jev-recipes 0.2.0 → 0.3.0

package/recipes/result-outcome/README.md

@@ -1,49 +0,0 @@
-# Interpret a reported tool outcome
-
-What outcome does result report for task? Judge only what the response claims. A success status for transport alone does not establish completion of the requested task.
-
-```ts
-import { resultOutcome } from 'jev-recipes/result-outcome';
-
-const result = await resultOutcome({
-  task: 'Export the customer report.',
-  result: 'The export could not be created because the account lacks export access.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `task`          | Non-empty text                               |
-| `result`        | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict           | Meaning                                                                              |
-| ----------------- | ------------------------------------------------------------------------------------ |
-| `success`         | The response explicitly reports the full requested task completed.                   |
-| `partial_success` | The response reports some requested work completed and some unfinished.              |
-| `failure`         | The response explicitly reports the requested work failed or could not be performed. |
-| `unclear`         | The response does not establish a task outcome.                                      |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Interprets a report; it does not independently confirm an external action occurred. Prefer structured result fields when available.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo result-outcome` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe result-outcome` shows the input and result schemas.

package/recipes/result-usefulness/README.md

@@ -1,49 +0,0 @@
-# Check tool result usefulness
-
-Does result provide information useful for task? A response can be technically successful but contain no substantive information or information about a different task.
-
-```ts
-import { resultUsefulness } from 'jev-recipes/result-usefulness';
-
-const result = await resultUsefulness({
-  task: 'Find password reset instructions.',
-  result: 'Search completed successfully. No matching documents were found.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `task`          | Non-empty text                               |
-| `result`        | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict                 | Meaning                                                                           |
-| ----------------------- | --------------------------------------------------------------------------------- |
-| `useful`                | The response contains substantive information useful for the task.                |
-| `no_useful_information` | The response reports no results, an inability to help, or only empty boilerplate. |
-| `irrelevant`            | The response contains substantive information about a different task or subject.  |
-| `unclear`               | The response cannot be interpreted well enough to assess usefulness.              |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Assesses response content, not transport success or source truth. Handle empty bodies and status codes in code.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo result-usefulness` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe result-usefulness` shows the input and result schemas.

package/recipes/retrieval-needed/README.md

@@ -1,48 +0,0 @@
-# Check whether retrieval is needed
-
-Does request require facts beyond context? A request to transform or summarize supplied material does not need outside facts unless the request explicitly asks for them.
-
-```ts
-import { retrievalNeeded } from 'jev-recipes/retrieval-needed';
-
-const result = await retrievalNeeded({
-  request: 'Summarize this policy in one sentence.',
-  context: 'Customers may cancel their subscription from Account settings.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `request`       | Non-empty text                               |
-| `context`       | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict       | Meaning                                                                       |
-| ------------- | ----------------------------------------------------------------------------- |
-| `needed`      | Answering the request requires information missing from the supplied context. |
-| `unnecessary` | The requested work can be completed using only the supplied context.          |
-| `unclear`     | The request is too ambiguous to establish its information needs.              |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Assesses information needs without searching. A no-retrieval decision does not establish source truth.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo retrieval-needed` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe retrieval-needed` shows the input and result schemas.

