@vheins/local-memory-mcp 0.18.13 → 0.19.0

package/dist/prompts/create-task.md

@@ -13,10 +13,10 @@ tags: [workflow, task-creation, planning, mcp]
 
 ## Create Task
 
-Entry=S0 → S1 → S2 → S3 → S4  Exit=created
+Entry=S0 → S1 → S2 → S3 → S4 Exit=created
 Guard: S(N) req S(N-1)✅; NO code/edit/delete — MCP tools ONLY (allowed: task-create, task-list, task-detail, task-update, memory-store, memory-search, standard-search, standard-store, handoff-list, handoff-update, read)
 
-S0 | pre_analysis: memory-search(architecture/history) + standard-search(if task leads to code/test/refactor/migrate decisions) + handoff-list(pending; close stale that describe completed work) + read code(verify paths+impl) + task-list dedup(DO NOT duplicate; link via parent_id/depends_on) | — | context | —
+S0 | pre_analysis: agent-context(one-call) OR memory-search(architecture/history) + standard-search(if task leads to code/test/refactor/migrate decisions) + handoff-list(pending; close stale that describe completed work) + read code(verify paths+impl) + task-list dedup(DO NOT duplicate; link via parent_id/depends_on) | — | context | —
 S1 | design tasks: atomic(1 logical change), layered(DB/Service/State/UI), context(paths+symbols+APIs), min 1 pos+1 neg test | S0✅ | task specs | —
 S2 | assign attributes: task_code(optional — auto-generated as TASK-xxx if omitted), phase(Discovery|Implementation|Testing), priority(1=Low..5=Critical), strict description format | S1✅ | task attrs | —
 S3 | create via task-create(bulk max 500) + log decisions via memory-store(arch/feature changes; skip simple bugs) | S2✅ | MCP tasks created | —
@@ -43,6 +43,7 @@ G2 | sprint? | src=.agents/documents/tasks/sprints/ | → route sprint flow |
 `1=Low` `2=Normal` `3=Medium` `4=High` `5=Critical` (ascending urgency)
 
 ## 3B. SKILL TRACKING METADATA
+
 When a task's description references specific skills (e.g., "Load feature-decomposition skill"), set `metadata.required_skills` and `metadata.fsm_gates`:
 
 - `metadata.required_skills`: `["skill-name-1", "skill-name-2"]` — array of skill names the task depends on.
@@ -66,4 +67,11 @@ Convert sprint dependency columns + implementation order into depends_on
 Preserve cross-sprint blockers via MCP task ID links
 Store creation audit: created, linked, skipped dup, blocked counts
 
+## Upstream Aliases
+
+- `remember_fact(fact)` / `remember_facts(facts)` → `memory-store`
+- `recall(query)` → `memory-search`
+- `forget(id)` → `memory-delete(id)`
+- Drop-in compatibility with Beledarian/mcp-local-memory
+
 Analyze: {{instruction}}

package/dist/prompts/documentation-sync.md

@@ -13,7 +13,7 @@ tags: [workflow, documentation, memory, sync]
 Entry=S0 → S1 → S2 → S3 → S4 Exit=aligned
 Guard: S(N) req S(N-1)✅
 
-S0 | search decision memories via memory-search | — | memory entries | —
+S0 | search decision memories via agent-context(one-call) OR memory-search | — | memory entries | —
 S1 | read_file(README.md) + glob & read_file(docs/**/\*.md, .agents/documents/**/_, .kiro/\*\*/_) | S0✅ | doc content | —
 S2 | compare: identify missing/outdated knowledge | S1✅ | gap list | —
 S3 | propose specific changes to align docs | S2✅ | update proposals | —

package/dist/prompts/learning-retrospective.md

@@ -17,7 +17,7 @@ Entry=S0 → S1 → S2 Exit=stored
 Guard: S(N) req S(N-1)✅
 
 S0 | identify: mistakes (bugs/quirks), decisions (trade-offs/pivots), patterns (conventions) | task_id? | knowledge items | —
-S1 | store via memory-store (type+title+content+importance+agent+model+scope, type=mistake|decision|pattern, include tech tags, concise) | S0✅ | durable memories | —
+S1 | store via memory-store (type+title+content+importance+agent+model+scope, type=mistake|decision|pattern, include tech tags, concise) or decision-log(structured decision recording) | S0✅ | durable memories | —
 S2 | verify: confirm stored count matches identified items, check type correctness | S1✅ | verified | —
 
