Skip to content

An always-on second brain you talk to. Voice notes in Telegram → typed, linked knowledge in your Obsidian vault. Runs 24/7 on the Claude subscription you already have.

docs: update vault CLAUDE.md for v2 with agent-memory bootstrap

smixs6862064b4c0f
diff --git a/vault/.claude/CLAUDE.md b/vault/.claude/CLAUDE.md
index 8f9e4b0..5c2271d 100644
--- a/vault/.claude/CLAUDE.md
+++ b/vault/.claude/CLAUDE.md
@@ -7,9 +7,11 @@ Voice-first personal assistant for capturing thoughts and managing tasks via Tel
**Before doing anything else, read these files in order:**
1. `vault/MEMORY.md` — curated long-term memory (preferences, decisions, context)
-2. `vault/daily/YYYY-MM-DD.md` — today's entries
-3. `vault/daily/YYYY-MM-DD.md` — yesterday's entries (for continuity)
-4. `vault/goals/3-weekly.md` — this week's ONE Big Thing
+2. `vault/.memory-config.json` — memory decay configuration
+3. `vault/daily/YYYY-MM-DD.md` — today's entries
+4. `vault/daily/YYYY-MM-DD.md` — yesterday's entries (for continuity)
+5. `vault/goals/3-weekly.md` — this week's ONE Big Thing
+6. `vault/.session/handoff.md` — previous session context (if exists)
**Don't ask permission, just do it.** This ensures context continuity across sessions.
@@ -33,6 +35,13 @@ Session summary: [what was discussed/decided/created]
- Important fact learned
- Active context changed significantly
+**Update `vault/.session/handoff.md`:**
+- Last Session: what was done
+- Key Decisions: if any
+- In Progress: unfinished work
+- Next Steps: what to do next
+- Observations: friction signals, patterns, ideas (type: `[friction]`, `[pattern]`, `[idea]`)
+
---
## Mission
@@ -48,6 +57,34 @@ Help user stay aligned with goals, capture valuable insights, and maintain clari
| `thoughts/` | Processed notes by category |
| `MOC/` | Maps of Content indexes |
| `attachments/` | Photos by date |
+| `business/` | Business data (CRM, network, events) |
+| `projects/` | Side projects (clients, leads) |
+
+## Business Context
+
+**Entry point:** `business/_index.md`
+
+```
+business/
+├── _index.md ← Start here (stats, overview)
+├── crm/ ← Client records (companies + deals in one file)
+├── network/ ← Company structure, partners
+└── events/ ← Events, conferences
+```
+
+Search: `business/crm/{kebab-case}.md` (e.g. `acme-corp.md`, `client-b.md`)
+
+## Projects Context
+
+**Entry point:** `projects/_index.md`
+
+```
+projects/
+├── _index.md ← Start here
+├── clients/ ← Project clients
+├── leads/ ← Leads
+└── projects/ ← Active projects
+```
## Current Focus
@@ -58,7 +95,7 @@ See [[goals/2-monthly]] for monthly priorities.
```
goals/0-vision-3y.md → 3-year vision by life areas
-goals/1-yearly-2025.md → Annual goals + quarterly breakdown
+goals/1-yearly-YYYY.md → Annual goals + quarterly breakdown
goals/2-monthly.md → Current month's top 3 priorities
goals/3-weekly.md → This week's focus + ONE Big Thing
```
@@ -76,21 +113,87 @@ Types: `[voice]`, `[text]`, `[forward from: Name]`, `[photo]`
Run daily processing via `/process` command or automatically at 21:00.
-### Process Flow:
-1. Read goals/ → understand priorities
-2. Check Todoist → know workload
-3. Read daily/ → classify entries
-4. Create tasks → aligned with goals
-5. Save thoughts → build [[links]]
-6. Generate HTML report → send to Telegram
+### 3-Phase Pipeline:
+1. **CAPTURE** — Read daily entries → classify → JSON
+2. **EXECUTE** — Create Todoist tasks, save thoughts, update CRM → JSON
+3. **REFLECT** — Generate HTML report, update MEMORY, record observations
+
+Each phase = fresh Claude context for better quality.
