eve 0.6.0-beta.9 → 0.7.0

package/dist/docs/public/advanced/evals.md

@@ -1,158 +0,0 @@
----
-title: "Evals"
-description: "Define repeatable scored checks for an Eve agent with defineEvalSuite and run them with eve eval."
----
-
-An eval is a scored check that runs your agent against real sessions and grades the result. Use it to catch regressions when you change a prompt or a tool: compare the output against expected text, JSON, SQL, or behavior, and optionally ship the results to Braintrust.
-
-## `defineEvalSuite`
-
-Eve discovers suites under the app-root `evals/` directory, in `.eval.ts` files. The file path is the suite's identity, so you don't author an `id` or `name`.
-
-```text
-my-agent/
-├── agent/
-├── evals/
-│   ├── smoke.eval.ts
-│   └── weather.eval.ts
-└── package.json
-```
-
-```ts title="evals/weather.eval.ts"
-import { defineEvalSuite } from "eve/evals";
-import { Checks } from "eve/evals/checks";
-import { Run } from "eve/evals/scores";
-
-export default defineEvalSuite({
-  description: "Basic message and tool-usage coverage for the weather agent.",
-  cases: [
-    { id: "brooklyn-forecast", input: "What is the weather in Brooklyn?", expected: "Sunny" },
-  ],
-  checks: [Checks.didNotFail(), Checks.toolCalled("get_weather")],
-  scores: [Run.didNotFail()],
-});
-```
-
-Every suite needs `scores` (an empty array is fine) and either `cases` or `load`. The rest are optional: `description`, `task`, `checks`, `model`, `thresholds`, `modelOptions`, `tags`, `metadata`, `maxConcurrency`, `timeoutMs`, `reporters`. The init template adds `evals/**/*.ts` to `tsconfig.json`, so your eval code type-checks alongside the app.
-
-`model` is only required when a model-backed scorer (one of the `Autoevals` wrappers) is present without its own per-scorer model override — and it's the scorer model, not the agent's. Eve only uses it for model-backed scoring, never to swap out the agent under test. Pass a string id (e.g. `"anthropic/claude-opus-4.8"`) to route through the Vercel AI Gateway, or hand it an AI SDK model instance to use that directly.
-
-## Checks and scores
-
-Suites grade a case on two distinct tiers:
-
-- **Checks are hard assertions.** Any failed check marks the case failed and `eve eval` exits non-zero. Use them for the things that must hold — the run completed, the right tool ran, the output parses.
-- **Scores are soft data.** They land in reports and artifacts, and a below-threshold score marks the case `scored` — visible but not fatal, unless you pass `--strict`.
-
-Both exist at the suite level (applied to every case) and the case level (appended to the suite's). Built-in checks live on `eve/evals/checks`:
-
-| Check                                                         | Asserts                                                                           |
-| ------------------------------------------------------------- | --------------------------------------------------------------------------------- |
-| `Checks.completed()`                                          | The run did not fail and did not park on unanswered HITL input                    |
-| `Checks.waiting()`                                            | The run parked on HITL input (for approval-shaped suites)                         |
-| `Checks.didNotFail()`                                         | No terminal failure and no `turn.failed`/`step.failed` events (parked runs pass)  |
-| `Checks.messageIncludes(token)`                               | Joined assistant text contains `token` (string or RegExp)                         |
-| `Checks.outputEquals(value)` / `Checks.outputMatches(schema)` | Deep equality / Standard Schema (e.g. Zod) validation of the parsed output        |
-| `Checks.toolCalled(name, opts?)`                              | A matching tool call happened (`input`, `output`, `isError`, `times` constraints) |
-| `Checks.toolNotCalled(name)`                                  | No call to `name`                                                                 |
-| `Checks.toolOrder([...names])`                                | Tool names appear in order (other calls may interleave)                           |
-| `Checks.noFailedActions()`                                    | No tool, subagent, or skill action reported a failure                             |
-| `Checks.subagentCalled(name, opts?)`                          | A subagent delegation happened (`remoteUrl`, `output` constraints)                |
-| `Checks.event(predicate, label)`                              | Escape hatch: any predicate over the typed event stream                           |
-
-Matcher options accept a literal (objects partial-deep-match), a RegExp, or a function — return a boolean to act as a predicate, or return a value to compare against (handy for runner-assigned values like environment-provided URLs):
-
-```ts
-Checks.toolCalled("bash", { input: { command: /^pwd/ }, isError: false, times: 1 });
-Checks.subagentCalled("weather", {
-  remoteUrl: () => process.env.WEATHER_AGENT_URL!,
-  output: /72F/,
-});
-```
-
-A custom check is a plain function receiving `{ case, result, target }` and returning `{ name, passed, message? }`.
-
-## Cases and tasks
-
-A case pairs an `input` with an `expected`, plus optional `checks`, `scores`, tags, and metadata. `input` can be a string or an object. `expected` is optional, which is handy when you only care about behavior. List cases inline, or load them dynamically:
-
-```ts
-import { defineEvalSuite } from "eve/evals";
-import { loadYaml } from "eve/evals/loaders";
-import { Text } from "eve/evals/scores";
-
-export default defineEvalSuite({
-  async load() {
-    const doc = await loadYaml("evals/data/cases.yaml");
-    return (doc.evals as readonly { task: string; prompt: string; sql: string }[]).map((row) => ({
-      id: row.task,
-      input: row.prompt,
-      expected: row.sql,
-    }));
-  },
-  scores: [Text.exact()],
-});
-```
-
-The loaders (`loadJson`, `loadYaml` from `eve/evals/loaders`) resolve paths relative to the app root. They're meant for fixtures, not runtime agent code. For the task itself, `task.prompt` covers the single-string case, `task.messages` sets up a multi-turn conversation, and `task.parseOutput` lets you compare a transformed result rather than the raw final message.
-
-## Scorers and rubrics
-
-Scorers live in namespaces on `eve/evals/scores`. Pick the cheapest one that captures what "correct" means here. The deterministic scorers run instantly for free; an LLM judge runs once per case and burns tokens, so save it for when nothing simpler will do.
-
-| Need                                                       | Use                                                                                           |
-| ---------------------------------------------------------- | --------------------------------------------------------------------------------------------- |
-| Grade agent behavior (run succeeded, used the right tools) | `Run.didNotFail()`, `Run.usedTool(name, opts?)`, `Run.usedNoTools()`, `Run.maxToolCalls(max)` |
-| Exact string match                                         | `Text.exact()`, `Text.includes()`                                                             |
-| Fuzzy text match (typos, whitespace)                       | `Text.levenshtein()`                                                                          |
-| Exact JSON match                                           | `Json.deepEqual()`                                                                            |
-| Exact SQL match (after normalization)                      | `Sql.exactNormalized()`                                                                       |
-| LLM-judged factual correctness vs an expected answer       | `Autoevals.factuality()`                                                                      |
-| LLM-judged summary quality                                 | `Autoevals.summary()`                                                                         |
-| LLM-judged SQL semantic equivalence                        | `Autoevals.sql()`                                                                             |
-| LLM-judged free-form criteria (no `expected` to match)     | `Autoevals.closedQA({ criteria: "..." })`                                                     |
-
-Each scorer gets the flattened `input`, `output`, and `expected` strings along with the full case and task result — including derived facts: typed tool calls (name, input, output, error state), subagent calls, HITL input requests, and whether the run parked. `Run.usedTool` accepts the same matcher options as `Checks.toolCalled`, so you can grade tool-input quality fractionally where a check would assert it absolutely. Return `null` from a scorer to skip a case.
-
-### Thresholds
-
-By default a scorer has to hit an exact match to pass. `thresholds` loosens that, mapping each scorer name to the minimum score you'll accept:
-
-```ts
-defineEvalSuite({
-  cases: [{ id: "hello", input: "Hello", expected: "Hello" }],
-  scores: [Run.didNotFail(), Text.includes()],
-  thresholds: { "run.didNotFail": 1, "text.includes": 0.5 },
-});
-```
-
-`maxConcurrency` caps parallelism and `timeoutMs` bounds each case. Leave them off and `eve eval` runs up to 8 cases per suite at once. For provider-specific scorer-model settings, use `modelOptions.providerOptions`. Individual Autoevals scorers can also take their own `model` / `modelOptions`, which win over the suite default.
-
-## Reporters
-
-Suite-level reporters live on `eve/evals/reporters`. Right now the public surface is `Braintrust(...)`, which uploads suite results to Braintrust experiments. Eve still runs and scores everything itself; the reporter just ships the results out. Checks log as binary scores under a `check:` prefix so experiments diff check regressions the same way they diff score regressions. Suite and case `metadata` ride along to the reporter, and the CLI prints a console summary by default.
-
-## Running evals
-
-```bash
-eve eval                       # run all discovered suites locally
-eve eval weather smoke         # run selected suites
-eve eval --url https://<app>   # target a remote app instead of a local host
-eve eval --tag fast            # only cases (or suites) carrying a tag
-eve eval --case brooklyn-forecast  # only specific case ids
-eve eval --strict              # below-threshold scores also fail the exit code
-eve eval --list                # print discovered suites and cases without running
-eve eval --json                # machine-readable output
-eve eval --skip-report         # skip suite-defined reporters (e.g. Braintrust)
-```
-
-Exit codes: `0` when every case passed its checks (and thresholds, under `--strict`), `1` when any case failed (a failed check, an execution error, or a strict threshold miss), `2` for configuration errors. Each run drops per-suite artifacts under `.eve/evals/<timestamp>-<suite-id>/`, including per-case check results and verdicts.
-
-## A good baseline
-
-Most apps do fine with a small smoke suite. Assert behavior with `Checks.didNotFail()` plus one or two content checks, keep case fixtures in `evals/data/`, and only reach for Braintrust once you actually need shared result review or experiment history. In CI, run `eve eval --strict` so threshold misses fail the build too.
-
-## What to read next
-
-- [Tools](../tools): the surface most evals assert on
-- [instrumentation.ts](./instrumentation): tracing and debugging