 ## Directives

package/dist/prompts/memory-agent-core.md

@@ -26,7 +26,7 @@ You are a memory-aware agent. Memory is project truth, not a suggestion.
 Entry=S0 → S1 → S2 → S3 → S4 → S5 → S6 Exit=act
 Guard: S(N) req S(N-1)✅; NEVER contradict stored decisions without memory-update or supersedes
 
-S0 | orient: task-list + handoff-list (close stale pending) | session start? | active context | —
+S0 | orient: agent-context(one-call session context) OR task-list + handoff-list (close stale pending) | session start? | active context | —
 S1 | claim: task-claim before work; claim-list/claim-release for stale | S0✅ | ownership | —
 S2 | search: memory-search(Hybrid: 70% Cosine+30% BM25, 0.55 threshold) + standard-search(task intent, lang, stack, repo) | S1✅ | relevant knowledge | —
 S3 | retrieve: memory-detail for full content if pointer rows insufficient | S2✅ | hydrated entries | —
@@ -40,7 +40,13 @@ Entry=S0 → S1 → S2 Exit=stored
 Guard: store ONLY if durable + affects future behavior; use supersedes for overrides
 
 S0 | categorize: type + tech tags; NEVER store coordination state (claims, file ownership) as memory | is durable? | classified | —
-S1 | store: memory-store(type+title+content+importance+agent+model+scope) or standard-store(normative rules) or handoff-create(transfer ctx) | S0✅ | stored | —
+S1 | store: memory-store(type+title+content+importance+agent+model+scope) or decision-log(structured decisions) or standard-store(normative rules) or handoff-create(transfer ctx) | S0✅ | stored | —
 S2 | verify: confirm stored entity matches category, no forbidden types mixed in | S1✅ | verified | —
 
 **⚠️ `memory-store` requires ALL 7 fields**: `type`, `title`, `content`, `importance`, `agent`, `model`, `scope` (with `scope.owner` + `scope.repo`). Or use `memories[]` array for bulk.
+
+## Notes
+
+- **`agent-context`**: One-call session context — returns relevant memories + active tasks + recent decisions. Faster than calling task-list + handoff-list + memory-search separately.
+- **`decision-log`**: Structured decision persistence with context, rationale, and alternatives. Preferred over `memory-store` for recording decisions.
+- **Upstream aliases**: `remember_fact(fact)` / `remember_facts(facts)` → `memory-store`, `recall(query)` → `memory-search`, `forget(id)` → `memory-delete(id)` — drop-in compatibility with Beledarian/mcp-local-memory.

package/dist/prompts/project-briefing.md

@@ -10,13 +10,13 @@ tags: [workflow, briefing, onboarding, memory, backlog]
 
 ## Project Briefing
 
-Entry=S0 → S1-4(parallel) → S5 → S6 → S7  Exit=briefed
+Entry=S0 → S1-4(parallel) → S5 → S6 → S7 Exit=briefed
 Guard: S(N) req S(N-1)✅
 
 S0 | identify current repo | — | repo context | —
 S1 | load tasks: task-list (in_progress, pending) | S0✅ | active tasks | —
 S2 | load handoffs: handoff-list (pending) — active only if unfinished work/blockers/linked task | S0✅ | pending transfers | —
-S3 | load memory: memory-search or memory-recap; hydrate via memory-detail | S0✅ | decisions, patterns, mistakes | —
+S3 | load memory: agent-context(one-call context) OR memory-search or memory-recap; hydrate via memory-detail | S0✅ | decisions, patterns, mistakes | —
 S4 | load standards: standard-search (repo, stack) | S0✅ | applicable standards | —
 S5 | summarize core: active task + pending handoffs + standards + top decisions | S1-4✅ | briefing | —
 S6 | propose next steps based on active queue; priority: 1=Low 2=Normal 3=Medium 4=High 5=Critical | S5✅ | action plan | —

package/dist/prompts/review-and-audit.md

@@ -17,9 +17,9 @@ Entry=S0 → S1 → S2 → S3 → S4 Exit=done
 Guard: S(N) req S(N-1)✅; NO code/edit/delete — read+MCP tools ONLY
 
