Briefing

Claude Plan Overview (Phases 1–3)

Source: docs/briefings/claude-overview.md · Back to index · Beta hub

Claude Plan Overview (Phases 1–3)

Digest of apps/stps/deploy_n_v2/v2/claude/overview_sum.md. Use this as the engineering brief once Phase 0 is deployed.

Phase 1 – Foundation

Enhanced Session Model

  • Add optional fields (activities, focus_mode, intention, memory, artifacts_summary, continues_from, branches_to) without schema-breaking changes.
  • Memory stays lightweight (arrays of strings) until richer storage lands later.

Universal Blocks (Phase 1 scope)

  • Define UniversalBlock union: static (node refs), dynamic (queries), interactive (handlers), ai-generated.
  • Basic AI actions: suggest title/tags (editable) and “Add AI key points” block with clear badging; no autonomous writes.

Feature Flags

  • Three groups: core, enhanced, experimental.
  • Default flags: sessionMemory=true, universal blocks/slash commands/node references disabled until ready; experimental flags off by default.
  • Helper isEnabled differentiates enhanced vs experimental keys.

Phase 2 – Intelligence & Grid Composer

Grid Composer

  • GridComposer model with rows, columns, cells, connections, and interactive metadata (resizable/draggable/executable).
  • Sessions can attach a composer snapshot + history and optionally save as templates.

Context Engine

  • Central service that tracks the current session, grid, recent nodes, and active queries.
  • Provides getContext() for UI hints and getAIContext() for AI helpers (session memory, grid state, recent interactions, active glossaries/patterns/templates).

Smart Session Navigation

  • Workspace layout built around sessions instead of features: Session bar (current + switcher + new), Grid Composer as main canvas, context-aware side panel (relevant nodes, suggestions, session memory).

Phase 3 – Full Platform Integration

Node/Edge Migration

  • Unified Node interface with top-level indexed fields (title/status/visibility/owner) plus props for flexible data and AI metadata (embeddings, extracted entities).
  • Edge objects become first-class with their own props, strength/confidence, temporal bounds, and origin (user/ai/system).
  • Privacy: embeddings stay server-side; extracted metadata is redactable.

Workspace System

  • Workspaces defined by purpose (learning | creating | managing | personal), feature sets, and layout presets.
  • Examples: translation-studio, content-studio, learning-space each map to grid templates and feature toggles.

Integration Points

  • Existing apps become grid-cell adapters (e.g., translation panel cell, AI content cell).
  • Key idea: start session → grid adapts → add cells → everything links → session memory grows.

Suggested Implementation Order

  1. Week 1–2: Phase 1 foundation (session props, universal blocks, flags).
  2. Week 3–4: Grid Composer prototype.
  3. Week 5–6: Session chains + memory.
  4. Week 7–8: Integrate existing apps as grid cells.
  5. Future: Full node/edge migration.

Keep this file handy when sequencing the post-Phase 0 work across engineering and product.