claude Code](https://claude.com/claude-code)  ## Summary by CodeRabbit * **New Features** * AI Assistant panel with modeful interface (chat/ask-notes modes), configurable via settings * AI command presets support (export/import functionality) * Sync status monitoring with error tracking and pending change count * Enable Sync onboarding flow for new users * Redesigned website with modernized design system * Settings panel for AI configuration (API key, model selection, context limits) * **Bug Fixes** * Improved token handling for deep-link authentication * Better sync error classification and exponential backoff * Enhanced device limit error handling  --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

## Summary - Adds `auto-tag.yml` workflow that was created during v0.9.0 development but missed the squash merge to main - Without this workflow, release PRs don't trigger automatic tag creation, breaking the release pipeline ## Context The release pipeline chain is: `release/* PR merge → auto-tag creates vX.Y.Z → release.yml builds + publishes` This workflow was in `release/0.9.0` but the squash merge of PR #149 didn't include it, so v0.9.0 never got tagged or released. ## After merge Once this is on main, I'll manually create the `v0.9.0` tag to trigger the release build. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

## Release v0.9.0 ### Highlights - **Auth → Payment → Sync flow**: License-gated sync with smart step routing in EnableSyncModal - **Stripe checkout**: Monthly (€2/mo) and annual (€20/year) plan selection - **Sidebar drag-and-drop**: Grip handle, circular reference prevention, child depth propagation - **CI/CD**: Auto-deploy API — develop → staging, main → production - **Design tokens**: SyncStatusIndicator uses CSS variables instead of hardcoded colors - **AI features**: Command execution, plugin bridge, knowledge system (Phase 4-5) - **UI cleanup**: Tailwind v4 canonical classes, removed dead design-system package ### Changes since v0.8.0 - feat(sync): connect auth → payment → sync flow with license gating (#154) - chore: Tailwind v4 canonical classes + tsconfig cleanup (#151) - feat(ai): complete Phase 4-5 — AI command execution + plugin bridge (#150) - feat(ai,plugins): complete Phase 4-5 — AI knowledge & extensibility - feat(desktop,api): complete Phase 1-3 roadmap implementation ### Checklist - [x] `pnpm typecheck` passes - [x] `pnpm test` passes - [x] `package.json` version is `0.9.0` - [x] All features merged to develop first - [x] API production secrets configured (Stripe, Resend) 🤖 Generated with [Claude Code](https://claude.com/claude-code)  ## Summary by CodeRabbit # Release Notes * **New Features** * Added AI-powered text commands (summarize, rewrite, tweet) with plugin integration. * Introduced drag-and-drop notebook reordering with automatic circular reference protection. * Enhanced sync onboarding with license-based feature gating. * Improved magic link authentication with automatic verification. * **Bug Fixes** * Fixed token refresh infinite loop handling. * Added URL encoding for verification links. * Improved cleanup for animated UI components. * **Improvements** * Enhanced error handling for authentication and sync failures. * Added accessibility attributes (ARIA labels, semantic HTML). * Updated visual styling with new token-driven design system. * **Refactor** * Migrated from design-system package to plugin-api package. * Consolidated design tokens and theme system. * **Chores** * Updated GitHub Actions release automation. * Refreshed project documentation and contribution guidelines.  --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…used imports Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements context assembly for LLM calls with token-aware budget management: drops oldest history messages and truncates/excludes notes that exceed limits. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements parseSSEStream as an async generator that handles chunked Uint8Array streams, double-newline event boundaries, event/data field parsing, comment skipping, and malformed JSON resilience. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Implements AnthropicProvider with full SSE streaming support, HTTP error classification, usage token tracking, static model list, and config validation. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…etry Implements AIServiceImpl that wires together ProviderRegistry, buildContext, and withRetry into a single chat() entry point returning an abortable AsyncIterable<LLMEvent> handle with start/done envelope events. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Replace legacy ai:query handler with ai-core-based streaming IPC: - Add @readied/ai-core dependency, remove @readied/ai-assistant - Create apps/desktop/src/main/ai/setup.ts: singleton AIService factory using ProviderRegistry + AnthropicProvider with net.fetch - Create apps/desktop/src/main/ai/ipc-ai.ts: registerAIHandlers with ai:chat (streaming), ai:cancel, ai:validate, ai:exportPreset, ai:importPreset; 50ms text batching + per-window handle tracking - Remove old registerAiHandlers() inline function from index.ts Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Replace ai.query() with ai.chat(), ai.onEvent(), ai.cancel(), ai.validate(). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Replace synchronous ai.query() calls with event-driven ai.chat() streaming. Add provider field to settings schema (v3 migration), copy aiCommandTypes to ai-core, and update all renderer imports from @readied/ai-assistant to @readied/ai-core. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