package/dist/src/cli/commands/setup.d.ts

@@ -1,55 +0,0 @@
-import { type DeploymentInfo } from "#setup/scaffold/primitives/index.js";
-import { type Prompter } from "#setup/prompter.js";
-import { type ArgsHeadlessAiGateway, type ArgsHeadlessProject, type ChannelKind, type ChatPreference } from "#setup/state.js";
-export interface SetupCliLogger {
-    error(message: string): void;
-    log(message: string): void;
-}
-/**
- * Preset answers for a headless `eve setup` run. They mirror the create flow's
- * preset options so a headless onboarding run is fully determined by its
- * inputs; each field is threaded into the matching onboarding box option.
- */
-export interface SetupCommandPresets {
-    presetName?: string;
-    presetModel?: string;
-    presetChannels?: ChannelKind[];
-    presetCreateSlackbot?: boolean;
-    presetNoDeploy?: boolean;
-    overwriteExisting?: boolean;
-    presetConnections?: string[];
-    presetChatPreference?: ChatPreference;
-}
-export interface SetupCommandOptions {
-    /**
-     * Programmatic-only for now: the `eve setup` CLI registers no headless flags,
-     * so a headless run is only reachable through this API with presets that
-     * answer every decision. The CLI fails fast on a non-interactive terminal
-     * instead of silently entering a headless run that cannot succeed.
-     */
-    headless?: boolean;
-    /** Override for the scaffolded `eve` dependency version. Defaults to "beta",
-     *  matching the create flow; ai/zod/tsgo come from build-stamped defaults. */
-    evePackageVersion?: string;
-    presets?: SetupCommandPresets;
-    /** Headless-only Vercel provisioning flags, mirroring the create flow. */
-    provisioning?: {
-        project: ArgsHeadlessProject;
-        aiGateway: ArgsHeadlessAiGateway;
-    };
-}
-export interface SetupCommandDependencies {
-    createPrompter: () => Prompter;
-    detectDeployment(projectPath: string): Promise<DeploymentInfo>;
-    isEveProject(projectPath: string): Promise<boolean>;
-    hasInteractiveTerminal(): boolean;
-}
-/**
- * The unified `eve setup` entry. In a directory that is not yet an Eve project
- * it runs the SAME onboarding composition as `pnpm create eve`, scaffolding in
- * place; inside an existing project it runs in-project setup: the channel
- * interview, then channel scaffolding, then deploy, against the detected
- * on-disk project link. Both compose the shared boxes through the shared
- * runner.
- */
-export declare function runSetupCommand(logger: SetupCliLogger, appRoot: string, options: SetupCommandOptions, dependencies?: SetupCommandDependencies): Promise<void>;

