memorymaster 3.2.0__tar.gz → 3.2.2__tar.gz

{memorymaster-3.2.0/memorymaster.egg-info → memorymaster-3.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: memorymaster
-Version: 3.2.0
+Version: 3.2.2
 Summary: Production-grade memory reliability system for AI coding agents. Lifecycle-managed claims with citations, conflict detection, steward governance, and MCP integration.
 Author: wolverin0
 License: MIT
@@ -42,9 +42,9 @@ Lifecycle-managed claims with citations, conflict detection, steward governance,
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
-[![Tests](https://img.shields.io/badge/tests-932-green.svg)]()
-[![MCP Tools](https://img.shields.io/badge/MCP%20tools-21-purple.svg)]()
-[![CLI Commands](https://img.shields.io/badge/CLI%20commands-54%2B-orange.svg)]()
+[![Tests](https://img.shields.io/badge/tests-1034-green.svg)]()
+[![MCP Tools](https://img.shields.io/badge/MCP%20tools-22-purple.svg)]()
+[![CLI Commands](https://img.shields.io/badge/CLI%20commands-64-orange.svg)]()
 
 ---
 
@@ -55,9 +55,9 @@ MemoryMaster gives AI coding agents **persistent, verifiable memory** with a ful
 | Metric | Count |
 |--------|-------|
 | Source modules | 35+ (20,000+ lines) |
-| Tests | 932 across 66 test modules |
-| MCP tools | 21 |
-| CLI commands | 54+ |
+| Tests | 1034 across 68 test modules |
+| MCP tools | 22 |
+| CLI commands | 64 |
 | Import connectors | 10+ (Git, Slack, Jira, email, GitHub, conversations) |
 | Utility scripts | 30+ (connectors, benchmarks, drills) |
 
@@ -68,7 +68,7 @@ MemoryMaster gives AI coding agents **persistent, verifiable memory** with a ful
 │                        Agent Runtime                            │
 │  (Claude Code / Codex / any MCP-compatible agent)               │
 └────────────┬────────────────────────────────┬───────────────────┘
-             │ MCP (21 tools)                 │ CLI (50+ commands)
+             │ MCP (22 tools)                 │ CLI (64 commands)
              v                                v
 ┌─────────────────────────────────────────────────────────────────┐
 │                      MemoryMaster Core                          │
@@ -141,13 +141,19 @@ pip install memorymaster
 memorymaster --db memorymaster.db init-db
 
 # Full setup: hooks, MCP, steward cron, Obsidian skills
+memorymaster-setup      # after pip install
+# or, from a cloned repo:
 python scripts/setup-hooks.py
 ```
 
-The setup script configures everything interactively:
+The setup command configures everything interactively:
 - **Recall hook** — injects relevant claims into every Claude Code prompt
-- **Auto-ingest hook** — uses a cheap LLM (Gemini Flash Lite/GPT-4o-mini/Haiku/Ollama) to extract learnings from each session
-- **MCP server** — 21 tools available in all Claude Code & Codex sessions
+- **Classify hook** — regex signal matcher (DECISION/BUG/GOTCHA/CONSTRAINT/ARCHITECTURE/ENVIRONMENT/REFERENCE) that injects routing hints, Spanish + English
+- **Validate-wiki hook** — PostToolUse warning for wiki articles missing frontmatter or wikilinks
+- **SessionStart hook** — injects recent claims + cycle summary + pending candidates at session start
+- **Auto-ingest hook** — uses a cheap LLM (Gemini Flash Lite/GPT-4o-mini/Haiku/Ollama) to extract learnings from each session, with a block-based checkpoint every 15 human messages
+- **PreCompact hook** — forces save to MemoryMaster before Claude Code compacts context (permanent context loss prevention)
+- **MCP server** — 22 tools available in all Claude Code & Codex sessions
 - **Steward cron** — validates and curates claims every 6 hours
 - **CLAUDE.md / AGENTS.md** — appends instructions so Claude and Codex actually use MemoryMaster
 - **Obsidian skills** — read/write/search your vault from Claude Code
@@ -205,7 +211,7 @@ Add to your `.mcp.json` (see [`.mcp.json.example`](.mcp.json.example)):
 }
 ```
 