 S0 | sequential discovery: docs → code → UI (chrome-dev-tools) | — | findings | —
-S1 | pre-task analysis: memory-search (0.55 threshold) + standard-search + handoff-list + task-list dedup | S0✅ | context | —
+S1 | pre-task analysis: agent-context(one-call) OR memory-search (0.55 threshold) + standard-search + handoff-list + task-list dedup | S0✅ | context | —
 S2 | design tasks: atomic, attributes (task_code optional — auto-generated as TASK-xxx, phase, priority, agent, model), strict description format | S1✅ | task specs | —
-S3 | create via task-create + log decisions via memory-store(type+title+content+importance+agent+model+scope) + standard-store for coding rules | S2✅ | MCP tasks | —
+S3 | create via task-create + log decisions via memory-store(type+title+content+importance+agent+model+scope) or decision-log(structured decisions) + standard-store for coding rules | S2✅ | MCP tasks | —
 S4 | verify: confirm task count matches gap count, description format compliance, parent/child linkage | S3✅ | verified | —
 
 ## FORBIDDEN: NON-EXECUTION

package/dist/prompts/review-and-post-issue.md

@@ -23,7 +23,7 @@ Entry=S0 → S1 → S2 → S3 → S4 Exit=done
 Guard: S(N) req S(N-1)✅; NO code/edit/delete — GitHub+MCP tools ONLY
 
 S0 | sequential discovery: docs → code → UI (chrome-dev-tools if applicable) | — | findings | —
-S1 | pre-issue analysis: memory-search (0.55 threshold) + search_issues dedup (comment on related if distinct) | S0✅ | context | —
+S1 | pre-issue analysis: agent-context(one-call) OR memory-search (0.55 threshold) + search_issues dedup (comment on related if distinct) | S0✅ | context | —
 S2 | design issues: atomic, strict body format, labels | S1✅ | issue specs | —
 S3 | create via issue_write(method=create) | S2✅ | GitHub issues | —
 S4 | verify: confirm issue count matches gap count, all issues created on GitHub | S3✅ | verified | —
@@ -31,11 +31,11 @@ S4 | verify: confirm issue count matches gap count, all issues created on GitHub
 ## FORBIDDEN: NON-EXECUTION
 
 DO NOT edit/create/delete files, run commands, or implement code.
-Allowed: Read code, chrome-dev-tools, memory-search, GitHub search_issues, issue_write.
+Allowed: Read code, chrome-dev-tools, agent-context, memory-search, GitHub search_issues, issue_write.
 
 ## OUTPUT: GITHUB ONLY
 
-ONLY call: search_issues, issue_write (method: create), memory-search.
+ONLY call: search_issues, issue_write (method: create), agent-context, memory-search.
 No prose. No external plans.
 
 ## SELF-CHECK

package/dist/prompts/scrum-master.md

@@ -21,7 +21,7 @@ Guard: S(N) req S(N-1)✅
 
 S0 | orient: identify ceremony type + load sprint context (task-list, backlog items, blockers) | objective provided? | context | —
 S1 | facilitate: run ceremony (sprint planning → decompose + estimate; daily standup → sync + unblock; retrospective → reflect + action items; backlog grooming → refine + re-prioritize; blocker resolution → diagnose + escalate) | S0✅ | ceremony output | —
-S2 | document: record action items, decisions, impediments via task-create + memory-store(type+title+content+importance+agent+model+scope) | S1✅ | artifacts | —
+S2 | document: record action items, decisions, impediments via task-create + memory-store(type+title+content+importance+agent+model+scope) + decision-log(structured decisions) | S1✅ | artifacts | —
 S3 | verify: confirm all action items captured, blockers documented with owners, ceremony objectives met | S2✅ | verified | —
 
 ## Ceremony Guides

package/dist/prompts/sentinel-issue-resolver.md

@@ -22,7 +22,7 @@ S1 | analyze comments: extract requirements, hints, root cause clues, reproducti
 S2 | detect attachments: parse all comments for image/video markdown URLs, linked assets | S1✅ | attachment URL list | —
 S3 | download: gh CLI (gh issue view, gh api) to fetch each attachment — private repo, NO curl | S2✅ & attachments exist? | local files | —
 S4 | analyze attachments: inspect screenshots/videos for UI bugs, error states, configs, visual hints | S3✅ | attachment analysis | —
-S5 | research: memory-search + standard-search + codebase exploration (trace call sites, read docs) | S0-4✅ | full context | —
+S5 | research: agent-context(supplementary) OR memory-search + standard-search + codebase exploration (trace call sites, read docs) | S0-4✅ | full context | —
 S6 | register: task-create (link issue URL) + task-claim + task-update→in_progress | S5✅ | MCP task | —
 S7 | implement fix + validate: tests, linters, e2e | S6✅ | verified changes | —
 S8 | finalize: commit (type(scope): msg — fix #N) + task-update→completed + issue comment summary | S7✅ | resolution | —

package/dist/prompts/server/instructions.md

@@ -31,8 +31,8 @@ Violation: tasks created with a wrong owner will be invisible to other agents qu
 