-## Available Skills
+## Card Template (agent-memory)
+
+**Skill:** `.claude/skills/agent-memory/SKILL.md`
+
+All new vault cards follow the agent-memory template:
+
+```yaml
+---
+type: crm|lead|contact|project|personal|note
+description: >-
+ One line — what a searcher will see in results
+tags: [tag1, tag2] # 2-5 tags, lowercase
+status: active|draft|pending|done|inactive
+industry: FMCG # for CRM/leads
+region: US # ISO codes
+created: YYYY-MM-DD
+updated: YYYY-MM-DD
+# Auto fields (don't edit manually):
+last_accessed: YYYY-MM-DD
+relevance: 0.85
+tier: active
+---
+```
+
+**Rules:**
+- `description` — REQUIRED. Write as a search snippet, NOT "contact" or "crm"
+- `tags` — REQUIRED. 2-5 tags, lowercase, hyphen-separated
+- `status` ≠ `tier`: status = business status, tier = memory (automatic)
+- One fact = one place (DRY). References via [[wikilinks]]
+- Decay engine: `uv run .claude/skills/agent-memory/scripts/memory-engine.py decay .`
+
+## Skills & References
| Skill | Purpose |
|-------|---------|
-| `dbrain-processor` | Main daily processing |
-| `todoist-ai` | Task management via MCP |
+| `dbrain-processor` | Main daily processing (3-phase pipeline) |
| `graph-builder` | Vault link analysis and building |
+| `vault-health` | Health scoring, MOC generation, link repair |
+| `agent-memory` | Card template, decay engine, tiered search |
+| `todoist-ai` | Todoist task management via MCP |
+
+- **Processing:** `.claude/skills/dbrain-processor/SKILL.md`
+- **Graph Builder:** `.claude/skills/graph-builder/SKILL.md`
+- **Vault Health:** `.claude/skills/vault-health/SKILL.md`
+- **Agent Memory:** `.claude/skills/agent-memory/SKILL.md`
+- **Todoist:** `.claude/skills/todoist-ai/SKILL.md`
+- **Rules:** `.claude/rules/` (daily, thoughts, goals, obsidian-markdown, weekly-reflection)
+- **Docs:** `.claude/docs/`
+
+## Graph Builder
+
+**Purpose:** Analysis and maintenance of vault link structure.
+
+**Architecture:**
+1. `scripts/analyze.py` — deterministic vault traversal
+2. `scripts/add_links.py` — batch link addition
+3. Agent — semantic links for orphan files
+
+**Usage:**
+```bash
+# Analyze vault
+uv run vault/.claude/skills/graph-builder/scripts/analyze.py
+
+# Result
+vault/.graph/vault-graph.json # JSON graph with stats
+vault/.graph/report.md # Human-readable report
+```
+
+**Domains:**
+| Domain | Path | Hub |
+|--------|------|-----|
+| Personal | thoughts/, goals/, daily/ | MEMORY.md |
+| Business | business/crm/, business/network/ | business/_index.md |
+| Projects | projects/clients/, projects/leads/ | projects/_index.md |
## Available Agents
@@ -108,33 +211,8 @@ See `.claude/rules/` for format requirements:
- `thoughts-format.md` — thought notes format
- `goals-format.md` — goals format
- `telegram-report.md` — HTML report format
-
-## MCP Servers
-
-- `todoist` — Task management (add, find, complete tasks)
-- `filesystem` — Vault file access
-
-## CRITICAL: Tool Usage Policy
-
-**ТЫ ОБЯЗАН ВЫЗЫВАТЬ MCP TOOLS НАПРЯМУЮ.**
-
-Не существует ситуации, когда MCP tools "недоступны". Если ты получил эту инструкцию — у тебя есть доступ к:
-
-- `mcp__todoist__*` — все Todoist операции
-- File read/write — все файловые операции
-
-ЗАПРЕЩЁННЫЕ ПАТТЕРНЫ (НИКОГДА не делай это):
-- "MCP недоступен в текущем режиме"
-- "Не удалось добавить автоматически"
-- "Добавь вручную: ..."
-- Любые инструкции для ручного выполнения
-
-ПРАВИЛЬНЫЙ ПАТТЕРН:
-1. Вызвать mcp__todoist__add-tasks tool
-2. Получить результат (успех или ошибка)
-3. Включить результат в HTML отчёт
-
-При ошибке — показать ТОЧНУЮ ошибку от tool, не придумывать отговорки.
+- `obsidian-markdown.md` — Obsidian syntax rules
+- `weekly-reflection.md` — weekly reflection template
## Report Format
@@ -154,59 +232,10 @@ Reports use Telegram HTML:
| `/organize` | Organize vault |
| `/graph` | Analyze vault links |
-## /do Command Context
-
-When invoked via /do, Claude receives arbitrary user requests. Common patterns:
-
-**Task Management:**
-- "перенеси просроченные задачи на понедельник"
-- "покажи задачи на сегодня"
-- "добавь задачу: позвонить клиенту"
-- "что срочного на этой неделе?"
-
-**Vault Queries:**
-- "найди заметки про AI"
-- "что я записал сегодня?"