-**21 MCP tools:** `init_db`, `ingest_claim`, `run_cycle`, `run_steward`, `classify_query`, `query_memory`, `query_for_context`, `list_claims`, `redact_claim_payload`, `pin_claim`, `compact_memory`, `list_events`, `open_dashboard`, `list_steward_proposals`, `resolve_steward_proposal`, `extract_entities`, `entity_stats`, `find_related_claims`, `quality_scores`, `recompute_tiers`, `federated_query`
+**22 MCP tools:** `init_db`, `ingest_claim`, `run_cycle`, `run_steward`, `classify_query`, `query_memory`, `query_for_context`, `list_claims`, `redact_claim_payload`, `pin_claim`, `compact_memory`, `list_events`, `search_verbatim`, `open_dashboard`, `list_steward_proposals`, `resolve_steward_proposal`, `extract_entities`, `entity_stats`, `find_related_claims`, `quality_scores`, `recompute_tiers`, `federated_query`
 
 ## How It All Works (E2E)
 
@@ -251,7 +257,7 @@ YOU SEND A MESSAGE
 |-----------|-------------|-----------|
 | **Recall hook** | Injects relevant claims into every prompt | Every message you send |
 | **Auto-ingest hook** | LLM extracts learnings from transcript | Every time Claude stops |
-| **MCP server** (global) | 21 tools for query/ingest/steward | Always available |
+| **MCP server** (global) | 22 tools for query/ingest/steward | Always available |
 | **CLAUDE.md append** | Instructions for Claude to use MemoryMaster | Read at session start |
 | **AGENTS.md append** | Instructions for Codex to use MemoryMaster | Read at session start |
 | **Steward cron** | Validates, decays, compacts claims | Every 6 hours |
@@ -616,7 +622,7 @@ These are optional but enhance the experience:
 
 | MCP | What it adds | Install |
 |-----|--------------|---------|
-| **memorymaster** | The 21 MCP tools (`ingest_claim`, `query_memory`, `run_cycle`, etc.) | `python scripts/setup-hooks.py` (interactive) |
+| **memorymaster** | The 22 MCP tools (`ingest_claim`, `query_memory`, `run_cycle`, etc.) | `memorymaster-setup` (interactive; or `python scripts/setup-hooks.py` from clone) |
 | **GitNexus** | Code-graph aware impact analysis before edits | See [GitNexus Integration](#gitnexus-integration-code-intelligence) |
 | **Obsidian CLI** | Vault-aware search via the obsidian CLI tool | `npm install -g obsidian-cli` (requires Obsidian 1.12+) |
 | **Qdrant** | Vector search backend for semantic recall | `docker run -p 6333:6333 qdrant/qdrant` |
@@ -624,16 +630,18 @@ These are optional but enhance the experience:
 ### Installation
 
 ```bash
-# 1. Clone and install MemoryMaster
-git clone https://github.com/your-org/memorymaster.git
-cd memorymaster
-pip install -e .
+# Option A — from PyPI (recommended for users)
+pip install memorymaster
+memorymaster-setup               # interactive installer
 
-# 2. Run the interactive installer (sets up MCP, hooks, env vars, steward cron)
-python scripts/setup-hooks.py
+# Option B — from a clone (recommended for contributors)
+git clone https://github.com/wolverin0/memorymaster.git
+cd memorymaster
+pip install -e ".[dev,mcp,security]"
+python scripts/setup-hooks.py    # 3-line shim calling memorymaster.setup_hooks:main
 
-# 3. The installer copies hooks from config-templates/hooks/ to ~/.claude/hooks/
-#    and registers them in ~/.claude/settings.json automatically.
+# Either way, the installer copies hooks from memorymaster/config_templates/hooks/
+# to ~/.claude/hooks/ and registers them in ~/.claude/settings.json automatically.
 
 # 4. Verify
 python -m memorymaster --db memorymaster.db query "test"
@@ -696,7 +704,7 @@ Key config groups:
 # Install with all dev dependencies
 pip install -e ".[dev,mcp,security,embeddings,qdrant]"
 
-# Run tests (932 tests)
+# Run tests (1034 tests)
 pytest tests/ -q
 
 # Lint and format

{memorymaster-3.2.0 → memorymaster-3.2.2}/README.md

@@ -6,9 +6,9 @@ Lifecycle-managed claims with citations, conflict detection, steward governance,
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
-[![Tests](https://img.shields.io/badge/tests-932-green.svg)]()
-[![MCP Tools](https://img.shields.io/badge/MCP%20tools-21-purple.svg)]()
-[![CLI Commands](https://img.shields.io/badge/CLI%20commands-54%2B-orange.svg)]()
+[![Tests](https://img.shields.io/badge/tests-1034-green.svg)]()
+[![MCP Tools](https://img.shields.io/badge/MCP%20tools-22-purple.svg)]()
+[![CLI Commands](https://img.shields.io/badge/CLI%20commands-64-orange.svg)]()
 
 ---
 
@@ -19,9 +19,9 @@ MemoryMaster gives AI coding agents **persistent, verifiable memory** with a ful
 | Metric | Count |
 |--------|-------|
 | Source modules | 35+ (20,000+ lines) |
-| Tests | 932 across 66 test modules |
-| MCP tools | 21 |
-| CLI commands | 54+ |
+| Tests | 1034 across 68 test modules |
+| MCP tools | 22 |
+| CLI commands | 64 |
 | Import connectors | 10+ (Git, Slack, Jira, email, GitHub, conversations) |
 | Utility scripts | 30+ (connectors, benchmarks, drills) |
 
@@ -32,7 +32,7 @@ MemoryMaster gives AI coding agents **persistent, verifiable memory** with a ful
 │                        Agent Runtime                            │
 │  (Claude Code / Codex / any MCP-compatible agent)               │
 └────────────┬────────────────────────────────┬───────────────────┘
-             │ MCP (21 tools)                 │ CLI (50+ commands)
+             │ MCP (22 tools)                 │ CLI (64 commands)
              v                                v
 ┌─────────────────────────────────────────────────────────────────┐
 │                      MemoryMaster Core                          │
@@ -105,13 +105,19 @@ pip install memorymaster
 memorymaster --db memorymaster.db init-db
 
 # Full setup: hooks, MCP, steward cron, Obsidian skills
+memorymaster-setup      # after pip install
+# or, from a cloned repo:
 python scripts/setup-hooks.py
 ```
 
