feat: add fake phase harness

apps/cli/package.json

@@ -9,7 +9,7 @@
   "scripts": {
     "build": "tsup src/index.ts --format esm --clean",
     "typecheck": "tsc -p ../../tsconfig.typecheck.json --noEmit",
-    "test": "vitest run"
+    "test": "cd ../.. && vitest run --project apps/cli"
   },
   "dependencies": {
     "commander": "12.1.0",

docs/plan.md

@@ -1,4 +1,4 @@
-# Devflow Implementation Plan v3 r4
+# Devflow Implementation Plan v3 r9
 
 ## 0. Document Status
 
@@ -11,6 +11,11 @@
 - r2 applied CC-6 through CC-10.
 - r3 applied CC-11 through CC-15.
 - r4 applies CC-16 through CC-18.
+- r5 applies CC-19.
+- r6 applies CC-20.
+- r7 applies CC-21 through CC-23.
+- r8 applies CC-24 through CC-26.
+- r9 applies CC-27 through CC-28.
 
 ## 1. Stack Decisions
 
@@ -40,7 +45,7 @@
 - Pre-commit:
   - `lefthook`.
   - Runs `biome check --write` on staged files.
-  - Runs `tsc -b --noEmit` on changed packages.
+  - Runs `tsc -p tsconfig.typecheck.json --noEmit`.
   - Runs related Vitest tests on changed packages.
 
 ### 1.3 Database
@@ -819,6 +824,7 @@ export interface TranscriptChunk {
 ### 8.3 Recovery Counters
 
 - `sendPrompt` retry: 2.
+  - Means one initial send plus two adapter-level retries, three physical send attempts max.
 - `resume` retry: 2.
 - `rebootstrap` retry: 1.
 - artifact repair retry: 1.
@@ -882,7 +888,7 @@ const PromptEnvelope = z.object({
 ### 9.3 Rules
 
 - Prompt identity is `dedupKey`.
-- Adapter refuses duplicate `dedupKey` for the same session within a run lifetime.
+- Adapter treats duplicate `dedupKey` for the same session within a run lifetime as idempotent success and does not reprocess the prompt.
 - `attempt` increments only when the engine intentionally re-sends after timeout or repair.
 - Adapter-level retry does not increment attempt.
 - Completion is never inferred from transcript text.
@@ -1152,7 +1158,7 @@ Transitions:
 | `awaiting_approval` | request_changes | `planning` | increment phase attempts |
 | `awaiting_approval` | timeout | `paused` | set `paused_from_state='awaiting_approval'` |
 | `executing` | phase ok, more phases | `executing` | next phase |
-| `executing` | phase needs gate | `awaiting_approval` | request gate |
+| `executing` | normal workflow approval gate | `awaiting_approval` | request gate |
 | `executing` | all phases done | `completed` | emit `run.completed`, write final report |
 | `executing` | unrecoverable error | `failed` | emit `run.failed` |
 | `executing` | manual `pauseRun` | `paused` | set `paused_from_state='executing'` |
@@ -1196,6 +1202,14 @@ Transitions:
 | `awaiting_approval` | reject / abort | `failed` |
 | `awaiting_approval` | request_changes | `running`, attempt + 1 |
 
+Replay rules:
+
+- `phase.started.payload.repair === true` marks that attempt as the single allowed repair attempt. Replaying that attempt MUST use repair instructions, `prompt.repaired`, and must not start a third attempt.
+- Repair replay from `running` may reuse an existing `READY` / bootstrapped session even if `last_prompt_hash` still contains the previous attempt's prompt hash; current-attempt prompt send has not happened yet.
+- If phase state is `validating` and no artifact row exists yet, replay re-reads and validates the current `expectedArtifactPath` instead of treating the state as corruption.
+- If phase state is `validating` and artifact rows already exist for the same phase/path/schema, replay may reuse only an artifact row created at or after the current session `last_prompt_at`; older rows are treated as stale previous-attempt outputs and the file is revalidated.
+- Session bootstrap DB row/state changes and `session.created` / `session.ready` events are written in one DB transaction after adapter start succeeds.
+
 ## 14. Approval State
 
 States:
@@ -1463,6 +1477,7 @@ Human required:
 
 - `artifact_invalid_after_repair`
 - `artifact_timeout_exhausted`
+- `prompt_send_exhausted`
 - `destructive_command_blocked`
 - `secret_access_blocked`
 - `backend_unavailable`
@@ -1486,7 +1501,7 @@ Fatal:
 Mapping:
 
 - recoverable -> retry; exhausted -> human_required.
-- human_required -> run paused and gate created.
+- human_required / recovery gate -> run paused and gate created. This is distinct from normal workflow approval gates in §13.1, which use `awaiting_approval`.
 - fatal -> run failed, sessions disposed, final report best-effort.
 
