Metadata-Version: 2.4
Name: notice-me-sensei
Version: 0.1.0
Summary: AI-powered personal learning coach with spaced repetition
Project-URL: Homepage, https://github.com/tdfirth/sensei
Project-URL: Repository, https://github.com/tdfirth/sensei
Project-URL: Issues, https://github.com/tdfirth/sensei/issues
Author: Tom Firth
License-Expression: MIT
License-File: LICENSE
Keywords: ai,claude,learning,mcp,spaced-repetition,srs,study
Classifier: Development Status :: 4 - Beta
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Education
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Education
Classifier: Typing :: Typed
Requires-Python: >=3.11
Requires-Dist: fastmcp>=2.0
Requires-Dist: pydantic>=2.0
Requires-Dist: typer>=0.21
Description-Content-Type: text/markdown

# Sensei

A personal learning coach that turns [Claude Code](https://claude.com/claude-code) into a tutor, practice partner, and spaced repetition system.

Sensei manages your study tasks, schedules reviews using spaced repetition, tracks your progress, and connects everything to your notes and code. When you open a sensei project in Claude Code, Claude becomes your sensei—guiding your learning with the patience of a teacher and the memory of a database.

## Installation

```bash
uv tool install notice-me-sensei
```

Or with pip:

```bash
pip install notice-me-sensei
```

## Tutorial

### 1. Initialize a study project

```bash
mkdir ~/study && cd ~/study
sensei init
```

This creates:
- `sensei.db` — SQLite database for tasks, assessments, and artifacts
- `topics/` — Directory for your study materials
- `practice/` — Directory for practice attempts
- `.claude/` — Claude Code integration (MCP server config and sensei skill)
- `CLAUDE.md` — Project context file

### 2. Open with Claude Code

```bash
claude
```

Claude automatically loads the sensei tools and skill. You can now have conversations like:

> **You:** I want to learn about attention mechanisms in transformers
>
> **Claude:** You will read the paper. Then implement it. Then we will see.
>
> I've created these tasks in `topics/attention/`:
> - `[a1b2c3]` Read "Attention Is All You Need" (learning)
> - `[d4e5f6]` Implement scaled dot-product attention (implementation)
> - `[g7h8i9]` Implement multi-head attention (implementation)
>
> What will you study first?

### 3. Work through tasks

As you complete tasks, Claude tracks your progress:

> **You:** I finished reading the paper and wrote some notes in topics/attention/notes.md
>
> **Claude:** *[reviews your notes]*
>
> Good. You covered the key concepts. A few gaps:
> - You didn't mention why we scale by √d_k
> - The masking section is light
>
> I'll create SRS cards for the concepts you should retain. These will resurface over the coming weeks.

### 4. Spaced repetition

SRS tasks automatically schedule based on your performance:

> **You:** What's due today?
>
> **Claude:** You have unfinished business.
>
> - `[j1k2l3]` Quiz: Explain the purpose of scaling in attention (due today)
>
> Ready?

When you pass, the interval increases. When you fail: "Again. Tomorrow."

### 5. Track progress

```bash
sensei status
```

```
Sensei Project Status
========================================
Pending tasks:  12
Active tasks:   1
Due today:      3
Overdue:        1
```

## CLI Reference

### Project Management

```bash
sensei init [--path PATH]    # Initialize a new project
sensei status                # Show project summary
```

### Tasks

```bash
sensei task list [OPTIONS]   # List tasks
  --status pending|active|completed|archived
  --type learning|implementation|srs|test|review
  --due-before DATE          # ISO8601 or "today"
  --path PATH                # Filter by topic path
  --limit N                  # Max results (default: 20)

sensei task create TITLE --type TYPE [OPTIONS]
  --description TEXT
  --due DATE
  --path PATH
  --parent ID

sensei task show ID          # Show task details
sensei task update ID [OPTIONS]
sensei task complete ID      # Mark as completed
sensei task schedule-next ID [--days N]  # Create follow-up SRS task
```

### Artifacts

Link files to tasks for context:

```bash
sensei artifact add TASK_ID PATH [--type TYPE]
sensei artifact list TASK_ID
```

Types: `note`, `code`, `notebook`, `pdf`, `test_harness`, `other`

Type is auto-detected from the file:
- `.md` → note
- `.py` → code (or `notebook` if it's a marimo notebook)
- `test_*.py` → test_harness
- `.ipynb` → notebook
- `.pdf` → pdf

### Assessments

```bash
sensei assess TASK_ID [OPTIONS]
  --passed / --failed
  --score N                  # 0-10 scale
  --feedback TEXT

sensei history TASK_ID       # Show assessment history
```

### Convenience Commands

```bash
sensei due [--days N]        # Tasks due within N days (default: 0 = today)
sensei search QUERY          # Search task titles and descriptions
```

## Task Types

| Type | Purpose | SRS? |
|------|---------|------|
| `learning` | Reading, watching, initial exploration | No |
| `implementation` | Write code for the first time | No |
| `srs` | Spaced repetition—questions, re-implementation from memory | Yes |
| `test` | One-off assessment | No |
| `review` | Review notes or code with Claude | No |

## How SRS Scheduling Works

Sensei uses an SM-2 inspired algorithm:

- **First pass:** Review tomorrow
- **Second pass:** Review in 3 days
- **Subsequent passes:** Interval × ease factor (default 2.5)
- **On failure:** Reset to 1 day, reduce ease factor

Scores (0-10) adjust the ease factor:
- Score ≥ 8: Ease factor increases (easier = longer intervals)
- Score < 5: Ease factor decreases (harder = shorter intervals)

## Project Structure

```
my-study-project/
├── .claude/
│   ├── skills/sensei/SKILL.md    # Claude's personality and workflows
│   └── settings.local.json       # MCP server configuration
├── topics/                       # Your study materials
│   └── attention/
│       ├── notes.md
│       ├── attention.py
│       └── practice/
│           └── attention-2025-01-15.py
├── practice/                     # General practice (no specific topic)
├── sensei.db                     # SQLite database
└── CLAUDE.md                     # Project goals and context
```

## Using Without Claude

Sensei works as a standalone CLI. You can manage tasks, record assessments, and track progress without Claude Code:

```bash
sensei task create "Read Chapter 5" --type learning --path topics/ml
sensei task list --status pending
sensei assess abc123 --passed --feedback "Understood gradient descent"
sensei task schedule-next abc123
sensei due
```

## License

MIT — Tom Firth