-The setup script configures everything interactively:
+The setup command configures everything interactively:
 - **Recall hook** — injects relevant claims into every Claude Code prompt
-- **Auto-ingest hook** — uses a cheap LLM (Gemini Flash Lite/GPT-4o-mini/Haiku/Ollama) to extract learnings from each session
-- **MCP server** — 21 tools available in all Claude Code & Codex sessions
+- **Classify hook** — regex signal matcher (DECISION/BUG/GOTCHA/CONSTRAINT/ARCHITECTURE/ENVIRONMENT/REFERENCE) that injects routing hints, Spanish + English
+- **Validate-wiki hook** — PostToolUse warning for wiki articles missing frontmatter or wikilinks
+- **SessionStart hook** — injects recent claims + cycle summary + pending candidates at session start
+- **Auto-ingest hook** — uses a cheap LLM (Gemini Flash Lite/GPT-4o-mini/Haiku/Ollama) to extract learnings from each session, with a block-based checkpoint every 15 human messages
+- **PreCompact hook** — forces save to MemoryMaster before Claude Code compacts context (permanent context loss prevention)
+- **MCP server** — 22 tools available in all Claude Code & Codex sessions
 - **Steward cron** — validates and curates claims every 6 hours
 - **CLAUDE.md / AGENTS.md** — appends instructions so Claude and Codex actually use MemoryMaster
 - **Obsidian skills** — read/write/search your vault from Claude Code
@@ -169,7 +175,7 @@ Add to your `.mcp.json` (see [`.mcp.json.example`](.mcp.json.example)):
 }
 ```
 
-**21 MCP tools:** `init_db`, `ingest_claim`, `run_cycle`, `run_steward`, `classify_query`, `query_memory`, `query_for_context`, `list_claims`, `redact_claim_payload`, `pin_claim`, `compact_memory`, `list_events`, `open_dashboard`, `list_steward_proposals`, `resolve_steward_proposal`, `extract_entities`, `entity_stats`, `find_related_claims`, `quality_scores`, `recompute_tiers`, `federated_query`
+**22 MCP tools:** `init_db`, `ingest_claim`, `run_cycle`, `run_steward`, `classify_query`, `query_memory`, `query_for_context`, `list_claims`, `redact_claim_payload`, `pin_claim`, `compact_memory`, `list_events`, `search_verbatim`, `open_dashboard`, `list_steward_proposals`, `resolve_steward_proposal`, `extract_entities`, `entity_stats`, `find_related_claims`, `quality_scores`, `recompute_tiers`, `federated_query`
 
 ## How It All Works (E2E)
 
@@ -215,7 +221,7 @@ YOU SEND A MESSAGE
 |-----------|-------------|-----------|
 | **Recall hook** | Injects relevant claims into every prompt | Every message you send |
 | **Auto-ingest hook** | LLM extracts learnings from transcript | Every time Claude stops |
-| **MCP server** (global) | 21 tools for query/ingest/steward | Always available |
+| **MCP server** (global) | 22 tools for query/ingest/steward | Always available |
 | **CLAUDE.md append** | Instructions for Claude to use MemoryMaster | Read at session start |
 | **AGENTS.md append** | Instructions for Codex to use MemoryMaster | Read at session start |
 | **Steward cron** | Validates, decays, compacts claims | Every 6 hours |
@@ -580,7 +586,7 @@ These are optional but enhance the experience:
 
 | MCP | What it adds | Install |
 |-----|--------------|---------|
-| **memorymaster** | The 21 MCP tools (`ingest_claim`, `query_memory`, `run_cycle`, etc.) | `python scripts/setup-hooks.py` (interactive) |
+| **memorymaster** | The 22 MCP tools (`ingest_claim`, `query_memory`, `run_cycle`, etc.) | `memorymaster-setup` (interactive; or `python scripts/setup-hooks.py` from clone) |
 | **GitNexus** | Code-graph aware impact analysis before edits | See [GitNexus Integration](#gitnexus-integration-code-intelligence) |
 | **Obsidian CLI** | Vault-aware search via the obsidian CLI tool | `npm install -g obsidian-cli` (requires Obsidian 1.12+) |
 | **Qdrant** | Vector search backend for semantic recall | `docker run -p 6333:6333 qdrant/qdrant` |
@@ -588,16 +594,18 @@ These are optional but enhance the experience:
 ### Installation
 
