feat(inference): call what a brain does by its functions - #84
Merged
Merged
Conversation
The remedy for a model the server is not set up for said "once it names a model from one of those". With a spec of inference now called a reason function, "it" would be the reason function, but it is the prompt that configures it which names the model. The remedy now reads "once its prompt names a model from one of those, which the details below list, it can be tried again." The phrasing and explanation tests use reason function as their sample noun instead of prompt. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The plain words of every result called a spec of inference a prompt. A prompt is what configures it; the thing a person defines is a reason function. The noun of inference is now reason function, so results read "Created the reason function “summary”.", "Ran the reason function “summary”." and "This brain has 1 reason function: “summary”.". The description of inference, which the spec tools carry, now opens: a spec of the inference primitive is a reason function that reasons with a language model, following a prompt, to turn an input into an answer; in conversation, call it a reason function, while the tools take the name inference. The rest of the description is unchanged. Tests that called the thing a prompt now say reason function: the server's plain words, the quick start and the leak check's sample. The spec operations' tests of an unknown primitive name it reason, the word a person might now try, instead of prompt. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The description of orchestration, which the spec tools carry, now opens with the words a person uses: a spec of the orchestration primitive is a workflow that coordinates the brain's other functions, running them in order, deciding what happens next and waiting for input; in conversation, call it a workflow, while the tools take the name orchestration. The rest of the description is unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
How a brain works now leads with what a brain can do: reason, interact, compute, recall and predict, coordinating those functions through workflows. Its table has a row per brain function, with what a person defines for it, the primitive that runs it, linked to its package, and whether it is built, in place of the table of primitives and its column What you make with it. A budget-review workflow shows the functions together. The ledger paragraph and the diagram stay. The quick start, the MCP section, the api README and the opening of the inference README call a spec of inference a reason function, and use prompt only for what configures it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CLAUDE.md now says, under What this is, which word a user reads for each primitive: a brain can reason (inference, whose specs are reason functions, each configured by a prompt), interact, compute, recall and predict, and coordinates them through workflows (orchestration); a primitive is a brain function and an execution a run, while code, the API and the architecture keep the primitive names. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follows #83.
What this changes
The words a user reads, nothing else: no tool, argument, route, package or type is renamed.
A brain can reason, interact, compute, recall and predict. It coordinates those functions through workflows. Those are the words in every sentence the server gives an assistant to pass on and in the README; the primitives keep their technical names (inference, orchestration and the rest) in the architecture, the code and the API, and each primitive's description says how the two relate, once:
CLAUDE.mdgives builders the mapping, so user-facing text uses the brain-function words and code keeps the primitive names.Verification
pnpm checkpasses: 2,274 tests across ten packages, 100% coverage on every file, nothing excluded.🤖 Generated with Claude Code