package/dist/src/cli/commands/setup.js

@@ -1 +0,0 @@
-import{basename}from"node:path";import{isEveProject}from"#setup/scaffold/index.js";import{createAddToAgentState}from"#setup/scaffold/steps/index.js";import{detectDeployment}from"#setup/scaffold/primitives/index.js";import{addChannels}from"#setup/boxes/add-channels.js";import{deployProject as deployProject$1}from"#setup/boxes/deploy-project.js";import{selectChannels}from"#setup/boxes/select-channels.js";import{composeOnboardingBoxes}from"#setup/onboarding.js";import{createPrompter}from"#setup/prompter.js";import{runHeadless,runInteractive}from"#setup/runner.js";import{createDefaultSetupState,snapshotSetupState}from"#setup/state.js";const defaultDependencies={createPrompter,detectDeployment,isEveProject,hasInteractiveTerminal:()=>!!(process.stdin.isTTY&&process.stdout.isTTY)};async function runSetupCommand(t,n,r,i=defaultDependencies){let a=!await i.isEveProject(n),o=r.evePackageVersion??`beta`,s=r.presets??{},c=r.headless!==!0;if(c&&!i.hasInteractiveTerminal()){t.error("eve setup is interactive and needs a terminal. For automation, use `create-eve --headless` to scaffold a new agent or `eve channels add` inside an existing project."),process.exitCode=1;return}let l=i.createPrompter(),u={write:e=>l.log.message(e)};l.intro(a?`Create a new Eve agent`:`Set up your Eve agent`);let d=a?`Agent created.`:`Setup complete.`,f,p;a?(f=createDefaultSetupState(),p=composeOnboardingBoxes({prompter:l,presetName:s.presetName,presetModel:s.presetModel,presetChannels:s.presetChannels,presetConnections:s.presetConnections,presetCreateSlackbot:s.presetCreateSlackbot,provisioning:r.provisioning,presetChatPreference:s.presetChatPreference??`skip`,targetDirectory:n,inPlace:!0,overwriteExisting:s.overwriteExisting,presetNoDeploy:s.presetNoDeploy,headless:!c,evePackageVersion:o})):(l.log.message(`Checking the current Vercel project...`),f={...createDefaultSetupState(),...createAddToAgentState(await i.detectDeployment(n)),agentName:basename(n),projectPath:{kind:`resolved`,inPlace:!0,path:n}},p=[selectChannels({prompter:l,presetChannels:s.presetChannels}),addChannels({prompter:l,evePackageVersion:o,presetCreateSlackbot:s.presetCreateSlackbot}),deployProject$1({prompter:l,skip:s.presetNoDeploy})]);try{if(c){let e=await runInteractive(p,f,u,{snapshot:snapshotSetupState});l.outro(e.kind===`cancelled`?`Cancelled.`:d);return}await runHeadless(p,f,u,{snapshot:snapshotSetupState}),l.outro(d)}catch(e){t.error(e instanceof Error?e.message:String(e)),process.exitCode=1}}export{runSetupCommand};

package/dist/src/cli/dev/repl/input-requests.d.ts

@@ -1,38 +0,0 @@
-import type { HandleMessageStreamEvent } from "#protocol/message.js";
-import type { InputRequest, InputResponse } from "#runtime/input/types.js";
-import { type CliTheme } from "#cli/ui/output.js";
-/**
- * Sentinel returned when the CLI user pressed Escape while responding to a
- * pending runtime input request.
- */
-export declare const ESCAPED_RUNTIME_INPUT_PROMPT: unique symbol;
-/**
- * One raw answer collected from the CLI while prompting for pending runtime
- * input requests.
- */
-export type RuntimeInputPromptValue = string | typeof ESCAPED_RUNTIME_INPUT_PROMPT | undefined;
-/**
- * Outcome of one interactive CLI prompt pass for pending runtime input.
- */
-export type RuntimeInputPromptResult = {
-    readonly inputResponses: readonly InputResponse[];
-    readonly kind: "submitted";
-} | {
-    readonly kind: "deferred";
-} | {
-    readonly kind: "aborted";
-};
-/**
- * Returns the last pending runtime-input batch emitted in one turn, if any.
- */
-export declare function extractPendingRuntimeInputRequests(events: readonly HandleMessageStreamEvent[]): readonly InputRequest[];
-/**
- * Interactively collects CLI-native responses for pending runtime input
- * requests.
- */
-export declare function promptForRuntimeInputRequests(input: {
-    ask(prompt: string): Promise<RuntimeInputPromptValue>;
-    print(message: string): void;
-    requests: readonly InputRequest[];
-    theme: CliTheme;
-}): Promise<RuntimeInputPromptResult>;