All functionality has been migrated to @readied/ai-core. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Merge duplicate AI panel systems (plugin + App.tsx) into single panel instance managed by App.tsx - Plugin's Sparkles button now dispatches CustomEvent instead of managing its own panel, eliminating the duplicate panel that appeared inside the editor zone - Fix AI panel CSS from fixed height (350px) to 100% to fill the side panel container properly - Default to 'ask-notes' mode (RAG) when opening panel via ⌘K - Make React DevTools a dynamic import so it stays out of production bundle Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucide-react@0.562 ships types against @types/react@19.x, but the desktop app uses React 18. pnpm hoisted the 19.x types from apps/web, causing a type identity mismatch (bigint not in React 18 ReactNode). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

vercel · 2026-03-14T17:37:49Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
readide	Ready	Preview, Comment	Mar 14, 2026 6:07pm

chatgpt-codex-connector · 2026-03-14T17:37:50Z

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

coderabbitai · 2026-03-14T17:41:45Z

Caution

Review failed

Pull request was closed or merged during review

📝 Walkthrough

Walkthrough

The pull request refactors the AI subsystem from a monolithic @readied/ai-assistant package to a new modular @readied/ai-core architecture. It introduces a provider-agnostic streaming protocol with provider registry support (including AnthropicProvider), IPC-based chat integration in the desktop main process, and corresponding updates to renderer components, settings schema, and preload APIs. The legacy AI module is removed entirely.

Changes

Cohort / File(s)	Summary
AI Core Package (New) `packages/ai-core/package.json`, `packages/ai-core/src/types.ts`, `packages/ai-core/src/provider.ts`, `packages/ai-core/src/provider-registry.ts`, `packages/ai-core/src/context-builder.ts`, `packages/ai-core/src/retry.ts`, `packages/ai-core/src/ai-service.ts`, `packages/ai-core/src/ai-command-types.ts`, `packages/ai-core/src/providers/anthropic.ts`, `packages/ai-core/src/providers/sse-parser.ts`, `packages/ai-core/src/index.ts`	New modular AI core with LLMEvent streaming protocol, ProviderRegistry, AnthropicProvider implementation, context builder with token budgeting, retry logic with exponential backoff, AIServiceImpl for managing streaming chats, and AI command types/validation. Comprehensive test coverage added across modules.
Desktop Main Process Integration `apps/desktop/src/main/ai/setup.ts`, `apps/desktop/src/main/ai/ipc-ai.ts`, `apps/desktop/src/main/index.ts`	New AI service instantiation (setup.ts with AnthropicProvider), IPC handler registration (ipc-ai.ts with ai:chat, ai:cancel, ai:validate channels), streaming consumer with event batching, and per-window handle tracking. Main process wired to initialize AI service at startup and replace legacy AI handlers.
Desktop Preload & IPC API `apps/desktop/src/preload/index.ts`	Replaced `query` method with streaming `chat()`, added `onEvent()` listener, `cancel()`, and `validate()` methods. Chat requests now include query, currentNote, relevantNotes, history, mode, provider, and providerConfig fields instead of simple message/system structure.
Desktop Renderer AI Components `apps/desktop/src/renderer/components/ai/AiPanel.tsx`, `apps/desktop/src/renderer/App.tsx`, `apps/desktop/src/renderer/plugins/aiAssistant.tsx`, `apps/desktop/src/renderer/hooks/useRegisterPluginAiCommands.ts`	AiPanel migrated to streaming with active request tracking, event-driven text accumulation, and output-target actions. App.tsx imports updated to use ai-core. Plugin refactored to CustomEvent-based toggle (removed bridge/panel coupling). Hook import paths updated. Added AiInitialCommand interface and event listener for `readied:ai:toggle-panel`.
Desktop Settings & Schema `apps/desktop/src/renderer/stores/settings/schema.ts`, `apps/desktop/src/renderer/stores/settings/settingsStore.ts`, `apps/desktop/src/renderer/pages/settings/sections/AiSection.tsx`	Settings version bumped to 3 with new provider field (anthropic/openai/ollama enum), model relaxed to string, and provider-specific apiKey handling. Migration logic added for v2→v3. AiSection now includes provider selection UI and uses `validate()` instead of `query` for testing connections.
Legacy AI Assistant Removal `packages/ai-assistant/src/claude-client.ts`, `packages/ai-assistant/src/prompts.ts`, `packages/ai-assistant/src/rag.ts`, `packages/ai-assistant/src/index.ts`	Entire claude-client module, prompts and templates, RAG functionality, and public exports removed. All symbols previously re-exported are eliminated, shifting responsibility to ai-core package.
Dependency & Configuration `apps/desktop/package.json`, `package.json`, `.claude/settings.local.json`	Changed ai-assistant workspace dependency to ai-core. Added pnpm overrides for `@types/react` and `@types/react-dom`. Added Bash permissions for tsc, git stash/rm, and turbo commands in Claude settings.
Styling `apps/desktop/src/renderer/styles/ai-panel.css`	Replaced fixed 350px height with 100% flex height and removed top border for improved layout flexibility.