 ## 19. Concurrent Runs and Crash Recovery
@@ -1721,6 +1736,16 @@ M5+:
 | CC-16 | Prompt hash used phaseId but envelope uses phaseKey | prompt hash uses phaseKey |
 | CC-17 | abortRun transition too narrow | abort from any non-terminal run state |
 | CC-18 | approval pending transition wording conflicted with pause epoch | pending can transition once per pending epoch; paused may unpause to pending |
+| CC-19 | `tsc -b --noEmit` is brittle with TypeScript 5.6 project references on clean worktrees | build still uses `tsc -b`; no-emit verification uses root `tsconfig.typecheck.json` |
+| CC-20 | `sendPrompt` retry count was ambiguous against Temporal activity attempts | §8.3 now states retry budget means initial attempt plus retries; §15.2 remains Temporal-level attempts only |
+| CC-21 | Duplicate prompt dedup handling conflicted with adapter retry idempotency | duplicate `dedupKey` returns idempotent success without reprocessing |
+| CC-22 | Normal workflow approval gates and human-required recovery gates were easy to conflate | §13.1 names normal workflow gates; §18 keeps human_required recovery gates paused |
+| CC-23 | Phase start and event append could diverge under retry/error | phase start and `phase.started` append occur in one DB transaction |
+| CC-24 | Repair attempt replay lost repair prompt identity and one-repair budget | repair attempts are derived from `phase.started.payload.repair`, replay uses repair instructions and `prompt.repaired`, and cannot start attempt 3 |
+| CC-25 | `validating` replay failed if crash happened before artifact row insert | replay revalidates the expected artifact file when state is `validating` but no artifact row exists |
+| CC-26 | Session bootstrap state/events could diverge | session row/state and `session.created` / `session.ready` events are committed in one DB transaction |
+| CC-27 | `validating` replay could reuse stale previous-attempt artifact rows | artifact-row replay requires `artifact.created_at >= tui_sessions.last_prompt_at`; otherwise the file is revalidated |
+| CC-28 | repair `running` replay rejected existing READY sessions with previous attempt prompt hash | current-attempt repair prompt is considered unsent, so replay may reuse the session and send `prompt.repaired` |
 
 ### Future Open Questions
 

packages/core/package.json

@@ -8,8 +8,8 @@
   "types": "./dist/index.d.ts",
   "scripts": {
     "build": "tsup src/index.ts --format esm,cjs --clean && tsc -p tsconfig.build.json",
-    "typecheck": "tsc -b --noEmit",
+    "typecheck": "tsc -p ../../tsconfig.typecheck.json --noEmit",
-    "test": "vitest run"
+    "test": "cd ../.. && vitest run --project packages/core"
   },
   "dependencies": {
     "ajv": "8.17.1",

packages/core/src/run-event.test.ts

@@ -1,9 +1,93 @@
 import { describe, expect, it } from "vitest";
 
-import { RunEventPayloadSchemas, RunEventTypeValues } from "./run-event.js";
+import { RunEvent, RunEventPayloadSchemas, RunEventTypeValues } from "./run-event.js";
 
 describe("run events", () => {
   it("keeps a payload schema for every closed run event type", () => {
     expect(Object.keys(RunEventPayloadSchemas).sort()).toEqual([...RunEventTypeValues].sort());
   });
+
+  it("rejects malformed payloads for structured event families", () => {
+    expect(
+      RunEventPayloadSchemas["prompt.sent"].safeParse({
+        roleId: "implementer",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["artifact.validated"].safeParse({
+        artifactId: "not-a-uuid",
+        hash: "not-a-sha",
+        path: "/tmp/spec.json",
+        schemaId: "dev/spec@1",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["run.paused"].safeParse({
+        cause: "human_required:artifact_repair_failed",
+      }).success,
+    ).toBe(false);
+    expect(RunEventPayloadSchemas["run.resumed"].safeParse({}).success).toBe(false);
+    expect(
+      RunEventPayloadSchemas["approval.resolved"].safeParse({
+        action: "pause",
+        approvalRequestId: "00000000-0000-4000-8000-000000000000",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["approval.resolved"].safeParse({
+        action: "approve",
+        approvalRequestId: "00000000-0000-4000-8000-000000000000",
+      }).success,
+    ).toBe(true);
+    expect(
+      RunEventPayloadSchemas["session.ready"].safeParse({
+        roleId: "implementer",
+        sessionId: "00000000-0000-4000-8000-000000000000",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["session.failed"].safeParse({
+        sessionId: "00000000-0000-4000-8000-000000000000",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["phase.started"].safeParse({
+        attempt: 0,
+        phaseKey: "implement",
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEventPayloadSchemas["artifact.expected"].safeParse({
+        attempt: 0,
+        path: "/tmp/spec.json",
+        schemaId: "dev/spec@1",
+      }).success,
+    ).toBe(false);
+    expect(RunEventPayloadSchemas["phase.skipped"].safeParse({}).success).toBe(false);
+    expect(
+      RunEventPayloadSchemas["review.batch_recorded"].safeParse({
+        attempt: 0,
+        reviewerRole: "reviewer",
+      }).success,
+    ).toBe(false);
+  });
+
+  it("binds exported RunEvent validation to each event type payload schema", () => {
+    expect(
+      RunEvent.safeParse({
+        type: "session.ready",
+        payload: {},
+      }).success,
+    ).toBe(false);
+    expect(
+      RunEvent.safeParse({
+        type: "session.ready",
+        payload: {
+          recoveryAttempts: 0,
+          roleId: "implementer",
+          sessionId: "00000000-0000-4000-8000-000000000000",
+        },
+      }).success,
+    ).toBe(true);
+  });
 });

packages/core/src/run-event.ts

@@ -1,5 +1,7 @@
 import { z } from "zod";
 