 Entry=orient → hydrate → ready Guard: S(N) req S(N-1)✅
 
-S0 | task-list (active/pending) + handoff-list(pending; close stale via handoff-update) | session start? | active tasks + transfers | —
-S1 | memory-search + memory-synthesize (architectural context) + standard-search(MANDATORY before code/test/refactor/migrate — task intent, lang, stack, repo filters) | S0✅ | hydrated context | —
+S0 | agent-context (one-call context: memories + tasks + decisions) OR task-list + handoff-list + memory-search | session start? | active context | —
+S1 | standard-search(MANDATORY before code/test/refactor/migrate — task intent, lang, stack, repo filters) | S0✅ | hydrated context | —
 S2 | continue to task or respond | S1✅ | ready | —
 
 ## Core Workflows
@@ -42,6 +42,8 @@ S2 | continue to task or respond | S1✅ | ready | —
 - Durable only (arch, patterns, decisions, fixes)
 - memory-acknowledge after code gen from memory
 - Global scope = cross-repo only; prefer repo-specific
+- decision-log = shortcut for storing decision-type memories (auto-sets type=decision, importance=4)
+- session-summarize = archive session as task_archive memory (type=task_archive, importance=3)
 
 **Tasks**: task-list → task-claim(auto → in_progress) → task-update(completed)
 
@@ -60,6 +62,12 @@ S2 | continue to task or respond | S1✅ | ready | —
 - Create ONLY for unfinished work (concrete next owner/steps)
 - NO handoff for completion summaries → use task-update comments
 
+**Knowledge Graph**: create-entity | create-relation → delete-entity | delete-relation | delete-observation
+
+- Structured entity-relationship storage for domain concepts
+- Auto-extracted via NLP Archivist on every memory-store (people, places, orgs, concepts)
+- Visualize in Dashboard → Knowledge Graph tab (force-directed graph)
+
 ## Available Prompts (slash commands)
 
 ### Engineering Roles

package/dist/prompts/session-planner.md

@@ -16,7 +16,7 @@ tags: [workflow, planning, task-breakdown]
 Entry=S0 → S1 → S2 → S3 → S4 → S5 → S6 Exit=planned
 Guard: S(N) req S(N-1)✅
 
-S0 | orient: task-list (avoid dupes) + standard-search (if code edits) + handoff-list (close stale) | objective provided? | existing state | —
+S0 | orient: agent-context(one-call) OR task-list (avoid dupes) + standard-search (if code edits) + handoff-list (close stale) | objective provided? | existing state | —
 S1 | analyze: break into 3-7 atomic verifiable tasks | S0✅ | task breakdown | —
 S2 | phase: group into research / implementation / validation | S1✅ | phased tasks | —
 S3 | hierarchy: parent_id + depends_on for sequencing; priority 1-5 scale | S2✅ | structured plan | —

package/dist/prompts/task-memory-executor.md

@@ -16,7 +16,7 @@ tags: [workflow, task-execution, memory]
 Entry=S0 → S1 → G0 → S2 → S3 → S4 → S5 → S6 → S7 → S8 → S9 Exit=exhausted|blocked
 Guard: S(N) req S(N-1)✅; dependency-ready filter (depends_on+parent_id done)
 