package/dist/src/cli/dev/repl/input-requests.js

@@ -1 +0,0 @@
-import{renderCliTaggedLine}from"#cli/ui/output.js";import{resolveTextToResponse}from"#channel/resolve-text.js";const ESCAPED_RUNTIME_INPUT_PROMPT=Symbol(`eve.cli.runtime-input.escaped`);function extractPendingRuntimeInputRequests(e){for(let t=e.length-1;t>=0;--t){let n=e[t];if(n?.type===`input.requested`)return n.data.requests}return[]}async function promptForRuntimeInputRequests(t){let n=[];t.print(renderCliTaggedLine(t.theme,{message:`pending input request${t.requests.length===1?``:`s`}: reply now, press Escape, or type "skip" to return to the normal prompt`,tag:`input`,tone:`info`}));for(let e of t.requests){let r=await promptForRuntimeInputRequest({ask:t.ask,print:t.print,request:e,theme:t.theme});if(r.kind!==`submitted`)return r;n.push(r.inputResponse)}return{inputResponses:n,kind:`submitted`}}function isApprovalRequest(e){return e.display===`confirmation`}async function promptForRuntimeInputRequest(e){return isApprovalRequest(e.request)?await promptForApprovalInputRequest({ask:e.ask,print:e.print,request:e.request,theme:e.theme}):await promptForQuestionInputRequest({ask:e.ask,print:e.print,request:e.request,theme:e.theme})}async function promptForApprovalInputRequest(t){for(t.print(renderCliTaggedLine(t.theme,{message:`approve tool "${t.request.action.toolName}"? (yes/no)`,tag:`approval`,tone:`warning`})),t.print(renderCliTaggedLine(t.theme,{message:JSON.stringify(t.request.action.input,null,2),tag:`tool`,tone:`muted`}));;){let r=await t.ask(`approval> `);if(r===void 0)return{kind:`aborted`};if(r===ESCAPED_RUNTIME_INPUT_PROMPT)return{kind:`deferred`};let i=r.trim().toLowerCase();if(i===`skip`||i===`/skip`)return{kind:`deferred`};if(i===`a`||i===`approve`||i===`y`||i===`yes`)return{inputResponse:{optionId:`approve`,requestId:t.request.requestId},kind:`submitted`};if(i===`d`||i===`deny`||i===`n`||i===`no`)return{inputResponse:{optionId:`deny`,requestId:t.request.requestId},kind:`submitted`};t.print(renderCliTaggedLine(t.theme,{message:`enter "yes" to approve, "no" to deny, press Escape, or type "skip" to return to the normal prompt`,tag:`input`,tone:`warning`}))}}async function promptForQuestionInputRequest(r){r.print(renderCliTaggedLine(r.theme,{message:r.request.prompt,tag:`question`,tone:`info`}));for(let[t,n]of(r.request.options??[]).entries())r.print(renderCliTaggedLine(r.theme,{message:`${t+1}. ${n.label}${n.description===void 0?``:` - ${n.description}`}`,tag:`option`,tone:`muted`}));let i=r.request.allowFreeform===!0||(r.request.options?.length??0)===0;for(;;){let a=await r.ask(`answer> `);if(a===void 0)return{kind:`aborted`};if(a===ESCAPED_RUNTIME_INPUT_PROMPT)return{kind:`deferred`};let o=a.trim();if(o===`skip`||o===`/skip`)return{kind:`deferred`};let s=resolveTextToResponse(o,r.request);if(s!==void 0)return{inputResponse:s,kind:`submitted`};r.print(renderCliTaggedLine(r.theme,{message:r.request.options===void 0||r.request.options.length===0?`enter a response, press Escape, or type "skip" to return to the normal prompt`:i?`enter an option number/id, freeform text, press Escape, or type "skip" to return to the normal prompt`:`enter an option number/id, press Escape, or type "skip" to return to the normal prompt`,tag:`input`,tone:`warning`}))}}export{ESCAPED_RUNTIME_INPUT_PROMPT,extractPendingRuntimeInputRequests,promptForRuntimeInputRequests};

package/dist/src/cli/dev/repl/input.d.ts

@@ -1,19 +0,0 @@
-/**
- * Parsed result for one line of interactive `eve dev` input.
- */
-export type DevReplInput = {
-    kind: "empty";
-} | {
-    kind: "exit";
-} | {
-    kind: "help";
-} | {
-    kind: "new";
-} | {
-    kind: "message";
-    message: string;
-};
-/**
- * Parse one line of input for the minimal development REPL.
- */
-export declare function parseDevReplInput(value: string): DevReplInput;

package/dist/src/cli/dev/repl/input.js

@@ -1 +0,0 @@
-function parseDevReplInput(e){let t=e.trim();return t.length===0?{kind:`empty`}:t===`/exit`||t===`/quit`?{kind:`exit`}:t===`/help`?{kind:`help`}:t===`/new`?{kind:`new`}:{kind:`message`,message:t}}export{parseDevReplInput};