 ```bash
-# 1. Clone and install MemoryMaster
-git clone https://github.com/your-org/memorymaster.git
-cd memorymaster
-pip install -e .
+# Option A — from PyPI (recommended for users)
+pip install memorymaster
+memorymaster-setup               # interactive installer
 
-# 2. Run the interactive installer (sets up MCP, hooks, env vars, steward cron)
-python scripts/setup-hooks.py
+# Option B — from a clone (recommended for contributors)
+git clone https://github.com/wolverin0/memorymaster.git
+cd memorymaster
+pip install -e ".[dev,mcp,security]"
+python scripts/setup-hooks.py    # 3-line shim calling memorymaster.setup_hooks:main
 
-# 3. The installer copies hooks from config-templates/hooks/ to ~/.claude/hooks/
-#    and registers them in ~/.claude/settings.json automatically.
+# Either way, the installer copies hooks from memorymaster/config_templates/hooks/
+# to ~/.claude/hooks/ and registers them in ~/.claude/settings.json automatically.
 
 # 4. Verify
 python -m memorymaster --db memorymaster.db query "test"
@@ -660,7 +668,7 @@ Key config groups:
 # Install with all dev dependencies
 pip install -e ".[dev,mcp,security,embeddings,qdrant]"
 
-# Run tests (932 tests)
+# Run tests (1034 tests)
 pytest tests/ -q
 
 # Lint and format

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/__init__.py

@@ -2,4 +2,4 @@
 
 __all__ = ["__version__"]
 
-__version__ = "3.2.0"
+__version__ = "3.2.2"

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/_storage_lifecycle.py

@@ -6,24 +6,18 @@ and `self.db_path`. Do not instantiate directly.
 """
 from __future__ import annotations
 
-import contextlib
-import hashlib
 import json
 import logging
 import sqlite3
 from datetime import datetime, timedelta, timezone
-from pathlib import Path
 
 from memorymaster.embeddings import EmbeddingProvider, cosine_similarity
 from memorymaster.models import (
     CLAIM_LINK_TYPES,
     CLAIM_STATUSES,
     STATUS_TRANSITION_EVENT_TYPES,
-    Citation,
-    CitationInput,
     Claim,
     ClaimLink,
-    Event,
     validate_event_payload,
     validate_event_type,
     validate_transition_event_type,
@@ -33,12 +27,7 @@ logger = logging.getLogger(__name__)
 
 from memorymaster._storage_shared import (
     EVENT_HASH_ALGO,
-    HUMAN_ID_PREFIX,
-    SQLITE_CONFIRMED_TUPLE_GUARD_TRIGGERS,
-    SQLITE_EVENTS_APPEND_ONLY_TRIGGERS,
     ConcurrentModificationError,
-    generate_human_id_hash,
-    generate_top_level_human_id,
     utc_now,
 )
 