Sequence Diagram(s)

sequenceDiagram
    participant Renderer as Renderer<br/>(AiPanel)
    participant IPC as Electron IPC
    participant Service as AIService
    participant Provider as AnthropicProvider
    participant Network as Anthropic API

    Renderer->>IPC: ai:chat { query, history,<br/>mode, provider }
    activate IPC
    IPC->>Service: chat(request)
    activate Service
    Service->>Service: buildContext(sources,<br/>budget)
    Service->>Provider: chat(options,<br/>config)
    activate Provider
    Provider->>Network: POST /v1/messages<br/>(stream SSE)
    activate Network
    Network-->>Provider: SSE stream<br/>(text_delta, etc)
    deactivate Network
    
    loop Streaming Events
        Provider->>Provider: parseSSEStream()
        Provider->>Service: yield LLMEvent<br/>(text, usage)
        Service->>Service: track tokens,<br/>accumulate
        Service->>IPC: emit text event<br/>(batched, 50ms)
        IPC->>Renderer: ai:event<br/>(text delta)
        Renderer->>Renderer: accumulate text,<br/>update UI
    end
    
    Provider-->>Service: yield done event
    deactivate Provider
    Service->>Service: cleanup handle,<br/>emit duration
    Service-->>IPC: done(duration,<br/>cancelled)
    deactivate Service
    IPC->>Renderer: ai:event (done)
    deactivate IPC
    Renderer->>Renderer: finalize stream,<br/>apply output target

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Possibly related PRs

chore: release v0.9.0 #155: Implements the same AI/phase-4–5 feature refactor with ai-assistant→ai-core migration, provider registry, streaming IPC handlers, and desktop renderer changes.
feat(ai): complete Phase 4-5 — AI command execution + plugin bridge #150: Modifies AI command and prompt surfaces, including AiPanel/command registration and AiInitialCommand wiring overlapping with this PR's component updates.
chore(release): v0.9.0 #149: Related release PR covering identical ai-core package introduction, desktop AI IPC/setup, settings schema migration, and AI panel refactoring.