package/dist/src/cli/dev/repl/repl.d.ts

@@ -1,62 +0,0 @@
-import { type HandleMessageStreamEvent } from "#protocol/message.js";
-import type { InputRequest } from "#runtime/input/types.js";
-import { type CliTheme } from "#cli/ui/output.js";
-import { createDevelopmentTerminal } from "#cli/dev/repl/terminal.js";
-export type ReplDisplayBlockKind = "content" | "meta";
-export type LiveContentKind = "message" | "reasoning";
-export interface FormattedContentEvent {
-    readonly finalized: boolean;
-    readonly kind: LiveContentKind;
-    readonly line: string;
-}
-export interface TurnDisplayState {
-    readonly activeLiveContentKind: LiveContentKind | null;
-    readonly lastPrintedBlockKind: ReplDisplayBlockKind | null;
-}
-interface TurnEventRenderOptions {
-    readonly isSubagent?: boolean;
-    readonly sourceLabel?: string;
-}
-export type TurnDisplayTerminal = Pick<ReturnType<typeof createDevelopmentTerminal>, "commitLive" | "print" | "updateLive">;
-export declare function formatContentEvent(theme: CliTheme, event: HandleMessageStreamEvent, options?: TurnEventRenderOptions): FormattedContentEvent | undefined;
-export declare function formatEvent(theme: CliTheme, event: HandleMessageStreamEvent, options?: TurnEventRenderOptions): string | undefined;
-export declare function createTurnDisplayState(): TurnDisplayState;
-export declare function renderTurnEvent(input: {
-    readonly event: HandleMessageStreamEvent;
-    readonly options?: TurnEventRenderOptions;
-    readonly state: TurnDisplayState;
-    readonly terminal: TurnDisplayTerminal;
-    readonly theme: CliTheme;
-}): TurnDisplayState;
-/**
- * Decides whether the REPL should drain in-flight subagent stream
- * subscriptions before redrawing the next user prompt.
- *
- * Returns `true` for ordinary turn boundaries (no pending input
- * requests on the parent stream). The drain ensures any final
- * subagent events — `session.completed`, `session.failed`, or
- * trailing messages — render in order with the parent boundary.
- *
- * Returns `false` when the parent's boundary carries one or more
- * pending input requests (HITL). A subagent that triggered the HITL
- * request runs in `task` mode (see `execution/subagent-tool.ts`)
- * and never emits a turn-boundary event on its own stream while
- * parked for HITL — `harness/tool-loop.ts` gates `emitTurnEpilogue`
- * on `mode === "conversation"`. Blocking on
- * `ReplSubagentStreamManager.waitForIdle()` here would stall the
- * prompt until the subagent's serverless function times out, which
- * is exactly the multi-minute delay this gate prevents.
- *
- * Subagent subscriptions stay alive across the prompt; they resume
- * rendering when the user answers and the subagent unparks.
- *
- * Exported for unit-test coverage of the gating contract.
- */
-export declare function shouldDrainSubagentStreamsOnBoundary(pendingInputRequests: readonly InputRequest[]): boolean;
-/**
- * Run the minimal interactive `eve dev` REPL against an Eve server.
- */
-export declare function runDevelopmentRepl(input: {
-    serverUrl: string;
-}): Promise<void>;
-export {};

