@oneharness/sdk 0.26.0 → 0.27.0

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" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | 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" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | undefined;
     status?: "nonzero" | "timeout" | "cancelled" | undefined;
     work: "done" | "none";
     [k: string]: unknown;
@@ -29,43 +29,49 @@ export type HistoryRecord = ({
     [k: string]: unknown;
 } | {
     observed_model: string;
-    schema_version?: "1.8" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | 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" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | undefined;
     [k: string]: unknown;
 }) & (({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | "app_server_lost" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | "app_server_lost" | "unsupported_input" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | undefined;
     [k: string]: unknown;
 }) & ({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "model_mismatch" | "server_overloaded" | "app_server_lost" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "model_mismatch" | "server_overloaded" | "app_server_lost" | "unsupported_input" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | undefined;
     [k: string]: unknown;
 }) & ({
-    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "server_overloaded" | "app_server_lost" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "server_overloaded" | "app_server_lost" | "unsupported_input" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.8" | "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.8" | "1.9" | "1.10" | "1.11" | "1.12" | 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" | "app_server_lost" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "app_server_lost" | "unsupported_input" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.9" | "1.10" | "1.11" | undefined;
+    schema_version?: "1.9" | "1.10" | "1.11" | "1.12" | 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" | "server_overloaded" | null;
+    failure_kind?: "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | "unsupported_input" | null;
     [k: string]: unknown;
 } | {
-    schema_version?: "1.10" | "1.11" | undefined;
+    schema_version?: "1.10" | "1.11" | "1.12" | 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" | "server_overloaded" | "app_server_lost" | null;
+    [k: string]: unknown;
+} | {
+    schema_version?: "1.12" | undefined;
     [k: string]: unknown;
 })) & ({
     duration_ms: number;
@@ -184,7 +190,7 @@ 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" | "1.9" | "1.10" | "1.11";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -487,7 +493,7 @@ 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" | "1.9" | "1.10" | "1.11";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -614,7 +620,7 @@ 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" | "1.9" | "1.10" | "1.11";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -741,7 +747,7 @@ 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" | "1.9" | "1.10" | "1.11";
+    schema_version: "1.2" | "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -868,7 +874,7 @@ export type HistoryRecord = ({
      * run's single prompt).
      */
     prompt: string;
-    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11";
+    schema_version: "1.3" | "1.4" | "1.5" | "1.6" | "1.7" | "1.8" | "1.9" | "1.10" | "1.11" | "1.12";
     /**
      * The oneharness session id this run belongs to (the history file's stem).
      */
@@ -1341,7 +1347,7 @@ 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" | "model_mismatch" | "server_overloaded" | "app_server_lost";
+export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | "app_server_lost" | "unsupported_input";
 /**
  * The outcome of attempting to run one harness.
  */

package/dist/generated/options.d.ts

@@ -26,6 +26,10 @@ export type OutputFormat = "text" | "json" | "stream-json";
  * free of `clap`.
  */
 export type RunMode = "parallel" | "fallback";
+/**
+ * A harness-agnostic class of built-in tool a turn may load (`tool_set`).
+ */
+export type ToolClass = "shell" | "read";
 /**
  * Options accepted by `OneHarness.run()` in the published Node SDK.
  *
@@ -38,6 +42,11 @@ export interface RunOptions {
      * Run against every supported harness.
      */
     all?: boolean | undefined;
+    /**
+     * Compact a session's context once it reaches this many tokens (a
+     * positive integer), through each harness's own knob.
+     */
+    autoCompactTokens?: number | undefined;
     /**
      * Further prompts, making this a **batch**: one harness fanned over each
      * prompt, sharing the cacheable `system`/model prefix. Combined order is
@@ -81,6 +90,10 @@ export interface RunOptions {
      * unset — leave the option out to name no file.
      */
     historyPointerFile?: string | undefined;
+    /**
+     * Image files attached to the prompt as images the model sees.
+     */
+    images?: readonly string[] | undefined;
     /**
      * Maximum harnesses (or, in a batch, prompts) to run concurrently.
      */
@@ -106,6 +119,14 @@ export interface RunOptions {
      * `ONEHARNESS_HISTORY` environment override.
      */
     noHistory?: boolean | undefined;
+    /**
+     * Load no MCP server or MCP tool.
+     */
+    noMcp?: boolean | undefined;
+    /**
+     * Load no skills list.
+     */
+    noSkills?: boolean | undefined;
     /**
      * Write each harness's raw stdout/stderr under this directory.
      */
@@ -166,6 +187,15 @@ export interface RunOptions {
      */
     systemFile?: string | undefined;
     timeoutSeconds?: number | undefined;
+    /**
+     * Load only these built-in tool classes; every other built-in tool is
+     * not loaded. At least one: each class is sent as its own `--tool-set`,
+     * so an empty array would send nothing at all and is refused rather than
+     * silently loading the harness's own tools.
+     *
+     * @minItems 1
+     */
+    toolSet?: readonly ToolClass[] | undefined;
 }
 export interface HistoryLabels {
     [k: string]: string;