Suggested labels

size/XL

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 30.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The pull request title clearly and accurately summarizes the primary change: introducing a provider-agnostic AI architecture with streaming capabilities. It is concise, specific, and reflects the main objective of the PR without being overly broad or vague.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings (stacked PR)
📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment
Commit unit tests in branch feature/ai-core-provider-abstraction

📝 Coding Plan

Generate coding plan for human review comments

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

- Accept all apps/web changes from develop - Move aiCommandTypes from ai-assistant into ai-core - Fix test imports for renamed ai-command-types module - Regenerate pnpm-lock.yaml Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

## Summary - Merges all develop changes into main for release preparation - Includes: AI core provider abstraction (#156), automated release pipeline (#157), and all other develop work - All conflicts resolved (develop is source of truth) ## What happens after merge 1. Go to **Actions → Release → Run workflow** (branch: main) 2. semantic-release calculates version, creates tag + draft release 3. Build workflow triggers automatically for mac/win/linux 4. Release is published when all builds succeed ## Test plan - [x] All tests pass (`pnpm test` — 16/16) - [x] No conflict markers remaining - [ ] CI checks pass on this PR 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Resolve conflicts keeping tool use features + CodeRabbit fixes. Incoming from develop: ai-core base (#156), release pipeline (#157), cleanUrls fix (#164). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tomymaritano and others added 18 commits March 13, 2026 18:49

chore: scaffold packages/ai-core package

f65830b

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(ai-core): define LLMEvent protocol and chat types

4d7f7a5

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(ai-core): add DOM lib for AbortSignal/ReadableStream and clean un…

38dd2cf

…used imports Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

feat(ai-core): add LLMProvider interface and ProviderRegistry

a992d06

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(ai-core): add ContextBuilder with token budgeting and trimming

10588b7

Implements context assembly for LLM calls with token-aware budget management: drops oldest history messages and truncates/excludes notes that exceed limits. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(ai-core): add retry logic with exponential backoff and jitter

2951ebe

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(ai-core): add AnthropicProvider with SSE streaming

42fb7e5

Implements AnthropicProvider with full SSE streaming support, HTTP error classification, usage token tracking, static model list, and config validation. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(preload): expose streaming AI chat API via IPC

b40fe5f

Replace ai.query() with ai.chat(), ai.onEvent(), ai.cancel(), ai.validate(). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chore: remove deprecated @readied/ai-assistant package

2f0837c

All functionality has been migrated to @readied/ai-core. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chore: merge develop and resolve conflicts

3629486

- Accept all apps/web changes from develop - Move aiCommandTypes from ai-assistant into ai-core - Fix test imports for renamed ai-command-types module - Regenerate pnpm-lock.yaml Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

github-actions Bot enabled auto-merge (squash) March 14, 2026 18:06

github-actions Bot added dependencies Pull requests that update a dependency file app:desktop size/XL labels Mar 14, 2026

vercel Bot deployed to Preview March 14, 2026 18:07 View deployment

github-actions Bot merged commit a24a231 into develop Mar 14, 2026
14 of 15 checks passed

tomymaritano mentioned this pull request Mar 14, 2026

chore: release preparation — merge develop into main #160

Merged

3 tasks

coderabbitai Bot mentioned this pull request Mar 18, 2026

feat(ai): add tool use (function calling) to AI assistant #165

Merged

6 tasks

coderabbitai Bot mentioned this pull request Apr 24, 2026

feat: local HTTP API, quick capture, mermaid/math plugins + ASAR fix #231

Merged

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(ai-core): provider-agnostic AI architecture with streaming#156

feat(ai-core): provider-agnostic AI architecture with streaming#156
github-actions[bot] merged 19 commits into
developfrom
feature/ai-core-provider-abstraction

tomymaritano commented Mar 14, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

vercel Bot commented Mar 14, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot commented Mar 14, 2026

Uh oh!

coderabbitai Bot commented Mar 14, 2026 •

edited

Loading

Review failed

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

Possibly related PRs

Suggested labels

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

tomymaritano commented Mar 14, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Architecture

What's included

packages/ai-core (new)

apps/desktop

Monorepo

Test plan

Summary by CodeRabbit

Uh oh!

vercel Bot commented Mar 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot commented Mar 14, 2026

Uh oh!

coderabbitai Bot commented Mar 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

Possibly related PRs

Suggested labels

❌ Failed checks (1 warning)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

tomymaritano commented Mar 14, 2026 •

edited by coderabbitai Bot

Loading

vercel Bot commented Mar 14, 2026 •

edited

Loading

coderabbitai Bot commented Mar 14, 2026 •

edited

Loading