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/_storage_read.py

@@ -6,41 +6,19 @@ and `self.db_path`. Do not instantiate directly.
 """
 from __future__ import annotations
 
-import contextlib
-import hashlib
-import json
 import logging
 import sqlite3
 from datetime import datetime, timedelta, timezone
-from pathlib import Path
 
-from memorymaster.embeddings import EmbeddingProvider, cosine_similarity
 from memorymaster.models import (
-    CLAIM_LINK_TYPES,
-    CLAIM_STATUSES,
-    STATUS_TRANSITION_EVENT_TYPES,
     Citation,
-    CitationInput,
     Claim,
     ClaimLink,
     Event,
-    validate_event_payload,
-    validate_event_type,
-    validate_transition_event_type,
 )
 
 logger = logging.getLogger(__name__)
 
-from memorymaster._storage_shared import (
-    EVENT_HASH_ALGO,
-    HUMAN_ID_PREFIX,
-    SQLITE_CONFIRMED_TUPLE_GUARD_TRIGGERS,
-    SQLITE_EVENTS_APPEND_ONLY_TRIGGERS,
-    ConcurrentModificationError,
-    generate_human_id_hash,
-    generate_top_level_human_id,
-    utc_now,
-)
 
 
 class _ReadMixin:

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/_storage_schema.py

@@ -6,40 +6,19 @@ and `self.db_path`. Do not instantiate directly.
 """
 from __future__ import annotations
 
-import contextlib
 import hashlib
 import json
 import logging
 import sqlite3
-from datetime import datetime, timedelta, timezone
-from pathlib import Path
-
-from memorymaster.embeddings import EmbeddingProvider, cosine_similarity
-from memorymaster.models import (
-    CLAIM_LINK_TYPES,
-    CLAIM_STATUSES,
-    STATUS_TRANSITION_EVENT_TYPES,
-    Citation,
-    CitationInput,
-    Claim,
-    ClaimLink,
-    Event,
-    validate_event_payload,
-    validate_event_type,
-    validate_transition_event_type,
-)
+
 
 logger = logging.getLogger(__name__)
 
 from memorymaster._storage_shared import (
     EVENT_HASH_ALGO,
-    HUMAN_ID_PREFIX,
     SQLITE_CONFIRMED_TUPLE_GUARD_TRIGGERS,
     SQLITE_EVENTS_APPEND_ONLY_TRIGGERS,
-    ConcurrentModificationError,
-    generate_human_id_hash,
     generate_top_level_human_id,
-    utc_now,
 )
 
 

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/_storage_write_claims.py

@@ -6,39 +6,19 @@ and `self.db_path`. Do not instantiate directly.
 """
 from __future__ import annotations
 
-import contextlib
-import hashlib
 import json
 import logging
 import sqlite3
-from datetime import datetime, timedelta, timezone
-from pathlib import Path
 
-from memorymaster.embeddings import EmbeddingProvider, cosine_similarity
 from memorymaster.models import (
-    CLAIM_LINK_TYPES,
-    CLAIM_STATUSES,
-    STATUS_TRANSITION_EVENT_TYPES,
-    Citation,
     CitationInput,
     Claim,
-    ClaimLink,
-    Event,
     validate_event_payload,
-    validate_event_type,
-    validate_transition_event_type,
 )
 
 logger = logging.getLogger(__name__)
 
 from memorymaster._storage_shared import (
-    EVENT_HASH_ALGO,
-    HUMAN_ID_PREFIX,
-    SQLITE_CONFIRMED_TUPLE_GUARD_TRIGGERS,
-    SQLITE_EVENTS_APPEND_ONLY_TRIGGERS,
-    ConcurrentModificationError,
-    generate_human_id_hash,
-    generate_top_level_human_id,
     utc_now,
 )
 

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/cli.py

@@ -3,19 +3,18 @@ from __future__ import annotations
 import argparse
 from pathlib import Path
 