package/recipes/route/README.md

@@ -1,47 +0,0 @@
-# Route a request
-
-Choose a handler from the routes you supply. The recipe returns the selection; your application runs the handler.
-
-```ts
-import { route } from 'jev-recipes/route';
-
-const result = await route({
-  request: 'I was charged twice.',
-  routes: {
-    billing: 'Payments, invoices, subscriptions, and refunds',
-    technical: 'Errors, outages, and broken integrations',
-  },
-});
-
-console.log(result.route);
-```
-
-## Input
-
-| Field           | Accepts                                                                   |
-| --------------- | ------------------------------------------------------------------------- |
-| `request`       | Non-empty text describing the request                                     |
-| `routes`        | 1 to 254 non-empty route names and descriptions; `__review__` is reserved |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8                              |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors. Types are inferred from this folder's Zod 4 schemas.
-
-## Result
-
-A confident selection returns `status: "ready"` and the route name in `route`. A low-confidence result or the reserved `__review__` choice returns `status: "review"` and a null `route`.
-
-`suggestedRoute` preserves a low-confidence suggestion. It is null when no route is selected. `confidence` and `probabilities` describe the model decision, including the `__review__` option.
-
-The result includes model and token usage. Inspect the outcome as well as its review status.
-
-## Reuse and calls
-
-Uses the shared choice helper. This folder owns the routing question, route criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Routes with overlapping descriptions can be difficult to distinguish. Describe their responsibilities clearly. Routing does not execute handlers or establish permission to act.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo route` shows an offline illustration, not an accuracy measurement. Use `npx jev-recipes describe route` to inspect the input and result schemas.

package/recipes/source-applicability/README.md

@@ -1,48 +0,0 @@
-# Check source applicability
-
-Does the scope described in passage apply to scenario? Distinguish applicable information from information about another product, environment, role, or condition.
-
-```ts
-import { sourceApplicability } from 'jev-recipes/source-applicability';
-
-const result = await sourceApplicability({
-  passage: 'Workspace owners can delete the workspace. These instructions are for owners only.',
-  scenario: 'A workspace guest wants to delete the workspace.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `passage`       | Non-empty text                               |
-| `scenario`      | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict          | Meaning                                                                 |
-| ---------------- | ----------------------------------------------------------------------- |
-| `applies`        | The passage explicitly covers or unambiguously applies to the scenario. |
-| `does_not_apply` | The passage is scoped to incompatible circumstances.                    |
-| `unclear`        | Required scope information is missing or ambiguous.                     |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Assesses semantic scope. Enforce tenant, access, exact version, and region constraints in application code.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo source-applicability` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe source-applicability` shows the input and result schemas.

package/recipes/step-complete/README.md

@@ -1,48 +0,0 @@
-# Check one completion condition
-
-Does evidence establish that condition has been met? A plan, attempted action, or unsupported assertion of completion is not enough unless the condition specifically concerns that report.
-
-```ts
-import { stepComplete } from 'jev-recipes/step-complete';
-
-const result = await stepComplete({
-  condition: 'The customer has received a reset email.',
-  evidence: 'A reset email was queued for delivery. Delivery has not been confirmed.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `condition`     | Non-empty text                               |
-| `evidence`      | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict   | Meaning                                                             |
-| --------- | ------------------------------------------------------------------- |
-| `met`     | The evidence establishes the full condition is satisfied.           |
-| `unmet`   | The evidence establishes the condition is not fully satisfied.      |
-| `unclear` | The evidence does not establish whether the condition is satisfied. |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Assesses supplied evidence for one condition. Use exact system state checks when the condition can be determined in code.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo step-complete` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe step-complete` shows the input and result schemas.

package/recipes/step-progress/README.md

@@ -1,51 +0,0 @@
-# Check step progress
-
-How does observation change progress toward objective relative to previousState? Base the decision on observed information or achieved conditions, not an unsupported claim of progress.
-
-```ts
-import { stepProgress } from 'jev-recipes/step-progress';
-
-const result = await stepProgress({
-  objective: 'Find the password reset documentation.',
-  previousState: 'No relevant page has been identified.',
-  observation: 'The search returned the official password reset guide.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `objective`     | Non-empty text                               |
-| `previousState` | Non-empty text                               |
-| `observation`   | Non-empty text                               |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict       | Meaning                                                                                        |
-| ------------- | ---------------------------------------------------------------------------------------------- |
-| `progress`    | The observation establishes a useful new fact or achieved condition toward the objective.      |
-| `no_progress` | The observation leaves the relevant state essentially unchanged.                               |
-| `setback`     | The observation establishes loss of previously achieved progress or a worsened relevant state. |
-| `unclear`     | The change in progress cannot be determined.                                                   |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Compares one observation with a supplied prior state. It does not measure elapsed time or schedule the next step.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo step-progress` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe step-progress` shows the input and result schemas.