package/dist/src/cli/dev/repl/repl.js

@@ -1,2 +0,0 @@
-import{EVE_CONTINUE_SESSION_ROUTE_PATTERN,EVE_CREATE_SESSION_ROUTE_PATH,EVE_MESSAGE_STREAM_ROUTE_PATTERN,createEveMessageStreamRoutePath}from"#protocol/routes.js";import{createCliTheme,renderCliBanner,renderCliSection,renderCliSpeakerLine,renderCliTaggedLine}from"#cli/ui/output.js";import{createInterface,emitKeypressEvents}from"node:readline";import{openStreamIterable}from"#client/open-stream.js";import{isCurrentTurnBoundaryEvent}from"#protocol/message.js";import{toErrorMessage}from"#shared/errors.js";import{createDevelopmentRequestHeadersAsync}from"#services/dev-client/request-headers.js";import{extractCurrentTurnBoundaryEvent}from"#services/dev-client/stream.js";import{resolveDevelopmentServerResourceUrl}from"#services/dev-client/url.js";import{formatVercelAuthChallengeMessage,isVercelAuthChallenge}from"#services/dev-client/vercel-auth-error.js";import{createDevClient}from"#services/dev-client.js";import{parseDevReplInput}from"#cli/dev/repl/input.js";import{ESCAPED_RUNTIME_INPUT_PROMPT,extractPendingRuntimeInputRequests,promptForRuntimeInputRequests}from"#cli/dev/repl/input-requests.js";import{createDevelopmentTerminal}from"#cli/dev/repl/terminal.js";function renderConnectionRows(r){let i=[{label:`Server`,tone:`info`,value:r.serverUrl},{label:`Create`,tone:`info`,value:`POST ${EVE_CREATE_SESSION_ROUTE_PATH}`},{label:`Continue`,tone:`info`,value:`POST ${EVE_CONTINUE_SESSION_ROUTE_PATTERN}`},{label:`Stream`,tone:`info`,value:`GET ${EVE_MESSAGE_STREAM_ROUTE_PATTERN}`}];return i.push({label:`Session`,value:`Follow-up messages reuse the active continuation token.`}),i}function renderCommandRows(){return[{label:`/help`,value:`Print the connection contract and available commands.`},{label:`/new`,value:`Clear the current durable session cursor.`},{label:`/exit`,value:`Exit the REPL.`}]}function renderIntro(e,t){return[renderCliBanner(e,{subtitle:`Interactive development REPL for the active Eve server.`,title:`Eve Dev`}),``,renderCliSection(e,{rows:renderConnectionRows(t),title:`Connection`}),``,renderCliSection(e,{rows:renderCommandRows(),title:`Commands`})].join(`
-`)}function normalizeStepMessage(e){let t=e.trim();return t.length>0?t:null}function getRenderTag(e){return e.options?.isSubagent===!0?`subagent`:e.fallback}function getRenderTone(e){return e.options?.isSubagent===!0?`subagent`:e.fallback}function prefixSourceLabel(e){return e.options?.sourceLabel===void 0?e.message:`${e.options.sourceLabel}${e.separator??` `}${e.message}`}function formatContentEvent(e,t,n){switch(t.type){case`message.appended`:return{finalized:!1,kind:`message`,line:renderCliSpeakerLine(e,{message:t.data.messageSoFar,speaker:n?.sourceLabel??`agent`,tone:getRenderTone({fallback:`accent`,options:n})})};case`message.completed`:if(t.data.message===null)return;if(t.data.finishReason===`tool-calls`){let r=normalizeStepMessage(t.data.message);return r===null?void 0:{finalized:!0,kind:`message`,line:renderCliTaggedLine(e,{message:prefixSourceLabel({message:r,options:n,separator:`: `}),tag:getRenderTag({fallback:`step`,options:n}),tone:getRenderTone({fallback:`accent`,options:n})})}}return{finalized:!0,kind:`message`,line:renderCliSpeakerLine(e,{message:t.data.message,speaker:n?.sourceLabel??`agent`,tone:getRenderTone({fallback:`accent`,options:n})})};case`reasoning.appended`:return{finalized:!1,kind:`reasoning`,line:renderCliTaggedLine(e,{message:prefixSourceLabel({message:t.data.reasoningSoFar,options:n,separator:`: `}),tag:getRenderTag({fallback:`reasoning`,options:n}),tone:getRenderTone({fallback:`info`,options:n})})};case`reasoning.completed`:return{finalized:!0,kind:`reasoning`,line:renderCliTaggedLine(e,{message:prefixSourceLabel({message:t.data.reasoning,options:n,separator:`: `}),tag:getRenderTag({fallback:`reasoning`,options:n}),tone:getRenderTone({fallback:`info`,options:n})})};default:return}}function formatEvent(e,t,n){let r=formatContentEvent(e,t,n);if(r!==void 0)return r.line;switch(t.type){case`message.received`:return;case`actions.requested`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.actions.length} action${t.data.actions.length===1?``:`s`})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`muted`,options:n})});case`input.requested`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.requests.length} request${t.data.requests.length===1?``:`s`})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`info`,options:n})});case`action.result`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${formatActionResultLabel(t.data.result)})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`muted`,options:n})});case`session.waiting`:case`session.completed`:return;case`authorization.required`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.name}: ${t.data.description}${formatAuthorizationChallengeSuffix(t.data.authorization)})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`warning`,options:n})});case`compaction.requested`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`compacting conversation history`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`muted`,options:n})});case`compaction.completed`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`conversation history compacted`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`muted`,options:n})});case`step.failed`:case`turn.failed`:case`session.failed`:return;case`subagent.called`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.name} -> ${t.data.childSessionId})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`info`,options:n})});case`subagent.started`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.subagentName})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`info`,options:n})});case`subagent.event`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.subagentName}: ${formatNestedSubagentEventLabel(t.data.event)})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`muted`,options:n})});case`subagent.completed`:return renderCliTaggedLine(e,{message:prefixSourceLabel({message:`${t.type} (${t.data.subagentName})`,options:n}),tag:getRenderTag({fallback:`event`,options:n}),tone:getRenderTone({fallback:`info`,options:n})});default:return}}function getEventDisplayBlockKind(e){switch(e.type){case`message.appended`:case`message.completed`:case`reasoning.appended`:case`reasoning.completed`:return`content`;default:return`meta`}}function createTurnDisplayState(){return{activeLiveContentKind:null,lastPrintedBlockKind:null}}function printDisplayLine(e){let t=e.state;return t.activeLiveContentKind!==null&&(e.terminal.commitLive(),t={activeLiveContentKind:null,lastPrintedBlockKind:`content`}),t.lastPrintedBlockKind!==null&&t.lastPrintedBlockKind!==e.kind&&e.terminal.print(``),e.terminal.print(e.line),{activeLiveContentKind:null,lastPrintedBlockKind:e.kind}}function renderTurnEvent(e){let t=formatContentEvent(e.theme,e.event,e.options);if(t!==void 0){let n=e.state;return n.activeLiveContentKind!==null&&n.activeLiveContentKind!==t.kind&&(e.terminal.commitLive(),n={activeLiveContentKind:null,lastPrintedBlockKind:`content`}),n.lastPrintedBlockKind!==null&&n.lastPrintedBlockKind!==`content`&&e.terminal.print(``),e.terminal.updateLive(t.line),t.finalized?(e.terminal.commitLive(),{activeLiveContentKind:null,lastPrintedBlockKind:`content`}):{activeLiveContentKind:t.kind,lastPrintedBlockKind:`content`}}let n=formatEvent(e.theme,e.event,e.options);return n===void 0?e.state:printDisplayLine({kind:getEventDisplayBlockKind(e.event),line:n,state:e.state,terminal:e.terminal})}function isAbortLikeError(e){return(e instanceof DOMException||e instanceof Error)&&e.name===`AbortError`}function shouldDrainSubagentStreamsOnBoundary(e){return e.length===0}var ReplSubagentStreamManager=class{#e=new Map;#t;#n;#r;#i;constructor(e){this.