@oneharness/sdk 0.12.1 → 0.13.0

package/dist/generated/history.d.ts

@@ -8,7 +8,7 @@ export type HistoryRecord = ({
     [k: string]: unknown;
 } | ({
     error: string;
-    schema_version?: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | undefined;
+    schema_version?: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | undefined;
     [k: string]: unknown;
 } & ({
     status?: "nonzero" | "timeout" | "cancelled" | "spawn-error" | "skipped" | undefined;
@@ -20,7 +20,7 @@ export type HistoryRecord = ({
     work?: null | undefined;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.7" | "1.8" | undefined;
+    schema_version?: "1.7" | "1.8" | "1.9" | undefined;
     status?: "nonzero" | "timeout" | "cancelled" | undefined;
     work: "done" | "none";
     [k: string]: unknown;
@@ -29,31 +29,37 @@ export type HistoryRecord = ({
     [k: string]: unknown;
 } | {
     observed_model: string;
-    schema_version?: "1.8" | undefined;
+    schema_version?: "1.8" | "1.9" | undefined;
     [k: string]: unknown;
 }) & ({
     status?: "ok" | "nonzero" | "timeout" | "spawn-error" | "skipped" | "planned" | undefined;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | undefined;
+    schema_version?: "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | undefined;
     [k: string]: unknown;
 }) & (({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.5" | "1.6" | "1.7" | "1.8" | undefined;
+    schema_version?: "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | undefined;
     [k: string]: unknown;
 }) & ({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "model_mismatch" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "model_mismatch" | "server_overloaded" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.6" | "1.7" | "1.8" | undefined;
+    schema_version?: "1.6" | "1.7" | "1.8" | "1.9" | undefined;
     [k: string]: unknown;
 }) & ({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "server_overloaded" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.8" | undefined;
+    schema_version?: "1.8" | "1.9" | undefined;
+    [k: string]: unknown;
+}) & ({
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | null;
+    [k: string]: unknown;
+} | {
+    schema_version?: "1.9" | undefined;
     [k: string]: unknown;
 })) & ({
     duration_ms: number;
@@ -169,7 +175,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -469,7 +475,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -596,7 +602,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -723,7 +729,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -850,7 +856,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
+    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -1317,13 +1323,13 @@ export type TimingSource = "provider_measured" | "stdout_observed";
  * harness's output. It is the single source for the `failure_kind` contract
  * value: serialized as the snake_case token a consumer reads in the report
  * (`auth`, `rate_limit`, `model_not_found`, `quota`, `session_not_found`,
- * `tool_deferred`), so the wire shape is unchanged — modeling it as an enum
+ * `tool_deferred`, `server_overloaded`), so the wire shape is unchanged — modeling it as an enum
  * keeps a misspelled or invalid kind unrepresentable and gives every
  * producer/consumer (classifier, `is_failure`, the fallback fall-through rule,
  * the report, history) one definition to share instead of scattered string
  * literals.
  */
-export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch";
+export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded";
 /**
  * The outcome of attempting to run one harness.
  */
@@ -1336,8 +1342,9 @@ export type Status = "ok" | "nonzero" | "timeout" | "cancelled" | "spawn-error"
  *
  * - **Tool events.** A recorded tool call is the harness acting on the task.
  * - **Usage accounting.** [`Usage::reports_billed_work`][billed] — the same definition
- *   `signals::record_reports_work` classifies a raw harness record with, so the
- *   two readings of "billed" are one contract with one implementation.
+ *   `signals::record_work_evidence` reads a raw harness record with, and it answers
+ *   in this same type, so the two readings of "billed" are one contract with one
+ *   implementation.
  *
  * It is also a **published reading**, not only an internal one: a run that
  * failed with nothing to classify carries it as [`RunResult::work`] and in its

package/dist/generated/options.d.ts

@@ -130,6 +130,10 @@ export interface RunOptions {
      * Max retries when a response fails schema validation.
      */
     schemaMaxRetries?: number | undefined;
+    /**
+     * Max retries for a zero-work Codex server-overloaded refusal.
+     */
+    serverOverloadedMaxRetries?: number | undefined;
     session?: string | undefined;
     /**
      * Directory the `session` store lives in.

package/dist/generated/run-stream-envelope.d.ts

@@ -79,19 +79,19 @@ export type ControlVerb = "interrupt";
  * new variant is a report `schema_version` bump like any other enum value, since
  * a consumer matching exhaustively learns of it only from the version.
  */
-export type FallThroughReason = "not-installed" | "spawn-error" | "auth" | "quota" | "session-not-found" | "untrusted-directory" | "input-too-large" | "model-mismatch" | "model-not-found" | "rate-limit";
+export type FallThroughReason = "not-installed" | "spawn-error" | "auth" | "quota" | "session-not-found" | "untrusted-directory" | "input-too-large" | "model-mismatch" | "model-not-found" | "rate-limit" | "server-overloaded";
 /**
  * The normalized, closed set of failure reasons oneharness can classify from a
  * harness's output. It is the single source for the `failure_kind` contract
  * value: serialized as the snake_case token a consumer reads in the report
  * (`auth`, `rate_limit`, `model_not_found`, `quota`, `session_not_found`,
- * `tool_deferred`), so the wire shape is unchanged — modeling it as an enum
+ * `tool_deferred`, `server_overloaded`), so the wire shape is unchanged — modeling it as an enum
  * keeps a misspelled or invalid kind unrepresentable and gives every
  * producer/consumer (classifier, `is_failure`, the fallback fall-through rule,
  * the report, history) one definition to share instead of scattered string
  * literals.
  */
-export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch";
+export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded";
 /**
  * How a harness emits its result, which decides how `text` is extracted.
  *
@@ -144,8 +144,9 @@ export type ExecutionTelemetry = {
  *
  * - **Tool events.** A recorded tool call is the harness acting on the task.
  * - **Usage accounting.** [`Usage::reports_billed_work`][billed] — the same definition
- *   `signals::record_reports_work` classifies a raw harness record with, so the
- *   two readings of "billed" are one contract with one implementation.
+ *   `signals::record_work_evidence` reads a raw harness record with, and it answers
+ *   in this same type, so the two readings of "billed" are one contract with one
+ *   implementation.
  *
  * It is also a **published reading**, not only an internal one: a run that
  * failed with nothing to classify carries it as [`RunResult::work`] and in its
@@ -494,11 +495,11 @@ export interface RunResult {
      * `status`, which records oneharness's relationship to the process. Two
      * families: coarse reasons for a non-zero run (`auth`, `rate_limit`,
      * `model_not_found`, `quota`, `session_not_found`, `untrusted_directory`,
-     * `input_too_large`, `model_mismatch`), and `tool_deferred` — a run that exited
-     * *cleanly* but only deferred a builtin tool call instead of executing it
-     * (Claude Code bridge/managed deployments), so it did no useful work. The
-     * deferred case is the only `failure_kind` that can appear on a `status: ok`
-     * run, and it also marks the run as failed for exit-code purposes.
+     * `input_too_large`, `model_mismatch`, `server_overloaded`), and
+     * `tool_deferred` — a run that only deferred a builtin tool call instead of
+     * executing it (Claude Code bridge/managed deployments), so it did no useful
+     * work. Both `tool_deferred` and `server_overloaded` can accompany a clean
+     * transport exit; either still marks the run as failed for exit-code purposes.
      * Serialized as its snake_case token (see [`FailureKind`]).
      */
     failure_kind: FailureKind | null;