-- "покажи итоги недели"
-
-**Combined:**
-- "создай задачу из первой записи сегодня"
-- "перенеси всё с сегодня на завтра"
-
-## MCP Tools Available
-
-**Todoist (mcp__todoist__*):**
-- `add-tasks` — создать задачи
-- `find-tasks` — найти задачи по тексту
-- `find-tasks-by-date` — задачи за период
-- `update-tasks` — изменить задачи
-- `complete-tasks` — завершить задачи
-- `user-info` — информация о пользователе
-
-**Filesystem:**
-- Read/write vault files
-- Access daily/, goals/, thoughts/
-
## Customization
For personal overrides: create `CLAUDE.local.md`
-## Graph Builder
-
-Analyze and maintain vault link structure. Use `/graph` command or invoke `graph-builder` skill.
-
-**Commands:**
-- `/graph analyze` — Full vault statistics
-- `/graph orphans` — List unconnected notes
-- `/graph suggest` — Get link suggestions
-- `/graph add` — Apply suggested links
-
-**Scripts:**
-- `uv run .claude/skills/graph-builder/scripts/analyze.py` — Graph analysis
-- `uv run .claude/skills/graph-builder/scripts/add_links.py` — Link suggestions
-
-See `skills/graph-builder/` for full documentation.
-
## Learnings (from experience)
1. **Don't rewrite working code** without reason (KISS, DRY, YAGNI)
@@ -217,5 +246,4 @@ See `skills/graph-builder/` for full documentation.
---
-*System Version: 2.3*
-*Updated: 2026-02-01*
+*System Version: 3.0*
diff --git a/vault/.claude/docs/notebooklm-quick-start.md b/vault/.claude/docs/notebooklm-quick-start.md
new file mode 100644
index 0000000..932c9f8
--- /dev/null
+++ b/vault/.claude/docs/notebooklm-quick-start.md
@@ -0,0 +1,280 @@
+---
+type: note
+title: NotebookLM MCP CLI - Quick Start
+---
+# NotebookLM MCP CLI - Quick Start
+
+> **Установлено:** 1 февраля 2026
+> **Версия:** 0.2.10
+> **GitHub:** https://github.com/jacob-bd/notebooklm-mcp-cli
+
+---
+
+## Зачем нужен NotebookLM
+
+**Для обработки больших массивов данных:**
+- Клиентские исследования и анализ
+- Подготовка презентаций и брифингов
+- Генерация подкастов из документации
+- Создание обучающих материалов (флэшкарты, майндмэпы)
+- Анализ конкурентов
+
+---
+
+## 🚀 Быстрый старт
+
+### 1. Установка
+
+```bash
+uv tool install notebooklm-mcp-cli
+```
+
+**Установлены компоненты:**
+- `nlm` — CLI команда
+- `notebooklm-mcp` — MCP сервер (уже в конфиге)
+- `notebooklm-mcp-auth` — аутентификация
+
+### 2. Аутентификация
+
+```bash
+notebooklm-mcp-auth --file
+```
+
+**Инструкция:**
+1. Открой Chrome → https://notebooklm.google.com
+2. Открой DevTools (Cmd+Option+I)
+3. Вкладка **Network** → фильтр: `batchexecute`
+4. Кликни на любой ноутбук (появится запрос)
+5. Найди **Request Headers** → строку `cookie:`
+6. Скопируй значение (правый клик → Copy value)
+7. Создай файл `~/cookies.txt` и вставь туда
+8. Запусти: `notebooklm-mcp-auth --file ~/cookies.txt`
+
+**Проверка:**
+```bash
+nlm list notebook
+```
+
+---
+
+## 📚 Основные команды
+
+### Управление ноутбуками
+
+```bash
+# Список всех ноутбуков
+nlm list notebook
+
+# Создать ноутбук
+nlm create notebook "Название"
+
+# Детали ноутбука
+nlm get notebook "Название"
+
+# Удалить ноутбук
+nlm delete notebook "Название"
+
+# Переименовать
+nlm rename notebook "Старое" --new-name "Новое"
+```
+
+### Добавление источников
+
+```bash
+# URL
+nlm add source --url https://example.com --notebook "My Research"
+
+# YouTube видео
+nlm add source --youtube https://youtube.com/watch?v=VIDEO_ID --notebook "My Research"
+
+# Текст
+nlm add source --text "Ваш текст здесь" --notebook "My Research"
+
+# Google Drive файл
+nlm add source --drive-file "FILE_ID" --notebook "My Research"
+
+# Список источников
+nlm list source --notebook "My Research"
+```
+
+### Запросы и анализ
+
+```bash
+# Задать вопрос к ноутбуку
+nlm query "В чём основные инсайты?" --notebook "My Research"
+
+# Получить описание
+nlm describe notebook "My Research"
+
+# Получить raw контент из источника
