Glossary
Every term evoke uses, one line each: the outcomes of a decision, the parts of a reflex, the files of a project, the
adapter, and the words of a weave.
| Term | Meaning |
|---|---|
| Abstain | The outcome when none wins the route, or the winner is under the route floor. Exit 2 |
| Account | What each run of the input's words that no value holds does: says what to do, answers an argument's ask, says what is wanted from the result, asks for nothing, or asks for another thing. why names the words that answer an ask, those a call leaves out and those about the result; the JSON line carries them as left |
| Adapter | The classifier behind a decision, as an object answering typed questions with probabilities. jev and openjev reach Jev, through TypeSafe AI's API or through OpenJEV; clef and clef_flash reach Clef on Cloudflare's Workers AI; replay answers from a recording |
| Argument | A question about the input and a value for the body: options, vocab, pick or flag |
| Ask | The outcome when a required argument is missing: evoke asks the argument's own question |
| Binding | A value of one step's result taken by a later step, by kind: filled into a required argument, or written into the words for an optional one; or the whole result, by the name its source returns, handed beside the decision |
| Body | What run names: a .mts/.mjs file exporting a function, or a program with its arguments; a playbook has steps instead |
| Branch | Steps of a playbook that run only under a value an earlier step yields, { say = "…", when = { landing = "yes" } }, listed right after that step; the plan prints every one, the result picks, the rest are skipped clean |
| Call | A reflex with its arguments filled, on one line: lights room="den" state="off" |
| Cap | A reason a complete call stops at confirm: destructive, no gate, under the floor, a value read from your words another way, words that ask for more, a call that holds less than you typed |
| Change | A reflex whose effect is write or destructive: it changes something, where a lookup only reads |
| Collection | A repository of reflex directories |
| Confidence | The lowest probability among the route and every value the call holds |
| Confirm | The outcome when a call is complete but capped: [y]es [n]o [t]each |
| Contract | What a user cannot override: run or steps, argument names and sources, option keys, ranges, config keys, what a reflex yields, returns and takes |
| Effect | What running a reflex does: read, write or destructive. Absent means destructive, so a manifest always writes it |
| Effective manifest | The shipped manifest with your overlay merged in; what show <name> prints |
| Fits | The yes/no question asked when a reflex is added: does it do what another reflex's example asks? At or above its floor, add names the example |
| Flag | A yes/no argument. The body receives true or nothing |
| Floor | The bar a decision must clear. A threshold the adapter ships, meaning the probability this is right: route, read, write, whole, and fits at add |
| Fold | A part of a sentence that repeats a step a playbook wrote: removed from the plan and named under it, folded "…" into <n> |
| Gate | The step that turns confidence and effect into an outcome |
| h1 | The content hash of a reflex directory, h1:<sha256>, in the lock and the store |
| Home project | ~/.config/evoke/: the project used when no evoke.toml is found upward |
| Hub | Where reflexes are found; today the collection evoke-build/reflexes |
| Identity | How utterances are keyed: NFC, lower-cased, whitespace collapsed, trailing punctuation dropped |
| Inactive | A reflex left out of every decision until a problem is fixed: an empty vocabulary, an unset setting, an overlay that does not read |
| Local name | The key under [reflexes]: what the classifier reads and what the overlay file is named after |
| Lock | evoke.lock: per remote reflex, the ref, tag, commit, h1 and consented effect. Also the adapter |
| Lookup | A reflex whose effect is read: it finds or shows something, and changes nothing |
| Manifest | reflex.toml: what the classifier reads and what the body receives |
| Near neighbour | What a person might type that sounds like a reflex's request and is not: named under not_for, and written as a false test |
| Overlay | overlays/<name>.toml: your wording for one reflex, merged over the shipped one |
| Pick | An argument read from an exact span of the input: number, duration, email, url, quoted, date, time, amount, code |
| Pin | A question evoke asks the classifier on its own account beside a reflex's own, keyed weave.<name> in a recording: a second view of a listed word, a yes on a word of the input, what a run of words does, whether a call holds all the input says. Also what a plan file is checked against before it runs: the evoke version, the adapter's id and gate, the set's digest, each reflex and vocabulary by hash, a remote reflex's h1 |
| Plan | The installed set compiled to questions. Its digest keys the cache, the baselines and every decision |
| Plan file | A weave saved by try --save: the sentence, the classifier's answers and the pins it was decided under; run <file> runs it exactly, or names the pin that moved |
| Playbook | A reflex whose body is a plan: steps, one sentence per step with {slots}, reached by a short sentence and decided step by step over the installed set |
| Project | A directory of files you own: evoke.toml, evoke.lock, overlays/, vocab/ |
| Recalled | A value an earlier body of this session returned under the field a pick's recent names, listed at the ask and chosen by you |
| Record | One line of [examples] or [tests]: `"utterance" = { assertions } |
| Recording | An answers.toml: an adapter's declaration and its answers by utterance identity |
| Ref | Where a reflex comes from: owner/repo[/dir][@tag], a git URL, or ./dir |
| Reflex | A directory: a manifest and the file or program it names. Plural: reflexes |
| Returns | The name a body's whole data goes by, returns = "deploys", for a later step of one sentence to take |
| Route | The one choice over every installed reflex plus none |
| Run | The outcome when a call clears its floor. Also the key naming the body |
| Runtime | What runs a body: Node for a file, nothing for an argv |
| Situation | What a person faces, typed in one sentence, a problem or a routine: checkout is failing in eu-west, close the month. A playbook is reached by its situation |
| Slot | An argument named in a step of a playbook, {service}: filled from the sentence or asked before the plan prints; in square brackets, written only when stated |
| Span | An exact piece of the input, by character offsets |
| Step | One part of a weave, decided as one input is, a playbook's sentences among them; numbered as the plan prints it, run at its turn |
| Store | ~/.cache/evoke/store/<h1>/: fetched code, re-hashed before every run |
| Tag | A word under tags for --tag to narrow by. Also a git version tag |
| Takes | An argument an earlier step's whole result fills by that name, takes = "deploys": never asked, never stated, the plan alone hands it |
| Trust | A project's four owned paths bound to their content. Required outside home |
| Unstated | The answer that an argument was not given in the input |
| Utterance | A sentence a record is keyed by |
| View | One of two questions about a listed word, the argument's own and a second one in other words; a value both give stands, and why says asked a second way. Where they differ, the input's own words decide, or the call waits for a yes |
| Vocabulary | vocab/<name>.toml: your closed list of words, shared by every argument that names it |
| Weave | A sentence read as several steps: each decided on its own, ordered by the words, a result threaded into a later step, the plan shown before anything runs |
| Wording | What a user may override: descriptions, questions, option meanings, records |
| Yield | A field of a result's data a later step may take, declared under [yields] with the kind that reads it |