Skip to content

commands.md

Latest commit

 

History

History
141 lines (112 loc) · 4.85 KB

commands.md

File metadata and controls

141 lines (112 loc) · 4.85 KB
Raw

Command reference

Run mentor <command> --help for any command's own help. Without a command, mentor generates a prompt from the flags you pass.

Global options

Flag Description
--field TEXT the field to teach (required to generate a prompt)
--level {beginner,intermediate,advanced,expert} learner level (default: beginner)
--goal TEXT the learner's objective
--context TEXT the learner's background / prior knowledge
--language TEXT teaching language
--style {socratic,coaching,direct,immersive} teaching style (default: socratic)
--mentor-name TEXT override the generated mentor name
--provider NAME generic, anthropic/claude, openai/chatgpt/gpt, google/gemini, ollama/local, llamacpp
--model TEXT target model id (defaults per provider)
--emit {text,json,modelfile,markdown} output format (default depends on provider)
--compact short prompt for small-context local models
--out FILE write output to a file
-q, --quiet suppress usage/save notices
--prompt-format {auto,markdown,xml} prompt shape (auto = XML for Claude)
--thinking / --thinking-budget N Anthropic extended thinking
--reasoning-effort {none,minimal,low,medium,high,xhigh,max} OpenAI reasoning models
--no-cache disable Anthropic system-prompt caching
--learner NAME load/update a persisted learner profile
--remember auto-name the learner from --field and persist progress
--history-turns N run: max past turns kept in context (default 40)

mentor — generate a prompt

mentor --field "quantum computing" --level beginner
mentor --field "Rust" --provider claude --emit json > request.json
mentor --field "Rust" --provider ollama --compact --emit modelfile > Modelfile
mentor prompt --field "Rust" --provider chatgpt    # equivalent explicit form

mentor run [NAME] — live tutoring session

mentor run --field "Rust" --provider ollama --model qwen2.5:7b
mentor run rustbuddy                       # load a saved mentor
mentor run --field "Rust" --provider claude --remember
mentor run --field "Rust" --provider chatgpt --reasoning-effort high
mentor run --field "Rust" --provider claude --thinking --show-thinking

Session options:

Flag Description
--once "message" send a single message and exit
--dry-run print the request payload; send nothing
--host URL backend host for Ollama / llama.cpp
--temperature F sampling temperature (defaults per provider)
--show-thinking stream model reasoning, dimmed, on stderr
--learner NAME / --remember persist the learner profile

mentor save / show / saved

mentor save rustbuddy --field "Rust" --level intermediate
mentor show rustbuddy
mentor saved

mentor curriculum

mentor curriculum --field "Rust" --duration "6 weeks, 5h/week"

Produces a curriculum-designer prompt (outcome statement, prerequisite check, module plan, assessments, week-by-week schedule, failure modes).

mentor models

mentor models                       # built-in catalog for Claude & ChatGPT
mentor models --provider claude --refresh   # live list with your API key

mentor fields / mentor providers

mentor fields        # curated field profiles
mentor providers     # providers, default models, temperatures

Learner memory

mentor learners
mentor progress rust --add-mastered ownership --add-shaky lifetimes \
    --add-misconception "Rc == Arc" --add-question "Rc vs Box?" --goal "ship idiomatic Rust"
mentor sessions rust
mentor transcript <session-id>
mentor review rust [--apply] [--dry-run]

progress flags: --field, --level, --goal, --note, --add-mastered, --add-shaky, --add-misconception, --add-question, --reset.

Flashcards & quiz

mentor cards rust --add "What is ownership? :: Each value has one owner."
mentor cards rust --generate            # from the latest session
mentor cards rust                       # list with due counts
mentor cards rust --remove <id>
mentor cards rust --reset
mentor quiz rust [--all] [--n 10] [--shuffle]

Grades during a quiz: again, hard, good, easy (Enter = good, q = quit).

Configuration & utilities

mentor config                                   # show effective config
mentor config --set provider=claude --set level=intermediate --set style=coaching
mentor config --unset style
mentor config --reset
mentor doctor
mentor skill [--dir PATH] [--force]    # install the Claude Code / opencode skill
mentor version
mentor interactive

Exit codes

Code Meaning
0 success
1 runtime error (e.g. provider unreachable)
2 usage error / unsupported provider
130 interrupted (Ctrl-C)