-from memorymaster.cli_helpers import (
+from memorymaster.cli_helpers import (  # noqa: F401 — re-export for backward compat with tests/external callers
     STEALTH_DB_NAME,
     _add_cycle_policy_args,
     _json_error,
-    _resolve_claim_id,  # re-export for backward-compat with tests/external callers
+    _resolve_claim_id,
     _resolve_db_path,
     _stealth_active,
     parse_citation,
     parse_scope_allowlist,
 )
 from memorymaster.context_optimizer import OUTPUT_FORMATS
-from memorymaster.models import CLAIM_LINK_TYPES, CLAIM_STATUSES, CitationInput, VOLATILITY_LEVELS
-from memorymaster.policy import POLICY_MODES
+from memorymaster.models import CLAIM_LINK_TYPES, CLAIM_STATUSES, VOLATILITY_LEVELS
 from memorymaster.retrieval import RETRIEVAL_MODES
 from memorymaster.service import MemoryService
 

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/cli_handlers_basic.py

@@ -9,14 +9,13 @@ from __future__ import annotations
 import argparse
 import json
 import os
-from dataclasses import asdict, is_dataclass
+from dataclasses import asdict
 from pathlib import Path
 import time
 
 from memorymaster.cli_helpers import (
     STEALTH_DB_NAME,
     _SCORE_KEYS,
-    _add_cycle_policy_args,
     _claim_to_dict,
     _event_to_timeline_entry,
     _json_default,
@@ -24,19 +23,14 @@ from memorymaster.cli_helpers import (
     _json_error,
     _print_claim_brief,
     _resolve_claim_id,
-    _resolve_db_path,
+    _score_str_from_payload,
     _stealth_active,
     parse_citation,
     parse_scope_allowlist,
     print_claim,
 )
-from memorymaster.context_optimizer import OUTPUT_FORMATS
-from memorymaster.models import CLAIM_LINK_TYPES, CLAIM_STATUSES, CitationInput, VOLATILITY_LEVELS
-from memorymaster.policy import POLICY_MODES
-from memorymaster.retrieval import RETRIEVAL_MODES
 from memorymaster.scheduler import run_daemon
 from memorymaster.security import resolve_allow_sensitive_access
-from memorymaster.service import MemoryService
 
 
 def _handle_create_snapshot(args: argparse.Namespace, db_resolved: Path) -> int:

{memorymaster-3.2.0 → memorymaster-3.2.2}/memorymaster/cli_handlers_curation.py

@@ -6,8 +6,6 @@ to its handler. cli.main() imports this dict and dispatches to it.
 from __future__ import annotations
 
 import argparse
-import json
-import os
 from pathlib import Path
 import time
 
@@ -43,12 +41,13 @@ from memorymaster.cli_handlers_basic import (
     _handle_steward_proposals,
 )
 from memorymaster.cli_helpers import (
+    _SCORE_KEYS,
     _claim_to_dict,
     _json_envelope,
     _json_error,
-    _resolve_claim_id,
+    print_claim,
 )
-from memorymaster.service import MemoryService
+from memorymaster.models import CitationInput
 
 
 def _handle_export_vault(args: argparse.Namespace, service, parser: argparse.ArgumentParser, effective_db: str) -> int:
@@ -571,9 +570,15 @@ def _handle_daily_note(args, service, parser, effective_db) -> int:
 
 def _handle_ghost_notes(args, service, parser, effective_db) -> int:
     from memorymaster.daily_notes import find_ghost_notes
+    t0 = time.perf_counter()
     ghosts = find_ghost_notes(str(effective_db))
+    elapsed_ms = (time.perf_counter() - t0) * 1000
     if args.json_output:
-        print(_json_envelope({"ghost_notes": ghosts, "count": len(ghosts)}))
+        print(_json_envelope(
+            {"ghost_notes": ghosts, "count": len(ghosts)},
+            total=len(ghosts),
+            query_ms=elapsed_ms,
+        ))
     else:
         if not ghosts:
             print("No ghost notes found (all queried topics have sufficient claims)")

