memorix 1.0.7 → 1.0.9

package/CHANGELOG.md

@@ -1,11 +1,65 @@
 # Changelog
 
-All notable changes to this project will be documented in this file.
-
-## [1.0.7] - Unreleased
-
-### Visual Semantic Layering (dashboard follow-up)
-- Team page headline is now **active-only**; `recent`/`historical` shown as secondary subtitle only.
+All notable changes to this project will be documented in this file.
+
+## [1.0.9] - 2026-05-19
+
+### Added -- Knowledge Base / LLM Wiki Foundation
+- **First-class Knowledge Base layer** -- Added a readable project knowledge layer generated from durable memory, git-backed facts, mini-skills, and project evidence. The Knowledge Base is a shared human-and-AI surface, not a replacement for raw observations and not a TUI-only feature.
+- **Knowledge graph projection** -- Added a semantic graph projection over the same eligible knowledge inputs, preserving source refs for navigation and traceability.
+- **Dashboard Knowledge foundation** -- Dashboard surfaces can consume the same generated knowledge contract as the TUI, keeping Knowledge Base and graph semantics aligned.
+
+### Added -- TUI Knowledge Workbench
+- **Tabbed terminal workbench** -- Reworked the TUI around `Home`, `Knowledge`, `Memory`, `Workbench`, and `Graph` tabs.
+- **Knowledge browsing** -- Added a Knowledge tab for generated wiki sections, item summaries, entities, and source refs.
+- **Memory browsing** -- Added a Memory tab for recent/search/detail flows with ref-focused navigation from Knowledge.
+- **Workbench session center** -- Added explicit session status, explicit bind/end actions, context source summary, and chat. Entering the tab does not auto-start a session.
+- **Graph text browser** -- Added a terminal-native Knowledge Graph tab with summary stats, cluster browsing, node detail, light filtering, and Graph -> Knowledge jumps.
+- **Cross-surface ref navigation** -- Added Knowledge -> Memory, Memory -> Knowledge, and Graph -> Knowledge navigation paths based on stable refs where available.
+
+### Changed
+- **Knowledge architecture clarified** -- Documented the memory layer as the primary input to the knowledge layer, while project evidence such as docs, source structure, and git facts act as secondary inputs.
+- **TUI scope tightened** -- The TUI now behaves as a knowledge-native workbench, not a placeholder coding-agent harness. No fake Run Task / Review controls are exposed.
+- **Session semantics preserved** -- `session_start` remains explicit and lightweight. Agent Team join remains opt-in.
+
+### Fixed
+- **Formation duplicate detection (#94)** -- Fixed Formation resolve comparing raw Orama/BM25 ranking scores against normalized similarity thresholds. Noisy log-dump memories with high backend scores no longer discard unrelated valid memories as duplicates.
+- **TUI chat input lock** -- Fixed a regression where after a free-text chat response the bottom input bar became unresponsive because the chat overlay disabled the CommandBar. Chat overlay now keeps input active so users can continue the conversation.
+- **TUI read-only overlay input lock** -- Fixed read-only overlays such as `/doctor` disabling the CommandBar. Users can now keep typing commands or questions while diagnostics and other non-action views are open.
+- **TUI shortcut/input conflicts** -- Fixed CommandBar conflicts with view shortcuts such as Graph `f`/`k`, Memory `k`, and Knowledge `m` while still allowing normal first-character input.
+- **Graph TypeScript build issue** -- Fixed a type-only re-export under `verbatimModuleSyntax`.
+
+### Known Limitations
+- The Knowledge Base and Knowledge Graph are read-only generated projections in 1.0.9.
+- 1.0.9 does not include a desktop app, realtime agent messaging layer, GraphRAG system, graph editing, or full coding-agent harness.
+
+### Stats
+- **Verified TUI subset:** 95 passing tests across `interaction`, `workbench`, and `graph` suites.
+- **Release gates run during cleanup:** `npm run lint`, `npm run build`, and full `npm test`.
+
+## [1.0.8] - 2026-04-19
+
+### Added -- Operator CLI Surface
+- **Namespaced operator commands** -- Added human-oriented CLI namespaces for `session`, `memory`, `reasoning`, `retention`, `formation`, `audit`, `transfer`, `skills`, `team`, `task`, `message`, `lock`, `handoff`, `poll`, `sync`, and `ingest` so Memorix-native operations no longer require raw MCP tool calls.
+- **Lightweight session start + explicit collaboration join** -- `memorix session start` now opens project-bound memory sessions without auto-registering a team identity. Operators can opt into collaboration with `--joinTeam`, or join later via `memorix team join`, while still getting default role mapping from `agentType` when they do join.
+- **CLI-first project ops** -- Terminal operators can now inspect memory detail/timeline, store and search reasoning traces, run retention and formation checks, audit attribution, export/import memory snapshots, inspect/generate skills, run explicit rules/workspace sync flows, and ingest images without leaving the shell.
+
+### Added -- Programmatic SDK
+- **`memorix/sdk` subpath export** -- `createMemoryClient()` factory returns a self-contained `MemoryClient` with `store`, `search`, `get`, `getAll`, `count`, `resolve`, and `close` methods. No MCP transport or CLI needed — initialize from a Git project root, read/write observations directly.
+- **`createMemorixServer` re-export** -- Embed the full MCP server into your own Node.js process and connect it to any transport.
+- **`detectProject` re-export** -- Standalone Git-based project detection.
+- **Three subpath exports** -- `memorix/sdk` (runtime API + types), `memorix/types` (type-only), `memorix` (MCP stdio entry).
+
+### Added -- Official Docker Deployment
+- **Official HTTP control-plane container path** -- `Dockerfile` now builds a real `serve-http` runtime image instead of defaulting to stdio MCP.
+- **Compose example** -- added `compose.yaml` with port `3211`, persistent data volume, and `/health` healthcheck.
+- **Docker docs** -- added `docs/DOCKER.md` and linked Docker deployment from README / setup docs.
+- **Runtime truth** -- documentation now explicitly states that Docker support is for the HTTP control plane and that project-scoped Git/config behavior requires the container to see the bound repo path.
+
+### Visual Semantic Layering (dashboard follow-up)
+- Team mental model tightened before release: docs and dashboard copy now say "explicit collaborators" instead of implying a persistent IDE-window roster.
+- `memorix team status` now defaults to active collaborators only; inactive/historical identities remain available with `--all` for audits without flooding the normal operator view.
+- Team page headline is now **active-only**; `recent`/`historical` shown as secondary subtitle only.
 - Team agents list gets four filter tabs: **Active** (default) / Recent / Historical / All. Historical rows no longer flood the view; they sit behind an explicit "Show historical (N)" toggle.
 - Each agent row carries an explicit tier badge (Active/Recent/Historical) instead of a single "offline" sea of red.
 - Historical agent rows are de-emphasized (reduced opacity, muted colors) so they don't dominate the main view.
@@ -14,9 +68,9 @@ All notable changes to this project will be documented in this file.
 - Project switcher groups items into **Current / Real projects / Temporary / Placeholder**, with temporary and placeholder folded behind a "Show temporary" toggle. Each group has a visible count so users can tell a real project list apart from the historical test/smoke scratch pile.
 - All new labels wired through the i18n system (English + Simplified Chinese).
 