+import { ApprovalDecisionAction } from "./enums.js";
+
 export const RunEventTypeValues = [
   "run.created",
   "run.started",
@@ -40,15 +42,146 @@ export const RunEventTypeValues = [
 export const RunEventType = z.enum(RunEventTypeValues);
 export type RunEventType = z.infer<typeof RunEventType>;
 
-const payloadSchema = z.record(z.unknown());
+const uuid = z.string().uuid();
+const sha256 = z.string().regex(/^[a-f0-9]{64}$/);
+const nonEmptyString = z.string().min(1);
+const phaseAttempt = z.number().int().positive();
 
-export const RunEventPayloadSchemas = Object.freeze(
+const looseObject = z.object({}).passthrough();
-  Object.fromEntries(RunEventTypeValues.map((type) => [type, payloadSchema])),
+const phasePayload = z
-) as Readonly<Record<RunEventType, typeof payloadSchema>>;
+  .object({
+    phaseKey: nonEmptyString,
+    attempt: phaseAttempt,
+  })
+  .passthrough();
+const promptPayload = z
+  .object({
+    roleId: nonEmptyString,
+    dedupKey: sha256,
+  })
+  .passthrough();
+const artifactWaitPayload = z
+  .object({
+    path: nonEmptyString,
+    schemaId: nonEmptyString,
+    attempt: phaseAttempt,
+  })
+  .passthrough();
+const artifactValidationPayload = z
+  .object({
+    artifactId: uuid,
+    hash: sha256,
+    path: nonEmptyString,
+    schemaId: nonEmptyString,
+  })
+  .passthrough();
+const sessionBasePayload = z
+  .object({
+    sessionId: uuid,
+    roleId: nonEmptyString,
+  })
+  .passthrough();
+const sessionPromptPayload = sessionBasePayload.extend({
+  dedupKey: sha256,
+});
+const sessionRecoveryPayload = sessionBasePayload.extend({
+  recoveryAttempts: z.number().int().nonnegative(),
+});
+const approvalRequestedPayload = z
+  .object({
+    approvalRequestId: uuid,
+    approvalIdempotencyKey: nonEmptyString,
+    gateKey: nonEmptyString,
+  })
+  .passthrough();
+const approvalResolvedPayload = z
+  .object({
+    action: ApprovalDecisionAction,
+    approvalRequestId: uuid,
+  })
+  .passthrough();
+const commandPayload = z
+  .object({
+    commandId: uuid,
+  })
+  .passthrough();
+const findingVerifierResolvedPayload = z
+  .object({
+    findingId: uuid,
+  })
+  .passthrough();
+const backtestIterationPayload = z
+  .object({
+    iterationId: uuid,
+  })
+  .passthrough();
+const reviewBatchRecordedPayload = z
+  .object({
+    reviewerRole: nonEmptyString,
+    attempt: phaseAttempt,
+  })
+  .passthrough();
+const runPausedPayload = z
+  .object({
+    cause: nonEmptyString,
+    pausedFromState: nonEmptyString,
+  })
+  .passthrough();
 
-export const RunEvent = z.object({
+export const RunEventPayloadSchemas = Object.freeze({
+  "run.created": looseObject,
+  "run.started": looseObject,
+  "run.paused": runPausedPayload,
+  "run.resumed": runPausedPayload.pick({ cause: true }).passthrough(),
+  "run.completed": looseObject,
+  "run.failed": looseObject,
+  "run.aborted": looseObject,
+  "phase.started": phasePayload,
+  "phase.completed": phasePayload,
+  "phase.failed": phasePayload.extend({ reason: nonEmptyString.optional() }),
+  "phase.skipped": phasePayload,
+  "prompt.sent": promptPayload,
+  "prompt.repaired": promptPayload,
+  "artifact.expected": artifactWaitPayload,
+  "artifact.validated": artifactValidationPayload,
+  "artifact.invalid": artifactValidationPayload.extend({ errors: z.array(z.unknown()) }),
+  "artifact.timeout": artifactWaitPayload,
+  "approval.requested": approvalRequestedPayload,
+  "approval.resolved": approvalResolvedPayload,
+  "session.created": sessionBasePayload.extend({ backend: nonEmptyString }),
+  "session.ready": sessionRecoveryPayload,
+  "session.busy": sessionPromptPayload,
+  "session.idle": sessionPromptPayload,
+  "session.crashed": sessionRecoveryPayload,
+  "session.recovered": sessionRecoveryPayload,
+  "session.failed": sessionBasePayload,
+  "command.started": commandPayload,
+  "command.completed": commandPayload,
+  "command.failed": commandPayload,
+  "review.batch_recorded": reviewBatchRecordedPayload,
+  "finding.verifier_resolved": findingVerifierResolvedPayload,
+  "backtest.iteration_started": backtestIterationPayload,
+  "backtest.iteration_completed": backtestIterationPayload,
+  "backtest.objective_evaluated": backtestIterationPayload,
+} satisfies Record<RunEventType, z.ZodTypeAny>) as Readonly<Record<RunEventType, z.ZodTypeAny>>;
+
+export const RunEvent = z
+  .object({
     type: RunEventType,
-  payload: payloadSchema,
+    payload: z.unknown(),
+  })
+  .superRefine((event, ctx) => {
+    const payload = RunEventPayloadSchemas[event.type].safeParse(event.payload);
+    if (payload.success) {
+      return;
+    }
+
+    for (const issue of payload.error.issues) {
+      ctx.addIssue({
+        ...issue,
+        path: ["payload", ...issue.path],
+      });
+    }
   });
 