package/recipes/summary-coverage/README.md

@@ -1,60 +0,0 @@
-# Check summary coverage
-
-Check whether a summary preserves each supplied point.
-
-```ts
-import { summaryCoverage } from 'jev-recipes/summary-coverage';
-
-const result = await summaryCoverage({
-  summary: 'We agreed to keep the API unchanged.',
-  points: [
-    {
-      id: 'api',
-      text: 'Keep the public API unchanged.',
-    },
-    {
-      id: 'docs',
-      text: 'Update the installation instructions.',
-    },
-  ],
-});
-
-console.log(result.status, result.checks);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `summary`       | Non-empty text                               |
-| `points`        | 1 to 50 `{ id, text }` items with unique IDs |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-`checks` contains the original ID, verdict, status, confidence, and probabilities for each item. `allPreserved` is true only when every check is ready and has verdict `preserved`. Any uncertain check makes overall status `review`.
-
-| Verdict     | Meaning                                                                  |
-| ----------- | ------------------------------------------------------------------------ |
-| `preserved` | The full material meaning of this point is retained.                     |
-| `partial`   | Some meaning is retained but a material detail or constraint is missing. |
-| `missing`   | This point is absent or its meaning is reversed.                         |
-| `unclear`   | The summary is too ambiguous to establish coverage.                      |
-
-`unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared item-check helper. This folder owns its question, verdict criteria, and aggregate decision. A live invocation makes one logical Jev request; SDK retries can add transport attempts. All item questions are sent in that request.
-
-## Limits
-
-Checks the points you supply. It does not select important points from a transcript or detect unrelated invented statements.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo summary-coverage` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe summary-coverage` shows the input and result schemas.

package/recipes/ticket-match/README.md

@@ -1,50 +0,0 @@
-# Compare support tickets
-
-Do firstTicket and secondTicket describe the same underlying reported issue? Similar symptoms alone establish relatedness, not identity. Use only supplied links between the reports.
-
-```ts
-import { ticketMatch } from 'jev-recipes/ticket-match';
-
-const result = await ticketMatch({
-  firstTicket: 'My workspace export fails with error EXPORT_TIMEOUT.',
-  secondTicket: 'Exports also time out with EXPORT_TIMEOUT in another workspace.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `firstTicket`   | Non-empty text                               |
-| `secondTicket`  | Non-empty text                               |
-| `context`       | Optional non-empty text                      |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict      | Meaning                                                                            |
-| ------------ | ---------------------------------------------------------------------------------- |
-| `same_issue` | The reports contain enough shared identifying context to establish the same issue. |
-| `related`    | The reports share symptoms or a subject but do not establish the same issue.       |
-| `different`  | The reports describe materially different issues.                                  |
-| `unclear`    | The descriptions are insufficient to establish even their relationship.            |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Does not merge tickets or establish a shared root cause from similar symptoms alone. Compare exact identifiers in code.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo ticket-match` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe ticket-match` shows the input and result schemas.

package/recipes/tone-check/README.md

@@ -1,55 +0,0 @@
-# Check writing criteria
-
-Check a draft against each supplied writing criterion.
-
-```ts
-import { toneCheck } from 'jev-recipes/tone-check';
-
-const result = await toneCheck({
-  draft: 'You caused this problem. Read the manual.',
-  criteria: [
-    {
-      id: 'blame',
-      text: 'Avoid blaming the customer.',
-    },
-  ],
-});
-
-console.log(result.status, result.checks);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `draft`         | Non-empty text                               |
-| `criteria`      | 1 to 50 `{ id, text }` items with unique IDs |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-`checks` contains the original ID, verdict, status, confidence, and probabilities for each item. `allPassed` is true only when every check is ready and has verdict `pass`. Any uncertain check makes overall status `review`.
-
-| Verdict   | Meaning                                              |
-| --------- | ---------------------------------------------------- |
-| `pass`    | The wording satisfies this criterion.                |
-| `fail`    | The wording violates this criterion.                 |
-| `unclear` | The criterion or wording is too ambiguous to assess. |
-
-`unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared item-check helper. This folder owns its question, verdict criteria, and aggregate decision. A live invocation makes one logical Jev request; SDK retries can add transport attempts. All item questions are sent in that request.
-
-## Limits
-
-Checks supplied writing criteria. It does not rewrite text, determine factual accuracy, or check exact character counts.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo tone-check` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe tone-check` shows the input and result schemas.