-### Team Page Cleanup (1.0.7 final)
-- **Team page semantics**: clearly presented as a **project collaboration space**, not an organization backend or staffing admin tool. Scope labels: "Project Collaboration" / "All Projects".
-- **`memorix_session_start` auto-registration**: agents are automatically registered in the team with a default role derived from `agentType` via `AGENT_TYPE_ROLE_MAP` (e.g. `windsurf`→`engineer`, `gemini-cli`→`researcher`). No separate `team_manage(join)` call needed. Explicit role overrides the default.
+### Team Page Cleanup (1.0.7 final)
+- **Team page semantics**: clearly presented as a **project collaboration space**, not an organization backend or staffing admin tool. Scope labels: "Project Collaboration" / "All Projects".
+- **Explicit collaboration join**: `memorix_session_start` is lightweight by default and no longer auto-registers a team identity. Collaboration is now opt-in via `joinTeam: true` or `team_manage(join)`, and default role mapping from `agentType` applies only when joining.
 - **"Continue This Project" resume area**: shows open tasks, available-to-claim tasks, open handoffs, unread messages, active locks, and active agent count at the top of the Team page. Provides a clear "pick up where you left off" entry point.
 - **Unified statistics**: stat cards clearly labeled — Active Agents (with session count + historical total), Locked Files, Tasks (by status), Messages (unread count). No conflicting or ambiguous labels.
 