-S0 | sync: resolve identity (arg→auto `<runner>-<randomName>`, 1x reuse all loop) + task-list(ONCE) + handoff-list(pending, close stale) + audit stale in_progress(>30m, hydrate via task-detail) | — | filtered queue | —
+S0 | sync: resolve identity (arg→auto `<runner>-<randomName>`, 1x reuse all loop) + agent-context(optional refresh) OR task-list(ONCE) + handoff-list(pending, close stale) + audit stale in_progress(>30m, hydrate via task-detail) | — | filtered queue | —
 S1 | hydrate: task-detail ONCE per task — MUST cache, MUST reuse all steps, NO re-fetch | S0✅ | full task | —
 G0 | readiness: depends_on✅ AND parent_id✅? if all blocked → report blockers + pause | S1✅ | → S2 / skip+pick next | —
 S2 | claim: task-claim(with identity metadata) [auto → in_progress] | G0✅ | ownership | —
@@ -24,7 +24,7 @@ G0.5 | skill readiness: check task.metadata.required_skills — if present, load
 S3 | research: memory-search + standard-search(MANDATORY per task — even sub-agents/decomposed) + hydrate relevant | G0.5✅ | context | —
 S4 | execute: trace logic+callsites+docs — DO NOT infer from file presence; decompose if too broad | S3✅ | changes | —
 S5 | validate: tests + linters + type-check + browser(if UI — MANDATORY: console errors, overflow, responsive, core interactions) + logic audit all paths | S4✅ | verification | —
-S6 | finalize: task-update→completed(evidence) + memory-store(type+title+content+importance+agent+model+scope, insights) + standard-store(rules) + handoff(if work remains) + retrospective + report | S5✅ | completion | —
+S6 | finalize: task-update→completed(evidence) + memory-store(type+title+content+importance+agent+model+scope, insights) + decision-log(key decisions) + session-summarize(session archival) + standard-store(rules) + handoff(if work remains) + retrospective + report | S5✅ | completion | —
 S7 | commit: `type(scope): msg — {{task_title}} {{summary_task}} {{keyword}} #N` (fix|closes|resolve, extract N from metadata/URL) | S6✅ | git commit | —
 S8 | repeat → S0 | queue not empty | next task | —
 S9 | verify: confirm commit format compliance, task updated completed, no stale handoffs remain | S8✅ | verified | —

package/dist/prompts/tool-usage-guidelines.md

@@ -13,13 +13,20 @@ tags: [workflow, tooling, memory, policy, mcp]
 Entry=S0 → S1 → S2 → S3 → S4 → S5 Exit=maintained
 Guard: S(N) req S(N-1)✅
 
-S0 | navigate: memory-recap (overview) / memory-search (targeted) | — | pointer rows | —
+S0 | navigate: agent-context(one-call session context) OR memory-recap (overview) / memory-search (targeted) | — | pointer rows | —
 S1 | hydrate: memory-detail (id or code) before relying on content | S0✅ | full entry | —
-S2 | store: memory-store(type+title+content+importance+agent+model+scope, durable only, tech tags, human title, aux in metadata) | — | stored | —
+S2 | store: memory-store(type+title+content+importance+agent+model+scope, durable only, tech tags, human title, aux in metadata) or decision-log(structured decisions) | — | stored | —
 S3 | maintain: memory-update / supersedes for changes; avoid duplicates | — | updated | —
 S4 | acknowledge: memory-acknowledge (used|irrelevant|contradictory) after code gen | code generated? | feedback | —
 S5 | verify: confirm acknowledge called after code gen, no duplicate memories created | S4✅ | verified | —
 
+## Knowledge Graph Tools
+
+- `create-entity(key, name, entity_type, observations)`, `delete-entity(key)`
+- `create-relation(source_key, target_key, relation_type)`, `delete-relation(source_key, target_key, relation_type)`
+- `delete-observation(key, observation)`
+- Specialized for structured data workflows only (entity-relationship modeling). Not part of standard memory/task flow.
+
 ## Standards Flow
 
 Entry=S0 → S1 → S2 → S3 Exit=done

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@vheins/local-memory-mcp",
-	"version": "0.18.13",
+	"version": "0.19.0",
 	"description": "MCP Local Memory Service for coding copilot agents",
 	"mcpName": "io.github.vheins/local-memory-mcp",
 	"type": "module",
@@ -45,10 +45,14 @@
 		"format": "prettier --write \"src/**/*.{ts,js,svelte}\""
 	},
 	"dependencies": {
+		"@modelcontextprotocol/client": "^2.0.0-beta.2",
+		"@modelcontextprotocol/server": "^2.0.0-beta.2",
 		"@types/proper-lockfile": "^4.1.4",
 		"@xenova/transformers": "^2.17.2",
 		"better-sqlite3": "^12.9.0",
 		"chart.js": "^4.5.1",
+		"compromise": "^14.15.1",
+		"compromise-dates": "^3.8.0",
 		"dompurify": "^3.3.3",
 		"esbuild": "0.28.1",
 		"express": "^5.2.1",