#t=e.displayStateRef,this.#n=e.serverUrl,this.#r=e.terminal,this.#i=e.theme}subscribe(e){if(this.#e.has(e.sessionId))return;let t=new AbortController,n=this.#a({controller:t,sessionId:e.sessionId,subagentName:e.subagentName}).finally(()=>{this.#e.delete(e.sessionId)});this.#e.set(e.sessionId,{controller:t,done:n,label:e.subagentName})}async waitForIdle(){for(;this.#e.size>0;)await Promise.all([...this.#e.values()].map(e=>e.done))}async close(){let e=[...this.#e.values()];for(let t of e)t.controller.abort();await Promise.allSettled(e.map(e=>e.done))}async#a(e){let t=resolveDevelopmentServerResourceUrl({resource:createEveMessageStreamRoutePath(e.sessionId),serverUrl:this.#n});try{for await(let n of openStreamIterable({host:this.#n,maxReconnectAttempts:3,resolveHeaders:async()=>await createDevelopmentRequestHeadersAsync({resourceUrl:t}),sessionId:e.sessionId,signal:e.controller.signal,startIndex:0}))if(this.#t.current=renderTurnEvent({event:n,options:{isSubagent:!0,sourceLabel:e.subagentName},state:this.#t.current,terminal:this.#r,theme:this.#i}),n.type===`subagent.called`&&this.subscribe({sessionId:n.data.childSessionId,subagentName:n.data.name}),isCurrentTurnBoundaryEvent(n))return}catch(t){if(isAbortLikeError(t))return;let n=toErrorMessage(t);this.#t.current=printDisplayLine({kind:`meta`,line:renderCliTaggedLine(this.#i,{message:`${e.subagentName} stream failed: ${n}`,tag:`subagent`,tone:`danger`}),state:this.#t.current,terminal:this.#r})}finally{e.controller.abort()}}};function formatActionResultLabel(e){switch(e.kind){case`load-skill-result`:return e.kind;case`subagent-result`:return`${e.kind}:${e.subagentName}`;case`tool-result`:return`${e.kind}:${e.toolName}`}}function formatAuthorizationChallengeSuffix(e){if(e===void 0)return``;let t=[];return e.url!==void 0&&t.push(e.url),e.userCode!==void 0&&t.push(`code ${e.userCode}`),e.instructions!==void 0&&t.push(e.instructions),t.length===0?``:` — ${t.join(`, `)}`}function formatNestedSubagentEventLabel(e){switch(e.type){case`actions.requested`:{let t=e.data.actions,n=t.map(e=>e.kind===`tool-call`?e.toolName:e.kind).join(`,`);return`${e.type} (${t.length} action${t.length===1?``:`s`}${n.length>0?`: ${n}`:``})`}case`action.result`:return`${e.type} (${formatActionResultLabel(e.data.result)})`;case`input.requested`:return`${e.type} (${e.data.requests.length} request${e.data.requests.length===1?``:`s`})`;default:return e.type}}function formatDispatch(e,t){return t.continuationToken?renderCliTaggedLine(e,{message:`resuming session ${t.continuationToken}`,tag:`session`,tone:`info`}):renderCliTaggedLine(e,{message:`starting a new session`,tag:`session`,tone:`info`})}function formatTurnDispatch(e,t){return t.turn.inputResponses!==void 0&&t.turn.message===void 0?renderCliTaggedLine(e,{message:`responding to pending input request${t.turn.inputResponses.length===1?``:`s`}`,tag:`session`,tone:`info`}):formatDispatch(e,t.session)}function formatSessionBoundary(e,t){let n=extractCurrentTurnBoundaryEvent(t),r=extractPendingRuntimeInputRequests(t);switch(n?.type){case`session.waiting`:return[renderCliTaggedLine(e,{message:r.length>0?`waiting for input approval/answer or the next message`:`waiting for the next message`,tag:`session`,tone:`success`})];case`session.completed`:return[renderCliTaggedLine(e,{message:`session completed; the next input starts a new session`,tag:`session`,tone:`success`})];case`session.failed`:{let t=n.data.details&&typeof n.data.details.name==`string`?n.data.details.name:void 0;return[renderCliTaggedLine(e,{message:t?`session failed (${t}): ${n.data.message}`:`session failed: ${n.data.message}`,tag:`session`,tone:`danger`}),renderCliTaggedLine(e,{message:`cleared; the next input starts a new session`,tag:`session`,tone:`warning`})]}default:return[]}}async function waitForInputLine(e,t,n={}){return await new Promise(r=>{let i=e.input,cleanup=()=>{e.off(`close`,handleClose),e.off(`line`,handleLine),i.off(`keypress`,handleKeypress)},handleClose=()=>{cleanup(),r(void 0)},handleLine=e=>{cleanup(),r(e)},handleKeypress=(t,i)=>{n.allowEscape!==!0||i.name!==`escape`||(cleanup(),e.write(null,{ctrl:!0,name:`u`}),r(ESCAPED_RUNTIME_INPUT_PROMPT))};e.setPrompt(t),e.once(`close`,handleClose),e.once(`line`,handleLine),n.allowEscape&&(emitKeypressEvents(i,e),i.on(`keypress`,handleKeypress)),e.prompt()})}async function runTurn(e){let t=e.turn,n={current:createTurnDisplayState()},i=new ReplSubagentStreamManager({displayStateRef:n,serverUrl:e.serverUrl,terminal:e.terminal,theme:e.theme}),ask=async t=>{e.terminal.startPrompt(e.rl,t);let n=await waitForInputLine(e.rl,t,{allowEscape:!0});return e.terminal.stopPrompt(),n};try{for(;;){let a=e.client.getSession();n.current=printDisplayLine({kind:`meta`,line:formatTurnDispatch(e.theme,{session:a,turn:t}),state:n.current,terminal:e.terminal});let o=await e.client.send({inputResponses:t.inputResponses,message:t.message,onEvent(t){n.current=renderTurnEvent({event:t,state:n.current,terminal:e.terminal,theme:e.theme}),t.type===`subagent.called`&&i.subscribe({sessionId:t.data.childSessionId,subagentName:t.data.name})},onResponseStart(t){t.sessionId&&a.sessionId!==t.sessionId&&(n.current=printDisplayLine({kind:`meta`,line:renderCliTaggedLine(e.theme,{message:t.sessionId,tag:`session`,tone:`accent`}),state:n.current,terminal:e.terminal}),n.current=printDisplayLine({kind:`meta`,line:renderCliTaggedLine(e.theme,{message:resolveDevelopmentServerResourceUrl({resource:createEveMessageStreamRoutePath(t.sessionId),serverUrl:e.serverUrl}).toString(),tag:`stream`,tone:`info`}),state:n.current,terminal:e.terminal}))}}),s=extractPendingRuntimeInputRequests(o.events);if(shouldDrainSubagentStreamsOnBoundary(s)){await i.waitForIdle();for(let t of formatSessionBoundary(e.theme,o.events))n.current=printDisplayLine({kind:`meta`,line:t,state:n.current,terminal:e.terminal});return`continue`}let c=await promptForRuntimeInputRequests({ask,print(t){n.current=printDisplayLine({kind:`meta`,line:t,state:n.current,terminal:e.terminal})},requests:s,theme:e.theme});if(c.kind===`aborted`)return`exit`;if(c.kind===`deferred`)return n.current=printDisplayLine({kind:`meta`,line:renderCliTaggedLine(e.theme,{message:`left pending input requests unresolved; send a new message to ignore them`,tag:`session`,tone:`warning`}),state:n.current,terminal:e.terminal}),`continue`;t={inputResponses:c.inputResponses}}}finally{await i.close()}}async function runDevelopmentRepl(e){let t=createDevelopmentTerminal(),n=createCliTheme({color:!0}),r=createInterface({input:process.stdin,output:t.output,terminal:!0});r.on(`SIGINT`,()=>{r.close()});let a=createDevClient({serverUrl:e.serverUrl});try{for(t.print(renderIntro(n,e)),t.print(``);;){t.print(``),t.startPrompt(r,`you> `);let i=await waitForInputLine(r,`you> `);if(t.stopPrompt(),i===void 0)return;let o=parseDevReplInput(i);switch(o.kind!==`empty`&&o.kind!==`exit`&&t.print(``),o.kind){case`empty`:continue;case`help`:t.print(renderIntro(n,e)),t.print(``);continue;case`exit`:return;case`new`:await a.clear(),t.print(renderCliTaggedLine(n,{message:`cleared; the next input starts a new session`,tag:`session`,tone:`warning`})),t.print(``);continue;case`message`:try{if(await runTurn({client:a,rl:r,serverUrl:e.serverUrl,terminal:t,theme:n,turn:{message:o.message}})===`exit`)return}catch(r){isVercelAuthChallenge(r)?t.printError(renderCliTaggedLine(n,{message:formatVercelAuthChallengeMessage({serverUrl:e.serverUrl}),tag:`auth`,tone:`warning`})):t.printError(renderCliTaggedLine(n,{message:toErrorMessage(r),tag:`error`,tone:`danger`}))}t.print(``)}}}finally{await a.close(),r.close(),t.dispose()}}export{createTurnDisplayState,formatContentEvent,formatEvent,renderTurnEvent,runDevelopmentRepl,shouldDrainSubagentStreamsOnBoundary};