@@ -30,7 +84,15 @@ All notable changes to this project will be documented in this file.
 ### Added -- Team Identity (Phase 4)
 - Team store with agent registration, heartbeat, task board, handoff artifacts, stale detection. Prompt identity contract, sticky attribution prevention.
 
-### Fixed
+### Fixed
+- **HTTP stale-session poisoning (#82)** — `serve-http` no longer has only a hardcoded 30-minute idle GC. Added `MEMORIX_SESSION_TIMEOUT_MS` so long-running HTTP MCP clients such as Codex can keep sessions alive for longer work blocks without getting stuck on stale `Mcp-Session-Id` failures.
+- **TUI responsive layout** — sidebar width now scales with terminal width (28% ratio, 26-40 range) instead of fixed 34 columns; content area uses `flexGrow` to fill remaining space; maximized terminal windows now show expanded UI instead of locked small-window layout
+- **TUI search score display** — replaced absurd raw-percentage display (e.g. "927%") with relative relevance dots (●●●○○) normalized against top result
+- **TUI /resume numeric index** — `/resume 2` now selects thread #2 from the list (1-based), in addition to `/resume <threadId>` by ID
+- **TUI CJK text wrapping** — Ink's `wrap="wrap"` doesn't understand double-width CJK characters, causing garbled text; replaced with CJK-aware manual line splitting in chat output and ContextRail sizing so mixed CJK content renders predictably
+- **TUI layout DRY** — sidebar/content width calculation extracted to shared `computeLayoutWidths()` in theme.ts; App.tsx and ChatView.tsx now use the same function, preventing drift
+- **TUI CommandBar overlay** — palette no longer pushes input bar down; rendered as overlay in App.tsx; mouse SGR mode centrally managed to avoid conflicts between palette clicks and chat scrolling
+- **TUI chat freeze under assistant output** — removed assistant Markdown rendering from the workbench chat path and render chat bodies as plain text with CJK-aware wrapping; this avoids heavy reparsing/rerender spikes that could freeze the TUI when the assistant started answering after tool calls
 - **#4** Parallel multi-agent — fully implemented via orchestrator
 - **#52** observations.json perf — migrated to SQLite
 - **#56** LLM rerank timeout — configurable via `MEMORIX_RERANK_TIMEOUT_MS`
@@ -92,7 +154,7 @@ All notable changes to this project will be documented in this file.
 - **E2e demo tests** (`tests/e2e/`) excluded from default `vitest run` — these test CLI-agent demo artifacts, not Memorix product code. Available via `npm run test:e2e`.
 - **Live LLM quality tests** excluded from default suite — require `MEMORIX_RUN_LIVE_LLM_TESTS=1`. Rules-only fallback test preserved in `tests/memory/formation-rules-fallback.test.ts`. Available via `npm run test:llm-live`.
 - **Coordinator merge-conflict test** made deterministic — synchronous conflict file writes in `spawn()` instead of async `setTimeout` race. Removed `{ retry: 3 }`.
-- **Default test suite**: 141 files, 1906 tests, **0 skipped, 0 failed**.
+- **Default test suite**: 156 files, 2064 tests, **0 failed**.
 
 ### Added -- Hooks Test Coverage
 - Audit ledger lost/corrupted → re-install recovers audit entry (codex).