@e2edev/e2e 0.11.0 → 0.13.0

package/README.md

@@ -1,7 +1,7 @@
 # e2e
 
 The `e2e` SDK, runner, and CLI for agentic end-to-end testing. Full
-documentation: [e2e-docs.vercel.app](https://e2e-docs.vercel.app).
+documentation: [e2e.mintlify.app](https://e2e.mintlify.app).
 
 ```bash
 pnpm dlx @e2edev/e2e init
@@ -10,7 +10,7 @@ pnpm dlx @e2edev/e2e init
 ## Usage
 
 ```bash
-APP_URL=http://localhost:3000 npx --no-install e2e run
+APP_URL=http://localhost:3000 npx e2e run
 ```
 
 Init adds the runner to `devDependencies`, offers Playwright or agent-device
@@ -62,7 +62,7 @@ ready. On every exit path the runner stops the app, then stops the services in
 reverse order, then runs their `teardown` commands in reverse order, so
 `docker compose down` runs after the migration step and the app are gone.
 Every service option is listed under
-[services](https://e2e-docs.vercel.app/reference/config#services).
+[services](https://e2e.mintlify.app/reference/config#services).
 
 The runner spawns the command, waits until `readyUrl` (defaults to `url`)
 answers with a 200-499 status, and terminates it when the run finishes, fails,
@@ -76,7 +76,7 @@ secrets included, must be passed explicitly through `command.env`. If the app
 never becomes ready the run fails with `APP_UNREACHABLE` and `.e2e/report.json`
 is still written; set `command.log: '.e2e/logs/app.log'` to keep what the
 server printed, since its output is otherwise discarded. Every option is listed under
-[the app under test](https://e2e-docs.vercel.app/reference/config#the-app-under-test).
+[the app under test](https://e2e.mintlify.app/reference/config#the-app-under-test).
 
 ```ts
 import { test } from '@e2edev/playwright';
@@ -127,7 +127,7 @@ directly:
 import { openai } from '@ai-sdk/openai';
 
 export default {
-  agents: { default: { model: openai('gpt-5.4-mini') } },
+  agents: { default: { model: openai('gpt-5.6-luna') } },
 } satisfies E2EConfig;
 ```
 
@@ -168,8 +168,8 @@ read a failing run. `npx skills add tester-army/e2e` installs the same skill
 from the repository. An agent without it can print the text:
 