memorymaster-3.2.2/memorymaster/config_templates/claude-md-append.md

@@ -0,0 +1,26 @@
+## MemoryMaster (Cross-Session Memory) — MANDATORY
+
+You have access to the `memorymaster` MCP server. It stores facts, decisions, and constraints across ALL sessions and providers (Claude, Codex, Gemini).
+
+### READ before assuming
+Before making architectural decisions or debugging unfamiliar code, call `mcp__memorymaster__query_memory` with the relevant topic. A previous session may have already solved this.
+
+### WRITE when you learn something non-obvious
+After completing a task where you discovered something that would save future sessions time, call `mcp__memorymaster__ingest_claim`:
+
+**Always ingest:** bug root causes, architectural decisions, environment gotchas, integration patterns, constraints ("never do X because Y")
+
+**Never ingest:** credentials, API keys, tokens, private IPs, personal paths, code snippets, routine actions
+
+```
+mcp__memorymaster__ingest_claim({
+  "text": "One-line factual description",
+  "claim_type": "fact|decision|constraint",
+  "subject": "entity",
+  "predicate": "aspect",
+  "object_value": "value",
+  "scope": "project:<project-name>"
+})
+```
+
+This is NOT optional. If you fixed a non-trivial bug, made an architecture decision, or discovered a gotcha — ingest it.

memorymaster-3.2.2/memorymaster/config_templates/codex-agents-md-append.md

@@ -0,0 +1,35 @@
+## MemoryMaster (Cross-Session Memory)
+
+You have access to the `memorymaster` MCP server with 21 tools for persistent memory across sessions and providers.
+
+### MUST DO — Query before assumptions
+Before making architectural decisions, call `query_memory` or `query_for_context` with the relevant topic. Another session (Claude, Codex, or Gemini) may have already solved this or made a decision.
+
+### MUST DO — Ingest non-obvious learnings
+When you discover something that would save future sessions time, call `ingest_claim`:
+
+**Always ingest:**
+- Decisions made ("we chose X over Y because Z")
+- Bug root causes ("the 403 was caused by missing RLS policy")
+- Environment gotchas ("Docker build crashes the VM, use docker pull instead")
+- Integration patterns ("MercadoPago webhooks need /api/mp/webhook endpoint")
+- Architecture constraints ("max 800 lines per file, immutable patterns only")
+
+**Never ingest:**
+- Credentials, API keys, tokens, passwords
+- Private IPs or personal paths
+- Code snippets (the code itself is in git)
+- Ephemeral task state ("currently working on X")
+
+### Ingest format
+```
+ingest_claim({
+  "text": "Clear one-line description of the fact/decision",
+  "claim_type": "fact" | "decision" | "constraint",
+  "subject": "the entity this is about",
+  "predicate": "what aspect",
+  "object_value": "the value",
+  "scope": "project:<project-name>",
+  "source": "codex-session"
+})
+```

memorymaster-3.2.2/memorymaster/config_templates/hooks/memorymaster-precompact.py

@@ -0,0 +1,35 @@
+"""PreCompact hook: ALWAYS block before compaction, force save everything.
+
+When Claude Code is about to compress context to free up the context window,
+this hook blocks and forces Claude to save ALL important context to MemoryMaster
+before it's lost. This is the safety net — compaction = permanent context loss.
+"""
+import json
+import os
+import sys
+from datetime import datetime
+
+STATE_DIR = os.path.join(os.path.expanduser("~"), ".memorymaster", "hook_state")
+os.makedirs(STATE_DIR, exist_ok=True)
+
+
+def main():
+    try:
+        data = json.loads(sys.stdin.read() or "{}")
+    except Exception:
+        data = {}
+
+    session_id = data.get("session_id", "unknown")
+
+    with open(os.path.join(STATE_DIR, "hook.log"), "a") as f:
+        f.write(f"[{datetime.now().strftime('%H:%M:%S')}] PRE-COMPACT triggered for {session_id}\n")
+
+    # ALWAYS block — compaction means permanent context loss
+    sys.stdout.write(json.dumps({
+        "decision": "block",
+        "reason": "COMPACTION IMMINENT — context will be permanently lost. Before compaction, save ALL important context to MemoryMaster using mcp__memorymaster__ingest_claim. Save: every decision made, every bug root cause, every architecture choice, every gotcha discovered, every constraint learned. Set source_agent to 'claude-session'. Be thorough — after compaction you will NOT remember any of this. NEVER ingest credentials, IPs, or tokens. After saving everything, allow compaction to proceed."
+    }))
+
+
+if __name__ == "__main__":
+    main()