 export type RunEvent = z.infer<typeof RunEvent>;

packages/db/package.json

@@ -8,8 +8,8 @@
   "types": "./dist/index.d.ts",
   "scripts": {
     "build": "tsup src/index.ts --format esm,cjs --clean && tsc -p tsconfig.build.json",
-    "typecheck": "tsc -b --noEmit",
+    "typecheck": "tsc -p ../../tsconfig.typecheck.json --noEmit",
-    "test": "vitest run"
+    "test": "cd ../.. && vitest run --project packages/db"
   },
   "dependencies": {
     "@devflow/core": "workspace:*",

packages/db/src/index.ts

@@ -1,3 +1,4 @@
 export { createDbClient, type DbClient } from "./client.js";
+export * from "./repositories/run-event.js";
 export * from "./repositories/transcript.js";
 export * from "./schema/index.js";

packages/db/src/repositories/run-event.ts

@@ -0,0 +1,278 @@
+import {
+  DevflowError,
+  RunEventPayloadSchemas,
+  RunEventType,
+  type RunEventType as RunEventTypeName,
+  canonicalize,
+} from "@devflow/core";
+import { and, desc, eq, sql } from "drizzle-orm";
+
+import type { DbClient } from "../client.js";
+import { runEvents, runPhases } from "../schema/index.js";
+
+type Database = DbClient["db"];
+type TransactionDatabase = Parameters<Parameters<Database["transaction"]>[0]>[0];
+
+export interface AppendRunEventInput {
+  runId: string;
+  phaseId?: string;
+  type: RunEventTypeName;
+  payload: Record<string, unknown>;
+  idempotencyKey: string;
+}
+
+export interface RunEventRow {
+  id: bigint;
+  runId: string;
+  phaseId: string | null;
+  seq: bigint;
+  type: string;
+  payload: unknown;
+  idempotencyKey: string;
+  ts: Date;
+}
+
+export class RunEventRepository {
+  constructor(private readonly db: Database) {}
+
+  async append(input: AppendRunEventInput): Promise<RunEventRow> {
+    return this.db.transaction(async (tx) => this.appendInTransaction(tx, input));
+  }
+
+  async appendInTransaction(
+    tx: TransactionDatabase,
+    input: AppendRunEventInput,
+  ): Promise<RunEventRow> {
+    const type = RunEventType.parse(input.type);
+    const payload = RunEventPayloadSchemas[type].parse(input.payload) as Record<string, unknown>;
+    if (isPhaseScopedEvent(type) && input.phaseId === undefined) {
+      throw new DevflowError("Run event phase id is required for phase-scoped event", {
+        class: "fatal",
+        code: "internal_state_corruption",
+        runId: input.runId,
+      });
+    }
+    if (input.idempotencyKey.length === 0) {
+      throw new DevflowError("Run event idempotency key is required", {
+        class: "fatal",
+        code: "internal_state_corruption",
+        runId: input.runId,
+        ...(input.phaseId === undefined ? {} : { phaseId: input.phaseId }),
+      });
+    }
+    const expectedIdempotencyKey = expectedRunEventIdempotencyKey(input, type, payload);
+    if (expectedIdempotencyKey !== undefined && input.idempotencyKey !== expectedIdempotencyKey) {
+      throw new DevflowError("Run event idempotency key does not match event contract", {
+        class: "fatal",
+        code: "internal_state_corruption",
+        runId: input.runId,
+        ...(input.phaseId === undefined ? {} : { phaseId: input.phaseId }),
+      });
+    }
+
+    await tx.execute(
+      sql`SELECT pg_advisory_xact_lock(hashtext('devflow:run-events'), hashtext(${input.runId}))`,
+    );
+
+    if (input.phaseId !== undefined) {
+      const [phase] = await tx
+        .select({ id: runPhases.id })
+        .from(runPhases)
+        .where(and(eq(runPhases.id, input.phaseId), eq(runPhases.runId, input.runId)))
+        .limit(1);
+      if (phase === undefined) {
+        throw new DevflowError("Run event phase does not belong to run", {
+          class: "fatal",
+          code: "internal_state_corruption",
+          runId: input.runId,
+          phaseId: input.phaseId,
+        });
+      }
+    }
+
+    const existing = await tx
+      .select()
+      .from(runEvents)
+      .where(
+        and(eq(runEvents.runId, input.runId), eq(runEvents.idempotencyKey, input.idempotencyKey)),
+      )
+      .limit(1);
+
+    if (existing[0] !== undefined) {
+      assertIdempotentReplayMatches(input, type, payload, existing[0]);
+      return existing[0];
+    }
+
+    const latest = await tx
+      .select({ seq: runEvents.seq })
+      .from(runEvents)
+      .where(eq(runEvents.runId, input.runId))
+      .orderBy(desc(runEvents.seq))
+      .limit(1);
+    const seq = (latest[0]?.seq ?? 0n) + 1n;
+
+    const inserted = await tx
+      .insert(runEvents)
+      .values({
+        runId: input.runId,
+        phaseId: input.phaseId,
+        seq,
+        type,
+        payload,
+        idempotencyKey: input.idempotencyKey,
+      })
+      .returning();
+
+    const event = inserted[0];
+    if (event === undefined) {
+      throw new DevflowError("Run event insert returned no row", {
+        class: "fatal",
+        code: "internal_state_corruption",
+        runId: input.runId,
+        ...(input.phaseId === undefined ? {} : { phaseId: input.phaseId }),