package/dist/src/cli/dev/repl/terminal.d.ts

@@ -1,21 +0,0 @@
-import type { Interface } from "node:readline";
-/**
- * Output controller used by the interactive `eve dev` REPL.
- *
- * It keeps the prompt stable by temporarily clearing and restoring it when
- * background server logs write to stdout or stderr.
- */
-export interface DevelopmentTerminal {
-    commitLive(): void;
-    readonly output: NodeJS.WriteStream;
-    dispose(): void;
-    print(message: string): void;
-    printError(message: string): void;
-    startPrompt(rl: Interface, prompt: string): void;
-    stopPrompt(): void;
-    updateLive(message: string): void;
-}
-/**
- * Creates the terminal surface used by the interactive development REPL.
- */
-export declare function createDevelopmentTerminal(): DevelopmentTerminal;

package/dist/src/cli/dev/repl/terminal.js

@@ -1,5 +0,0 @@
-const ansiSgrPattern=RegExp(`${String.fromCodePoint(27)}\\[[0-9;]*m`,`gu`);function chunkToString(e,t){return typeof e==`string`?e:Buffer.from(e).toString(t)}function endsWithLineBreak(e){return e.endsWith(`
-`)||e.endsWith(`\r`)}function stripAnsi(t){return t.replace(ansiSgrPattern,``)}function countRenderedLines(e){let t=process.stdout.columns;return e.split(/\r\n|\r|\n/u).reduce((e,n)=>{let r=Array.from(stripAnsi(n)).length;return t===void 0||t<=0?e+1:e+Math.max(1,Math.ceil(r/t))},0)}function getPromptCursor(e){let t=e.cursor;return typeof t==`number`?t:e.line.length}function createDevelopmentTerminal(){let e=process.stdout.write.bind(process.stdout),t=process.stderr.write.bind(process.stderr),n=null,r=null,writeOriginal=(e,t,n,r,i)=>Reflect.apply(t,e,[n,r,i]),clearPrompt=()=>{n!==null&&writeOriginal(process.stdout,e,`\x1B[2K\r`)},redrawPrompt=()=>{if(n===null)return;let{prompt:t,rl:r}=n,i=getPromptCursor(r),a=r.line.length-i;writeOriginal(process.stdout,e,`${t}${r.line}`),a>0&&writeOriginal(process.stdout,e,`\u001B[${a}D`)},clearLiveBlock=()=>{if(r!==null)for(let t=0;t<r.lineCount;t+=1)writeOriginal(process.stdout,e,`\x1B[1A`),writeOriginal(process.stdout,e,`\x1B[2K\r`)},renderLiveBlock=t=>{let n=t.length===0||!endsWithLineBreak(t);writeOriginal(process.stdout,e,t),n&&writeOriginal(process.stdout,e,`
-`),r={lineCount:countRenderedLines(t),message:t}},printMessage=(e,t,r)=>{let i=n!==null,a=e.length===0||!endsWithLineBreak(e);i&&clearPrompt(),writeOriginal(r,t,e),a&&writeOriginal(r,t,`
-`),i&&redrawPrompt()},patchStream=(e,t)=>(e.write=((i,a,o)=>{if(n===null&&r===null)return writeOriginal(e,t,i,a,o);let s=chunkToString(i,typeof a==`string`?a:void 0),c=r;n!==null&&clearPrompt(),c!==null&&(clearLiveBlock(),r=null);let l=writeOriginal(e,t,i,a,o);return s.length>0&&!endsWithLineBreak(s)&&writeOriginal(e,t,`
-`),c!==null&&renderLiveBlock(c.message),n!==null&&redrawPrompt(),l}),()=>{e.write=t}),i=patchStream(process.stdout,e),a=patchStream(process.stderr,t);return{commitLive(){r=null},dispose(){n=null,r=null,i(),a()},output:new Proxy(process.stdout,{get(t,n,r){if(n===`write`)return(t,n,r)=>writeOriginal(process.stdout,e,t,n,r);let i=Reflect.get(t,n,r);return typeof i==`function`?i.bind(t):i}}),print(t){printMessage(t,e,process.stdout)},printError(e){printMessage(e,t,process.stderr)},startPrompt(e,t){n={prompt:t,rl:e}},stopPrompt(){n=null},updateLive(e){let t=n!==null;t&&clearPrompt(),r!==null&&(clearLiveBlock(),r=null),renderLiveBlock(e),t&&redrawPrompt()}}}export{createDevelopmentTerminal};