+nlm content source "SOURCE_ID"
+```
+
+### Генерация контента
+
+```bash
+# Audio overview (подкаст)
+nlm audio create --notebook "My Research"
+
+# Video overview
+nlm video create --notebook "My Research"
+
+# Slides (презентация)
+nlm slides create --notebook "My Research"
+
+# Mind map
+nlm mindmap create --notebook "My Research"
+
+# Infographic
+nlm infographic create --notebook "My Research"
+
+# Flashcards
+nlm flashcards create --notebook "My Research"
+
+# Briefing document
+nlm report create --notebook "My Research"
+
+# Quiz
+nlm quiz create --notebook "My Research"
+```
+
+### Синхронизация
+
+```bash
+# Синхронизировать Google Drive источники
+nlm sync --notebook "My Research"
+
+# Список устаревших источников
+nlm stale --notebook "My Research"
+```
+
+### Шеринг
+
+```bash
+# Сделать ноутбук публичным
+nlm share notebook "My Research" --public
+
+# Поделиться с конкретным пользователем
+nlm share notebook "My Research" --email [email protected]
+```
+
+---
+
+## 🔧 Использование через MCP
+
+### Через mcp-cli (рекомендуется)
+
+```bash
+# Список инструментов
+mcp-cli info notebooklm
+
+# Создать ноутбук
+mcp-cli call notebooklm create_notebook '{"title": "Research"}'
+
+# Добавить источник
+mcp-cli call notebooklm add_source '{
+ "notebook_id": "...",
+ "url": "https://example.com"
+}'
+
+# Запрос
+mcp-cli call notebooklm query '{
+ "notebook_id": "...",
+ "query": "What are the main points?"
+}'
+
+# Генерация подкаста
+mcp-cli call notebooklm create_audio '{
+ "notebook_id": "..."
+}'
+```
+
+### Из Claude Code
+
+```markdown
+Используй NotebookLM для анализа этих статей:
+- https://example.com/article1
+- https://example.com/article2
+
+Создай ноутбук "Competitor Analysis" и сгенерируй:
+1. Briefing document
+2. Mind map
+3. Audio overview (подкаст)
+```
+
+---
+
+## 💡 Use Cases
+
+### 1. Client Research
+
+```bash
+nlm create notebook "Client Research: Acme Campaign"
+nlm add source --url https://competitor1.com --notebook "Client Research: Acme Campaign"
+nlm add source --url https://competitor2.com --notebook "Client Research: Acme Campaign"
+nlm query "What patterns in mechanics?" --notebook "Client Research: Acme Campaign"
+nlm slides create --notebook "Client Research: Acme Campaign"
+nlm report create --notebook "Client Research: Acme Campaign"
+```
+
+### 2. Training Materials
+
+```bash
+nlm create notebook "AI Training Materials"
+nlm add source --url https://docs.anthropic.com --notebook "AI Training Materials"
+nlm add source --youtube https://youtube.com/watch?v=... --notebook "AI Training Materials"
+nlm flashcards create --notebook "AI Training Materials"
+nlm mindmap create --notebook "AI Training Materials"
+nlm quiz create --notebook "AI Training Materials"
+```
+
+### 3. Content Generation
+
+```bash
+nlm create notebook "Content Ideas"
+nlm add source --url https://research-article.com --notebook "Content Ideas"
+nlm audio create --notebook "Content Ideas" # Podcast
+nlm infographic create --notebook "Content Ideas" # Infographic
+```
+
+---
+
+## ⚠️ Важно
+
+- **Cookie-based auth** — токены хранятся локально, обновляются автоматически
+- **Undocumented API** — может сломаться после обновлений Google
+- **Personal use** — не для продакшена, только эксперименты
+- **Rate limits** — Google может ограничить при частых запросах
+
+---
+
+## 📝 Профили
+
+Если нужно несколько Google аккаунтов:
+
+```bash
+# Аутентификация с профилем
+nlm login --profile work
+nlm login --profile personal
+
+# Переключение
+nlm login switch work
+
+# Список профилей
+nlm login profile list
+
+# Использование конкретного профиля
+nlm list notebook --profile work
+```
+
+---
+
+## 🔗 Ссылки
+
+- **GitHub:** https://github.com/jacob-bd/notebooklm-mcp-cli
+- **MEMORY.md:** `vault/MEMORY.md` — секция "NotebookLM"
+- **MCP Config:** `~/.config/mcp/mcp_servers.json`
+
+---
+
+*Created: 2026-02-01*