package/dist/generated/registry.d.ts

@@ -83,6 +83,21 @@ export interface HarnessInfo {
      * into that file (see the README support matrix).
      */
     supports_allowed_tools: boolean;
+    /**
+     * Whether `auto_compact_tokens` reaches this harness's own compaction
+     * knob. Every `supports_*` flag below is read from the translation table
+     * ([`crate::domain::lean::TRANSLATIONS`]): `false` means a run asking for
+     * the setting carries a diagnostic (or, for images, refuses the harness).
+     */
+    supports_auto_compact: boolean;
+    /**
+     * Whether a result's `calls` can carry per-call usage.
+     */
+    supports_call_usage: boolean;
+    /**
+     * Whether `oneharness compact` can compact this harness's sessions.
+     */
+    supports_compact_now: boolean;
     supports_denied_tools: boolean;
     /**
      * Whether `run --resume <session> --fork` is supported — branching a new
@@ -90,6 +105,14 @@ export interface HarnessInfo {
      */
     supports_fork: boolean;
     supports_hooks: boolean;
+    /**
+     * Whether `images` reach the model as images.
+     */
+    supports_images: boolean;
+    /**
+     * Whether `load_mcp = false` loads no MCP server.
+     */
+    supports_mcp_off: boolean;
     /**
      * Whether `oneharness mock <id>` can express a pre-tool *deny* for this
      * harness (the same protocol `oneharness gate` speaks).
@@ -118,6 +141,10 @@ export interface HarnessInfo {
      * Whether `run --resume <session>` is supported for this harness.
      */
     supports_resume: boolean;
+    /**
+     * Whether `load_skills = false` loads no skills list.
+     */
+    supports_skills_off: boolean;
     /**
      * Whether a large system prompt can be delivered off the argv via a file
      * flag (Claude Code's `--append-system-prompt-file`). `false` does not mean a
@@ -125,6 +152,11 @@ export interface HarnessInfo {
      * travels on stdin with the prompt; see the README large-prompt matrix.
      */
     supports_system_file: boolean;
+    /**
+     * Whether `tool_set` loads exactly the named classes (`false` for codex,
+     * which applies it in part and says so).
+     */
+    supports_tool_set: boolean;
     /**
      * The project-scoped config file `oneharness sync` writes for this
      * harness; `null` when it has none (sync settings are then rejected).

