@oneharness/sdk 0.11.4 → 0.12.1

package/dist/generated/history-records.d.ts

@@ -69,7 +69,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms: number;
@@ -78,6 +81,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -93,7 +112,7 @@ export type HistoryRecord = {
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -211,7 +230,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels1 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms: number;
@@ -220,6 +242,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -331,7 +369,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels2 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -340,6 +381,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -355,7 +412,7 @@ export type HistoryRecord = {
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -439,7 +496,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels3 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -448,6 +508,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms: number;
     /**
      * The normalized approval mode requested for the run.
@@ -463,7 +539,7 @@ export type HistoryRecord = {
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -547,7 +623,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels4 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -556,6 +635,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -571,7 +666,7 @@ export type HistoryRecord = {
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -655,7 +750,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels5 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -664,6 +762,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -679,7 +793,7 @@ export type HistoryRecord = {
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7";
+    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -766,7 +880,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels6 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -775,6 +892,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -889,7 +1022,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels7 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: never | undefined;
@@ -898,6 +1034,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     observed_tool_ms?: never | undefined;
     /**
      * The normalized approval mode requested for the run.
@@ -1000,7 +1152,10 @@ export type HistoryRecord = {
     history_id: string;
     labels?: HistoryLabels8 | undefined;
     /**
-     * The effective top-level model for the run, if any.
+     * The model this run **requested** — the value oneharness put on the
+     * harness's model flag or protocol frame — if any. Copied from
+     * [`RunResult::model`], so it keeps meaning the requested model even where
+     * `observed_model` below is present.
      */
     model: string | null;
     model_ms?: number | null | undefined;
@@ -1009,6 +1164,22 @@ export type HistoryRecord = {
      * every record so a reader can resolve a session by name from any line.
      */
     name: string;
+    /**
+     * The model the harness **itself** reported the conversation would run
+     * under, read off its own protocol before any token was spent — for codex
+     * over `app-server`, `result.model` of the `thread/start` /
+     * `thread/resume` response. Copied from [`RunResult::observed_model`]:
+     * never inferred and never copied from `model`, and omitted on the wire on
+     * every path that reports none.
+     *
+     * On a completed app-server turn the two agree, because a turn whose
+     * reported model differed from the requested one was refused before it
+     * ran (`model_mismatch`) — so a completed record's `model` is what codex
+     * ran, and the record can no longer say one model while another was
+     * billed. They differ only on such a refused record, and where no model
+     * was requested at all. Gated to [`FIRST_MODEL_OBSERVATION_SCHEMA_VERSION`].
+     */
+    observed_model?: string | null | undefined;
     /**
      * Union of tool intervals observed at the stdout pipe. Unlike `tool_ms`,
      * this is not provider-measured and has no model-latency counterpart.
@@ -1095,7 +1266,7 @@ export type TimingSource = "provider_measured" | "stdout_observed";
  * 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";
+export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch";
 /**
  * The outcome of attempting to run one harness.
  */
@@ -1125,7 +1296,7 @@ export type HistoryRecords = (({
     [k: string]: unknown;
 } | ({
     error: string;
-    schema_version?: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | undefined;
+    schema_version?: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | undefined;
     [k: string]: unknown;
 } & ({
     status?: "nonzero" | "timeout" | "cancelled" | "spawn-error" | "skipped" | undefined;
@@ -1137,27 +1308,40 @@ export type HistoryRecords = (({
     work?: null | undefined;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.7" | undefined;
+    schema_version?: "1.7" | "1.8" | undefined;
     status?: "nonzero" | "timeout" | "cancelled" | undefined;
     work: "done" | "none";
     [k: string]: unknown;
+}) & ({
+    observed_model?: null | undefined;
+    [k: string]: unknown;
+} | {
+    observed_model: string;
+    schema_version?: "1.8" | 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" | undefined;
+    schema_version?: "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | undefined;
     [k: string]: unknown;
 }) & (({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | null;
+    [k: string]: unknown;
+} | {
+    schema_version?: "1.5" | "1.6" | "1.7" | "1.8" | undefined;
+    [k: string]: unknown;
+}) & ({
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "model_mismatch" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.5" | "1.6" | "1.7" | undefined;
+    schema_version?: "1.6" | "1.7" | "1.8" | undefined;
     [k: string]: unknown;
 }) & ({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.6" | "1.7" | undefined;
+    schema_version?: "1.8" | undefined;
     [k: string]: unknown;
 })) & HistoryRecord)[];
 /**