55 lines (43 loc) · 2.33 KB

Roadmap

Completed ✅

v0.1–v0.5 — Digest browsing, SQLite storage, i18n, Google OAuth, Sources CRUD, Source Packs sharing, JSON/RSS Feed, Marks
v0.6 — Soft Delete Sources (prevent pack zombie resurrection)

Near-term 🔨

Priority	Feature	Description
P0	Multi-tenant Phase 1	`raw_items` table + collection pipeline; decouple source fetching from digest generation
P1	Multi-tenant Phase 2	Personalized digests based on user subscription combinations
P1	Sources → Cron integration	Cron reads `/api/sources?active=true` instead of hardcoded Twitter

Mid-term Directions 🧭

1. AI Agent Embed

Embed an AI editorial assistant (Chat Widget) for interactive digest exploration:

Floating chat bubble (bottom-right)
Behavior-aware: observe browsing patterns, proactive suggestions
Q&A: deep-dive into current digest topics
Example: "What's the background on this?" "Track this topic for me"

2. Agent Friendly

Make the entire system AI-agent-friendly, lowering automation barriers:

Structured API output (JSON Schema)
MCP Server support (let Claude/GPT agents operate Digest directly)
Webhook callbacks (source update, digest generation events)
Idempotent operation design (safe agent retries)

3. Channel Push

Multi-channel proactive digest delivery — users choose how to receive:

Telegram Bot — Scheduled push + on-demand queries
Feishu/Lark — Group bot / DM delivery
Email — Periodic email summaries (daily/weekly)
Slack — Webhook / Bot integration
Discord — Channel push
RSS/JSON Feed — Done ✅
Per-user: choose delivery channels + frequency preferences

Long-term 🔭

Feature	Description
Source Market	Community source discovery, trending packs, category browsing
Subscription combo caching	Same subscription set = shared digest, reducing LLM costs
Multi-language Digest	Same source pool → different language outputs
Paid tiers	Source limits, premium source types, higher generation frequency

AI Testing (Exploring)

Current: curl E2E scripts (66 assertions, 18 categories)
Planned: Playwright + Midscene for UI-level AI testing
Direction: Natural language assertions, aligned with Agent Friendly roadmap