Agent + CLI that turns Claude into a private French tutor: track progress, get an overview, get advice, read stories, and learn & practice French (written and spoken) online from your phone.
- An agent-first CLI cited from teken
(
afi-cli) — the runtime package has no third-party dependencies. - A mesh identity —
culture.yaml(suffix+backend) and the matching resident prompt file (AGENTS.colleague.md, since this template runsbackend: colleague). - The canonical guildmaster skill kit (11 skills) under
.claude/skills/, vendored cite-don't-import. Seedocs/skill-sources.md. - A build + deploy baseline — pytest, lint, the agent-first rubric gate, and PyPI Trusted Publishing wired into GitHub Actions.
uv sync
uv run pytest -n auto # run the test suite
uv run french whoami # identity from culture.yaml
uv run french learn # self-teaching prompt (add --json)
uv run teken cli doctor . --strict # the agent-first rubric gate CI runsThe installed command is french; the PyPI package and repo are
french-cli. french explain french and french explain french-cli both
resolve to the root doc entry.
french-cli is an LLM-free tutor engine that implements the
learn subject-plugin contract: it
owns the committed French content (stories, lessons, exercises) and each
learner's mastery state, resolves what to teach next, and emits structured
teaching directives. The driving agent (or human) does the conversational
tutoring the directive describes, grades pass|partial|fail, and writes the
result back with record. french never converses, grades free text, or computes
scores — learn-cli's motivation layer does the scoring.
| Verb | What it does | Payload kind |
|---|---|---|
french overview |
Subject self-description: modules + content counts. | subject_overview |
french progress |
The learner's mastery, counters, and next step. | progress |
french advice |
Deterministic study advice from stored state. | advice |
french story list | read <id> |
Graded stories + a reading directive. | story_list / story_read |
french lesson start | next | repeat |
Teaching directives from the curriculum. | lesson_directive |
french practice [<scope>] |
A batch of exercises to run (no scope = review). | practice_directive |
french record --item <id> --result … |
Write back one graded outcome; updates mastery. | record_ack |
french doctor |
Self-check + the pinned contract version. | subject_doctor |
| Verb | What it does |
|---|---|
french whoami |
Report this agent's nick, version, backend, and model from culture.yaml. |
french learn |
Print a structured self-teaching prompt. |
french explain <path> |
Markdown docs for any noun/verb path. |
french cli overview |
Describe the CLI surface itself. |
Every command supports --json. Results go to stdout, errors/diagnostics to
stderr (never mixed). Exit codes: 0 success, 1 user error, 2 environment
error, 3+ reserved.
Learner-scoped verbs take --learner <id> (default: $FRENCH_CLI_LEARNER or the
OS user; the resolved id is echoed back). State persists one JSON file per
learner under $FRENCH_CLI_LEARN_HOME (or $XDG_DATA_HOME/french_cli/learn, or
~/.local/share/french_cli/learn), written atomically and resumed across
sessions.
Stories are committed JSON files under content/stories/*.json (flat, one file
per story, filename == id), each validated against the shared story schema in
CI (tests/test_story_content.py). The starter set ships dev-prefixed dev
stories; the full graded ladder is authored separately. The curriculum (modules,
lessons, items, exercises) lives as structured data in
french/tutor/curriculum.py.
The contract schemas + validator are cited (cite-don't-import) into
french/contract_cite/ — see docs/contract-provenance.md.
See CLAUDE.md for the architecture and the full conventions
(version-bump-every-PR, the cicd PR lane, deploy setup).
Apache 2.0 — see LICENSE.