+      });
+    }
+
+    return event;
+  }
+}
+
+function isPhaseScopedEvent(type: RunEventTypeName): boolean {
+  return (
+    type.startsWith("phase.") || type.startsWith("artifact.") || type === "review.batch_recorded"
+  );
+}
+
+function assertIdempotentReplayMatches(
+  input: AppendRunEventInput,
+  type: RunEventTypeName,
+  payload: Record<string, unknown>,
+  existing: RunEventRow,
+) {
+  const sameType = existing.type === type;
+  const samePhase = !isPhaseScopedEvent(type) || existing.phaseId === (input.phaseId ?? null);
+  const samePayload = canonicalize(normalizeJson(existing.payload)) === canonicalize(payload);
+  if (sameType && samePhase && samePayload) {
+    return;
+  }
+
+  throw new DevflowError("Run event idempotency key replay does not match existing event", {
+    class: "fatal",
+    code: "internal_state_corruption",
+    runId: input.runId,
+    ...(input.phaseId === undefined ? {} : { phaseId: input.phaseId }),
+  });
+}
+
+function expectedRunEventIdempotencyKey(
+  input: AppendRunEventInput,
+  type: RunEventTypeName,
+  payload: Record<string, unknown>,
+): string | undefined {
+  switch (type) {
+    case "run.created":
+    case "run.started":
+    case "run.completed":
+    case "run.failed":
+    case "run.aborted":
+      return `${type}:${input.runId}`;
+    case "run.paused":
+    case "run.resumed":
+      return `${type}:${input.runId}:${stringPayload(payload, "cause")}`;
+    case "phase.started":
+    case "phase.completed":
+    case "phase.failed":
+    case "phase.skipped":
+      return `${type}:${requiredPhaseId(input)}:${numberPayload(payload, "attempt")}`;
+    case "prompt.sent":
+    case "prompt.repaired":
+      return `${type}:${stringPayload(payload, "dedupKey")}`;
+    case "artifact.expected":
+    case "artifact.timeout":
+      return `${type}:${requiredPhaseId(input)}:${numberPayload(payload, "attempt")}:${stringPayload(payload, "path")}`;
+    case "artifact.validated":
+    case "artifact.invalid":
+      return `${type}:${requiredPhaseId(input)}:${stringPayload(payload, "path")}:${stringPayload(payload, "hash")}`;
+    case "approval.requested":
+      return `approval.requested:${stringPayload(payload, "approvalIdempotencyKey")}`;
+    case "approval.resolved":
+      return `approval.resolved:${stringPayload(payload, "approvalRequestId")}:${stringPayload(payload, "action")}`;
+    case "session.created":
+    case "session.failed":
+      return `${type}:${stringPayload(payload, "sessionId")}`;
+    case "session.busy":
+    case "session.idle":
+      return `${type}:${stringPayload(payload, "sessionId")}:${stringPayload(payload, "dedupKey")}`;
+    case "session.ready":
+    case "session.crashed":
+    case "session.recovered":
+      return `${type}:${stringPayload(payload, "sessionId")}:${numberPayload(payload, "recoveryAttempts")}`;
+    case "command.started":
+    case "command.completed":
+    case "command.failed":
+      return `${type}:${stringPayload(payload, "commandId")}`;
+    case "review.batch_recorded":
+      return `review.batch_recorded:${requiredPhaseId(input)}:${stringPayload(payload, "reviewerRole")}:${numberPayload(payload, "attempt")}`;
+    case "finding.verifier_resolved":
+      return `finding.verifier_resolved:${stringPayload(payload, "findingId")}`;
+    case "backtest.iteration_started":
+    case "backtest.iteration_completed":
+    case "backtest.objective_evaluated":
+      return `${type}:${stringPayload(payload, "iterationId")}`;
+    default:
+      return undefined;
+  }
+}
+
+function requiredPhaseId(input: AppendRunEventInput): string {
+  if (input.phaseId === undefined) {
+    throw new DevflowError("Run event phase id is required for idempotency key", {
+      class: "fatal",
+      code: "internal_state_corruption",
+      runId: input.runId,
+    });
+  }
+
+  return input.phaseId;
+}
+
+function stringPayload(payload: Record<string, unknown>, key: string): string {
+  const value = payload[key];
+  if (typeof value !== "string" || value.length === 0) {
+    throw new DevflowError(`Run event payload is missing string field ${key}`, {
+      class: "fatal",
+      code: "internal_state_corruption",
+    });
+  }
+
+  return value;
+}
+
+function numberPayload(payload: Record<string, unknown>, key: string): number {
+  const value = payload[key];
+  if (typeof value !== "number" || !Number.isInteger(value)) {
+    throw new DevflowError(`Run event payload is missing integer field ${key}`, {
+      class: "fatal",
+      code: "internal_state_corruption",
+    });
+  }
+
+  return value;
+}
+
+function normalizeJson(value: unknown): unknown {
+  if (Array.isArray(value)) {
+    return value.map((item) => normalizeJson(item));
+  }
+
+  if (value !== null && typeof value === "object") {
+    return Object.fromEntries(
+      Object.entries(value as Record<string, unknown>).map(([key, child]) => [
+        key,
+        normalizeJson(child),
+      ]),
+    );
+  }
+
+  return value;
+}