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

@@ -90,7 +90,11 @@ 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" | "server-overloaded" | "app-server-lost";
+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" | "app-server-lost" | "unsupported-input";
+/**
+ * One of the settings this module translates.
+ */
+export type Setting = "auto_compact_tokens" | "compact_now" | "tool_set" | "load_skills" | "load_mcp" | "images" | "calls";
 /**
  * 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
@@ -102,7 +106,7 @@ export type FallThroughReason = "not-installed" | "spawn-error" | "auth" | "quot
  * 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" | "server_overloaded" | "app_server_lost";
+export type FailureKind = "auth" | "rate_limit" | "model_not_found" | "quota" | "session_not_found" | "tool_deferred" | "untrusted_directory" | "input_too_large" | "model_mismatch" | "server_overloaded" | "app_server_lost" | "unsupported_input";
 /**
  * How a harness emits its result, which decides how `text` is extracted.
  *
@@ -504,10 +508,26 @@ export interface RunResult {
      * The binary name or path oneharness resolved and would invoke.
      */
     bin: string;
+    /**
+     * One usage per model call the turn made, in order, read from the
+     * harness's own stream or session record (see
+     * [`crate::domain::lean::Setting::Calls`]). Like `usage`, it covers the
+     * run's final attempt when a retry ran it again. Absent where the harness
+     * exposes no per-call figures, and then `diagnostics` says why. The
+     * run's `usage` total is unchanged by it. Added in report schema `0.13`.
+     */
+    calls?: Usage[] | null | undefined;
     /**
      * The exact argv oneharness built; its first element is the binary.
      */
     command: string[];
+    /**
+     * Each setting this run's harness could not apply, or applied in part
+     * (`auto_compact_tokens`, `tool_set`, `load_skills`, `load_mcp`, `calls`):
+     * the run proceeded, and this says what it did instead. Absent when
+     * empty. Added in report schema `0.13`.
+     */
+    diagnostics?: Diagnostic[] | undefined;
     /**
      * Wall-clock duration of the run; `null` when not executed.
      */
@@ -648,7 +668,7 @@ export interface RunResult {
      * How `text` was extracted (e.g. `json:result`, `raw`); `null` when absent.
      */
     text_source: string | null;
-    usage: Usage;
+    usage: Usage1;
     /**
      * How `usage` was read (e.g. `json`); `null` when nothing was found.
      */
@@ -676,8 +696,9 @@ export interface RunResult {
     [k: string]: unknown;
 }
 /**
- * Best-effort token/cost accounting; every field is `null` when the harness
- * does not report it. Always present so consumers can read a stable shape.
+ * Normalized token/cost accounting. Every field is best-effort and independently
+ * nullable: a harness may report tokens but not dollar cost (cost is commonly
+ * absent on subscription auth), or report nothing at all (plain-text harnesses).
  */
 export interface Usage {
     /**
@@ -706,6 +727,53 @@ export interface Usage {
     output_tokens: number | null;
     [k: string]: unknown;
 }
+/**
+ * A setting a harness could not apply, named in the run's report.
+ */
+export interface Diagnostic {
+    /**
+     * The harness id the run used.
+     */
+    harness: string;
+    /**
+     * What happened instead.
+     */
+    message: string;
+    setting: Setting;
+    [k: string]: unknown;
+}
+/**
+ * Normalized token/cost accounting. Every field is best-effort and independently
+ * nullable: a harness may report tokens but not dollar cost (cost is commonly
+ * absent on subscription auth), or report nothing at all (plain-text harnesses).
+ */
+export interface Usage1 {
+    /**
+     * Prompt tokens served from the provider's prompt cache (a cheap read of a
+     * previously-written prefix), when the harness reports them. `None` when the
+     * harness does not surface cache counts — never `0` as a guess.
+     */
+    cache_read_tokens: number | null;
+    /**
+     * Prompt tokens written to the provider's prompt cache (a.k.a. cache
+     * creation), when the harness reports them. `None` when not surfaced.
+     */
+    cache_write_tokens: number | null;
+    /**
+     * Total cost in USD, when the harness reports it (often absent on
+     * subscription auth, where there is no per-call dollar figure).
+     */
+    cost_usd: number | null;
+    /**
+     * Prompt/input tokens billed, when the harness reports them.
+     */
+    input_tokens: number | null;
+    /**
+     * Completion/output tokens billed, when the harness reports them.
+     */
+    output_tokens: number | null;
+    [k: string]: unknown;
+}
 /**
  * The uniform session handle for a run (`--session`). Present on
  * [`RunReport::session`] only when `--session <name>` was requested.