 ```bash
-npx --no-install e2e guide                # the overview and the topic list
-npx --no-install e2e guide writing-tests  # one topic
+npx e2e guide                # the overview and the topic list
+npx e2e guide writing-tests  # one topic
 ```
 
 ## Current limitations

package/dist/agent/act.d.ts

@@ -1,16 +1,41 @@
 /**
  * Harness-owned dispatch of one `agent.act()` step.
  *
- * The harness opens the step, owns the deadline, the action budget, origin
- * policy, observation redaction, and recording — then hands the step to the
- * configured executor and maps its verdict back onto the runner's error
- * taxonomy. The executor never touches the engine: everything bottoms out in
- * the context built here, on the same accounting core (phases.ts) the
- * locate/judgment tier runs on. Trace-cache participation — replay, live
- * recording, staging — lives beside the dispatch in `StepTraceSession`.
+ * The harness opens the step and hands only the thinking to the configured
+ * executor; everything the executor can touch is built here from four
+ * collaborators, each owning one concern of the step:
+ *
+ * - `StepAccounting`: the clock, the budgets, the hard stop, the metrics.
+ * - `ObservationFeed`: every capture, the settle, the id ring, the pixel decision.
+ * - `ActionDispatcher`: the verb table, target resolution, recording.
+ * - `StepTraceSession`: replay, live recording, and the stage-or-evict decision.
+ *
+ * Capabilities beyond the grammar (secret fills, the trace cache) are modules
+ * over those collaborators, borrowed through a small host each. What stays in
+ * this file is the wiring, the executor's context, and the mapping of the
+ * executor's verdict onto the runner's error taxonomy.
  */
-import type { ActOptions, ActResult } from '../types.ts';
-import type { AgentContext } from './invocation.ts';
+import type { ActOptions, ActResult, AgentErrorCode, JsonValue, Secret } from '../types.ts';
+import type { AgentContext, AgentSelection } from './invocation.ts';
+/** What a dispatch needs of the agent a step runs with; an interactive step supplies its own. */
+export type DispatchAgent = Pick<AgentSelection, 'name' | 'config' | 'executor' | 'agentContext'>;
+/** Everything one dispatched step is, resolved before the step opens. */
+export interface DispatchSpec {
+    /** The public API the step ran as, or `session` for a step a host drives from outside the loop. */
+    readonly api: 'agent.act' | 'agent.assert' | 'session';
+    readonly kind: 'act' | 'assert';
+    readonly instruction: string;
+    readonly params: Readonly<Record<string, JsonValue>> | undefined;
+    /** Secrets declared in the params, by stable name. */
+    readonly secrets: ReadonlyMap<string, Secret>;
+    /** The code a code-less non-passing verdict maps to. */
+    readonly defaultFailureCode: AgentErrorCode;
+    readonly timeout: number | undefined;
+    readonly maxSteps: number | undefined;
+    readonly maxModelCalls: number | undefined;
+    /** The configured agent the call named, if any. */
+    readonly agent: string | undefined;
+}
 /** Runs one `agent.act()` call as a harness-dispatched executor step. */
 export declare function runActStep(runtime: AgentContext, instruction: string, options: ActOptions | undefined, extraArguments?: number): Promise<ActResult>;
 /**
@@ -25,4 +50,13 @@ export declare function runAssertStep(runtime: AgentContext, assertion: string,
     screenshot?: boolean;
     agent?: string;
 } | undefined): Promise<void>;
+/**
+ * Opens the step, tries a cached replay, delegates to the executor when the
+ * cache misses or diverges, settles the verdict, and hands the outcome to the
+ * cache session, which owns the stage-or-evict decision
+ * (`StepTraceSession.conclude`). An assert is a verification step: its
+ * passing is what confirms the traces staged before it at attempt end
+ * (cache/context.ts).
+ */
+export declare function dispatchAgentStep(runtime: AgentContext, spec: DispatchSpec, agent?: DispatchAgent): Promise<ActResult>;
 //# sourceMappingURL=act.d.ts.map

package/dist/agent/act.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"act.d.ts","sourceRoot":"","sources":["../../src/agent/act.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAWH,OAAO,KAAK,EACV,UAAU,EACV,SAAS,EAOV,MAAM,aAAa,CAAC;AAoBrB,OAAO,KAAK,EAAE,YAAY,EAAkB,MAAM,iBAAiB,CAAC;AAoFpE,yEAAyE;AACzE,wBAAsB,UAAU,CAC9B,OAAO,EAAE,YAAY,EACrB,WAAW,EAAE,MAAM,EACnB,OAAO,EAAE,UAAU,GAAG,SAAS,EAC/B,cAAc,SAAI,GACjB,OAAO,CAAC,SAAS,CAAC,CAgBpB;AAED;;;;;GAKG;AACH,wBAAsB,aAAa,CACjC,OAAO,EAAE,YAAY,EACrB,SAAS,EAAE,MAAM,EACjB,OAAO,EAAE;IAAE,OAAO,CAAC,EAAE,MAAM,CAAC;IAAC,MAAM,CAAC,EAAE,OAAO,CAAC;IAAC,UAAU,CAAC,EAAE,OAAO,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,SAAS,GAChG,OAAO,CAAC,IAAI,CAAC,CAsBf"}
+{"version":3,"file":"act.d.ts","sourceRoot":"","sources":["../../src/agent/act.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;GAgBG;AAMH,OAAO,KAAK,EAAE,UAAU,EAAE,SAAS,EAAE,cAAc,EAAE,SAAS,EAAiB,MAAM,EAAE,MAAM,aAAa,CAAC;AAM3G,OAAO,KAAK,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AAUpE,iGAAiG;AACjG,MAAM,MAAM,aAAa,GAAG,IAAI,CAAC,cAAc,EAAE,MAAM,GAAG,QAAQ,GAAG,UAAU,GAAG,cAAc,CAAC,CAAC;AAElG,yEAAyE;AACzE,MAAM,WAAW,YAAY;IAC3B,mGAAmG;IACnG,QAAQ,CAAC,GAAG,EAAE,WAAW,GAAG,cAAc,GAAG,SAAS,CAAC;IACvD,QAAQ,CAAC,IAAI,EAAE,KAAK,GAAG,QAAQ,CAAC;IAChC,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC,MAAM,CAAC,MAAM,EAAE,SAAS,CAAC,CAAC,GAAG,SAAS,CAAC;IACjE,sDAAsD;IACtD,QAAQ,CAAC,OAAO,EAAE,WAAW,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAC9C,wDAAwD;IACxD,QAAQ,CAAC,kBAAkB,EAAE,cAAc,CAAC;IAC5C,QAAQ,CAAC,OAAO,EAAE,MAAM,GAAG,SAAS,CAAC;IACrC,QAAQ,CAAC,QAAQ,EAAE,MAAM,GAAG,SAAS,CAAC;IACtC,QAAQ,CAAC,aAAa,EAAE,MAAM,GAAG,SAAS,CAAC;IAC3C,mDAAmD;IACnD,QAAQ,CAAC,KAAK,EAAE,MAAM,GAAG,SAAS,CAAC;CACpC;AAED,yEAAyE;AACzE,wBAAsB,UAAU,CAC9B,OAAO,EAAE,YAAY,EACrB,WAAW,EAAE,MAAM,EACnB,OAAO,EAAE,UAAU,GAAG,SAAS,EAC/B,cAAc,SAAI,GACjB,OAAO,CAAC,SAAS,CAAC,CAgBpB;AAED;;;;;GAKG;AACH,wBAAsB,aAAa,CACjC,OAAO,EAAE,YAAY,EACrB,SAAS,EAAE,MAAM,EACjB,OAAO,EAAE;IAAE,OAAO,CAAC,EAAE,MAAM,CAAC;IAAC,MAAM,CAAC,EAAE,OAAO,CAAC;IAAC,UAAU,CAAC,EAAE,OAAO,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,SAAS,GAChG,OAAO,CAAC,IAAI,CAAC,CAsBf;AAED;;;;;;;GAOG;AACH,wBAAsB,iBAAiB,CACrC,OAAO,EAAE,YAAY,EACrB,IAAI,EAAE,YAAY,EAGlB,KAAK,GAAE,aAA0C,GAChD,OAAO,CAAC,SAAS,CAAC,CAoBpB"}