packages/run-engine/package.json

@@ -0,0 +1,20 @@
+{
+  "name": "@devflow/run-engine",
+  "version": "0.0.0",
+  "private": true,
+  "type": "module",
+  "main": "./dist/index.cjs",
+  "module": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "scripts": {
+    "build": "tsup src/index.ts --format esm,cjs --clean && tsc -p tsconfig.build.json",
+    "typecheck": "tsc -p ../../tsconfig.typecheck.json --noEmit",
+    "test": "cd ../.. && vitest run --project packages/run-engine"
+  },
+  "dependencies": {
+    "@devflow/core": "workspace:*",
+    "@devflow/db": "workspace:*",
+    "@devflow/session": "workspace:*",
+    "drizzle-orm": "0.45.2"
+  }
+}

packages/run-engine/src/index.ts

@@ -0,0 +1,2 @@
+export * from "./fake-phase-harness.js";
+export * from "./run-event-repository.js";

packages/run-engine/src/run-event-repository.ts

@@ -0,0 +1 @@
+export { RunEventRepository } from "@devflow/db";

packages/run-engine/tsconfig.build.json

@@ -0,0 +1,10 @@
+{
+  "extends": "./tsconfig.json",
+  "compilerOptions": {
+    "composite": false,
+    "emitDeclarationOnly": true,
+    "noEmit": false
+  },
+  "references": [],
+  "exclude": ["src/**/*.test.ts"]
+}

packages/run-engine/tsconfig.json

@@ -0,0 +1,10 @@
+{
+  "extends": "../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "dist",
+    "types": ["node", "vitest"]
+  },
+  "include": ["src/**/*.ts"],
+  "references": [{ "path": "../core" }, { "path": "../db" }, { "path": "../session" }]
+}

packages/session/package.json

@@ -9,7 +9,7 @@
   "scripts": {
     "build": "tsup src/index.ts --format esm,cjs --clean && tsc -p tsconfig.build.json",
     "typecheck": "tsc -p ../../tsconfig.typecheck.json --noEmit",
-    "test": "vitest run"
+    "test": "cd ../.. && vitest run --project packages/session"
   },
   "dependencies": {
     "@devflow/core": "workspace:*"

packages/session/src/fake.test.ts

@@ -97,7 +97,25 @@ describe("FakeSessionAdapter", () => {
     expect(chunks.every((chunk, index) => chunk.seq === BigInt(index + 1))).toBe(true);
   });
 
-  it("refuses duplicate prompt dedup keys for the same session", async () => {
+  it("classifies unsupported backend as human-required backend_unavailable", async () => {
+    const fixtureRoot = makeFixtureRoot();
+    tempRoots.push(fixtureRoot);
+    const adapter = new FakeSessionAdapter({ fixtureRoot, writeDelayMs: 0 });
+
+    await expect(
+      adapter.start({
+        runId,
+        roleId: "implementer",
+        backend: "codex",
+        cwd: fixtureRoot,
+      }),
+    ).rejects.toMatchObject({
+      class: "human_required",
+      code: "backend_unavailable",
+    });
+  });
+
+  it("treats duplicate prompt dedup keys as idempotent success without reprocessing", async () => {
     const fixtureRoot = makeFixtureRoot();
     tempRoots.push(fixtureRoot);
     const adapter = new FakeSessionAdapter({ fixtureRoot, writeDelayMs: 0 });
@@ -116,13 +134,12 @@ describe("FakeSessionAdapter", () => {
 
     await adapter.sendPrompt(handle, first);
 