package/recipes/tool-fit/README.md

@@ -1,49 +0,0 @@
-# Check a tool fit
-
-Can the capabilities explicitly described in tool perform task? Do not assume capabilities that are not described. Capability fit is separate from permission to invoke the tool.
-
-```ts
-import { toolFit } from 'jev-recipes/tool-fit';
-
-const result = await toolFit({
-  task: 'Read the current incident status.',
-  tool: 'Status reader: retrieves active incidents. It cannot create or modify incidents.',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `task`          | Non-empty text                               |
-| `tool`          | Non-empty text                               |
-| `context`       | Optional non-empty text                      |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict        | Meaning                                                                    |
-| -------------- | -------------------------------------------------------------------------- |
-| `fits`         | The described capabilities can perform the requested task.                 |
-| `does_not_fit` | The described capabilities do not cover the task or explicitly exclude it. |
-| `unclear`      | The capability description lacks facts needed to decide.                   |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Assesses a supplied capability description. It does not discover tools, validate credentials, or grant permission.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo tool-fit` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe tool-fit` shows the input and result schemas.

package/recipes/topic-shift/README.md

@@ -1,50 +0,0 @@
-# Detect a topic shift
-
-Does message stay with currentTopic, introduce a different topic, or contain both? A new detail within the same subject is not automatically a new topic.
-
-```ts
-import { topicShift } from 'jev-recipes/topic-shift';
-
-const result = await topicShift({
-  currentTopic: 'Resetting a password.',
-  message: 'Also, where can I download my invoices?',
-});
-
-console.log(result.status, result.verdict);
-```
-
-## Input
-
-| Field           | Accepts                                      |
-| --------------- | -------------------------------------------- |
-| `currentTopic`  | Non-empty text                               |
-| `message`       | Non-empty text                               |
-| `context`       | Optional non-empty text                      |
-| `minConfidence` | Optional number from 0 to 1; defaults to 0.8 |
-
-See [shared options and behavior](../README.md#shared-options-and-behavior) for client configuration, validation, and errors.
-
-## Result
-
-| Verdict      | Meaning                                                                 |
-| ------------ | ----------------------------------------------------------------------- |
-| `same_topic` | The message remains within the current subject.                         |
-| `new_topic`  | The message moves to a different subject.                               |
-| `mixed`      | The message substantively addresses both the current and a new subject. |
-| `unclear`    | The subject relationship cannot be resolved.                            |
-
-The result includes `verdict`, `confidence`, and all choice `probabilities`. `unclear` always requires review, even at high confidence. Other verdicts can be ready, including negative assessments.
-
-Results include model and token usage. Decisions below `minConfidence` require review. Inspect the outcome as well as the status.
-
-## Reuse and calls
-
-Uses the shared choice helper for the Jev call and response parsing. This folder owns its question, verdict criteria, and review policy. A live invocation makes one logical Jev request; SDK retries can add transport attempts.
-
-## Limits
-
-Compares one message with a supplied topic. It does not generate topic labels or split conversation threads.
-
-## Example input
-
-[demo.json](demo.json) contains editable input and a hand-authored response. After installing `jev-recipes`, `npx jev-recipes demo topic-shift` shows the result offline. The fixture is an illustration, not an accuracy measurement. `npx jev-recipes describe topic-shift` shows the input and result schemas.