-    await expect(adapter.sendPrompt(handle, duplicate)).rejects.toMatchObject({
+    await expect(adapter.sendPrompt(handle, duplicate)).resolves.toEqual({ promptId: dedupKey });
-      code: "duplicate_prompt_dedup_key",
-    });
     await waitForFile(first.expectedArtifact);
+    expect(existsSync(duplicate.expectedArtifact)).toBe(false);
   });
 
-  it("preserves prompt dedup history across crash and rebootstrap recovery", async () => {
+  it("records dedup history only after a fake prompt is accepted", async () => {
     const fixtureRoot = makeFixtureRoot();
     tempRoots.push(fixtureRoot);
     const adapter = new FakeSessionAdapter({ fixtureRoot, writeDelayMs: 0 });
@@ -141,7 +158,7 @@ describe("FakeSessionAdapter", () => {
       code: "prompt_send_transient",
     });
     await expect(adapter.sendPrompt(handle, crash)).rejects.toMatchObject({
-      code: "duplicate_prompt_dedup_key",
+      code: "prompt_send_transient",
     });
 
     const ok = envelope({ dedupKey: "d".repeat(64) });
@@ -150,9 +167,7 @@ describe("FakeSessionAdapter", () => {
     await waitForFile(ok.expectedArtifact);
     await adapter.rebootstrap(handle);
 
-    await expect(adapter.sendPrompt(handle, ok)).rejects.toMatchObject({
+    await expect(adapter.sendPrompt(handle, ok)).resolves.toEqual({ promptId: "d".repeat(64) });
-      code: "duplicate_prompt_dedup_key",
-    });
   });
 
   it("rejects prompts whose run or role do not match the session", async () => {
@@ -207,6 +222,26 @@ describe("FakeSessionAdapter", () => {
     expect(existsSync(prompt.expectedArtifact)).toBe(false);
   });
 
+  it("does not record dedup history when fixture resolution fails", async () => {
+    const fixtureRoot = mkdtempSync(join(tmpdir(), "devflow-empty-fake-fixtures-"));
+    tempRoots.push(fixtureRoot);
+    const adapter = new FakeSessionAdapter({ fixtureRoot, writeDelayMs: 0 });
+    const handle = await adapter.start({
+      runId,
+      roleId: "implementer",
+      backend: "fake",
+      cwd: fixtureRoot,
+    });
+    const missing = envelope({ dedupKey: "f".repeat(64) });
+
+    await expect(adapter.sendPrompt(handle, missing)).rejects.toMatchObject({
+      code: "fake_fixture_missing",
+    });
+    await expect(adapter.sendPrompt(handle, missing)).rejects.toMatchObject({
+      code: "fake_fixture_missing",
+    });
+  });
+
   it("supports invalid, timeout, and crash sentinel scenarios", async () => {
     const fixtureRoot = makeFixtureRoot();
     tempRoots.push(fixtureRoot);

packages/session/src/fake.ts

@@ -21,7 +21,7 @@ import type {
 
 export interface FakeSessionAdapterOptions {
   fixtureRoot?: string;
-  writeDelayMs?: number;
+  writeDelayMs?: number | ((envelope: PromptEnvelope) => number);
   sessionIdFactory?: () => string;
   now?: () => Date;
 }
@@ -40,14 +40,16 @@ interface FakeSessionRecord {
 
 export class FakeSessionAdapter implements SessionAdapter {
   private readonly fixtureRoot: string;
-  private readonly writeDelayMs: number;
+  private readonly writeDelayMs: (envelope: PromptEnvelope) => number;
   private readonly sessionIdFactory: () => string;
   private readonly now: () => Date;
   private readonly sessions = new Map<string, FakeSessionRecord>();
 
   constructor(options: FakeSessionAdapterOptions = {}) {
     this.fixtureRoot = options.fixtureRoot ?? defaultFixtureRoot();
-    this.writeDelayMs = options.writeDelayMs ?? 50;
+    const writeDelayMs = options.writeDelayMs;
+    this.writeDelayMs =
+      typeof writeDelayMs === "function" ? writeDelayMs : () => writeDelayMs ?? 50;
     this.sessionIdFactory = options.sessionIdFactory ?? randomUUID;
     this.now = options.now ?? (() => new Date());
   }
@@ -55,7 +57,7 @@ export class FakeSessionAdapter implements SessionAdapter {
   async start(input: StartInput): Promise<SessionHandle> {
     if (input.backend !== "fake") {
       throw new DevflowError("FakeSessionAdapter only supports the fake backend", {
-        class: "fatal",
+        class: "human_required",
         code: "backend_unavailable",
         runId: input.runId,
       });
@@ -87,15 +89,10 @@ export class FakeSessionAdapter implements SessionAdapter {
       });
     }
     if (record.sentDedupKeys.has(envelope.dedupKey)) {
-      throw new DevflowError("Duplicate prompt dedup key refused by fake session", {
+      return { promptId: envelope.dedupKey };
-        class: "recoverable",
-        code: "duplicate_prompt_dedup_key",
-        runId: record.runId,
-      });
     }
 
     const scenarioName = scenarioFromInstructions(envelope.instructions);
-    record.sentDedupKeys.add(envelope.dedupKey);
 
     if (scenarioName === "crash") {
       this.appendTranscript(record, `[fake] received prompt ${envelope.uuid}; crashing`);
@@ -107,6 +104,7 @@ export class FakeSessionAdapter implements SessionAdapter {
     }
 
     if (scenarioName === "timeout") {
+      record.sentDedupKeys.add(envelope.dedupKey);
       this.appendTranscript(record, `[fake] received prompt ${envelope.uuid}; timeout`);
       return { promptId: envelope.dedupKey };
     }
@@ -121,9 +119,12 @@ export class FakeSessionAdapter implements SessionAdapter {
             envelope.runId,
           );
 
+    record.sentDedupKeys.add(envelope.dedupKey);
+
+    const writeDelayMs = this.writeDelayMs(envelope);
     this.appendTranscript(
       record,
-      `[fake] received prompt ${envelope.uuid}; will write ${envelope.expectedArtifact} in ${this.writeDelayMs}ms`,
+      `[fake] received prompt ${envelope.uuid}; will write ${envelope.expectedArtifact} in ${writeDelayMs}ms`,
     );
     const timer = setTimeout(() => {
       record.timers.delete(timer);
@@ -145,7 +146,7 @@ export class FakeSessionAdapter implements SessionAdapter {
         return;
       }
       this.appendTranscript(record, `[fake] wrote artifact ${envelope.expectedArtifact}`);
-    }, this.writeDelayMs);
+    }, writeDelayMs);
     record.timers.add(timer);
 
     return { promptId: envelope.dedupKey };

pnpm-lock.yaml

@@ -100,6 +100,21 @@ importers:
         specifier: 8.20.0
         version: 8.20.0
 
+  packages/run-engine:
+    dependencies:
+      '@devflow/core':
+        specifier: workspace:*
+        version: link:../core
+      '@devflow/db':
+        specifier: workspace:*
+        version: link:../db
+      '@devflow/session':
+        specifier: workspace:*
+        version: link:../session
+      drizzle-orm:
+        specifier: 0.45.2
+        version: 0.45.2(@types/pg@8.20.0)(pg@8.20.0)
+
   packages/session:
     dependencies:
       '@devflow/core':

tsconfig.json

@@ -8,6 +8,7 @@
   "references": [
     { "path": "./packages/core" },
     { "path": "./packages/db" },
+    { "path": "./packages/run-engine" },
     { "path": "./packages/session" },
     { "path": "./apps/cli" }
   ]

tsconfig.typecheck.json

@@ -10,6 +10,7 @@
     "paths": {
       "@devflow/core": ["packages/core/src/index.ts"],
       "@devflow/db": ["packages/db/src/index.ts"],
+      "@devflow/run-engine": ["packages/run-engine/src/index.ts"],
       "@devflow/session": ["packages/session/src/index.ts"]
     }
   },

vitest.workspace.ts

@@ -1,39 +1,31 @@
+import { resolve } from "node:path";
+import { fileURLToPath } from "node:url";
 import { defineWorkspace } from "vitest/config";
 
+const root = fileURLToPath(new URL(".", import.meta.url));
+const alias = {
+  "@devflow/core": resolve(root, "packages/core/src/index.ts"),
+  "@devflow/db": resolve(root, "packages/db/src/index.ts"),
+  "@devflow/run-engine": resolve(root, "packages/run-engine/src/index.ts"),
+  "@devflow/session": resolve(root, "packages/session/src/index.ts"),
+};
+
+function nodeProject(name: string, include: string[]) {
+  return {
+    resolve: { alias },
+    test: {
+      name,
+      include,
+      environment: "node",
+    },
+  };
+}
+
 export default defineWorkspace([
-  {
+  nodeProject("root", ["tests/**/*.test.ts"]),
-    test: {
+  nodeProject("packages/db", ["packages/db/src/**/*.test.ts"]),
-      name: "root",
+  nodeProject("packages/core", ["packages/core/src/**/*.test.ts"]),
-      include: ["tests/**/*.test.ts"],
+  nodeProject("packages/session", ["packages/session/src/**/*.test.ts"]),
-      environment: "node",
+  nodeProject("packages/run-engine", ["packages/run-engine/src/**/*.test.ts"]),
-    },
+  nodeProject("apps/cli", ["apps/cli/src/**/*.test.ts"]),
-  },
-  {
-    test: {
-      name: "packages/db",
-      include: ["packages/db/src/**/*.test.ts"],
-      environment: "node",
-    },
-  },
-  {
-    test: {
-      name: "packages/core",
-      include: ["packages/core/src/**/*.test.ts"],
-      environment: "node",
-    },
-  },
-  {
-    test: {
-      name: "packages/session",
-      include: ["packages/session/src/**/*.test.ts"],
-      environment: "node",
-    },
-  },
-  {
-    test: {
-      name: "apps/cli",
-      include: ["apps/cli/src/**/*.test.ts"],
-      environment: "node",
-    },
-  },
 ]);