@bitfab/sdk 0.28.2 → 0.28.4

package/dist/index.d.cts

@@ -137,7 +137,7 @@ interface ActiveSpanContext$2 {
  * Bitfab spans with proper parent-child hierarchy.
  *
  * The TypeScript Claude Agent SDK exposes a single `query()` entry point (there
- * is no `ClaudeSDKClient` class — that exists only in the Python SDK). Wrap the
+ * is no `ClaudeSDKClient` class - that exists only in the Python SDK). Wrap the
  * `query()` async iterator with `wrapQuery`; tool and subagent spans come from
  * the hooks injected by `instrumentOptions`.
  *
@@ -221,7 +221,7 @@ declare class BitfabClaudeAgentHandler {
      * SDK's `wrapResponse` (which wraps `ClaudeSDKClient.receiveResponse()`);
      * in TypeScript, prefer `wrapQuery` around `query()`.
      *
-     * Pass `{ input }` (the prompt) to record a replayable root span — see
+     * Pass `{ input }` (the prompt) to record a replayable root span - see
      * `wrapQuery`.
      */
     wrapResponse(stream: AsyncIterable<unknown>, opts?: {
@@ -233,8 +233,8 @@ declare class BitfabClaudeAgentHandler {
      * Tool and subagent spans are captured separately via the hooks injected
      * by `instrumentOptions` into the `options` passed to `query()`.
      *
-     * Pass `{ input }` — the prompt (or the serializable args that produced it)
-     * — to make a handler-only run replayable: the handler records a root `agent`
+     * Pass `{ input }` - the prompt (or the serializable args that produced it)
+     * - to make a handler-only run replayable: the handler records a root `agent`
      * span with that input, so `replay(key, fn)` can re-feed it. Omit it only
      * when an enclosing `withSpan` already supplies the replayable root.
      *
@@ -445,7 +445,7 @@ declare class BitfabLangGraphCallbackHandler {
  *
  *   1. A process-wide `TracingProcessor` (see `getOpenAiTracingProcessor` /
  *      `BitfabOpenAITracingProcessor`) registered once with `addTraceProcessor`.
- *      It captures everything *inside* a run — LLM calls, tool calls, handoffs —
+ *      It captures everything *inside* a run - LLM calls, tool calls, handoffs -
  *      as Bitfab spans.
  *   2. This handler's `wrapRun`, which owns the *root*. The processor never sees
  *      the caller's input (the SDK's trace events don't carry it), so a
@@ -516,7 +516,7 @@ declare class BitfabOpenAIAgentHandler {
      * argument, so `replay(key, fn)` re-feeds it), and the run's `finalOutput`
      * is recorded as the root output. For streaming runs (`{ stream: true }`),
      * the result is handed back immediately and the final output is recorded
-     * once the stream completes — first-byte latency is untouched.
+     * once the stream completes - first-byte latency is untouched.
      *
      * The process-wide tracing processor (`getOpenAiTracingProcessor`) must still
      * be registered: it captures the LLM/tool/handoff spans that nest beneath
@@ -542,7 +542,7 @@ declare class BitfabOpenAIAgentHandler {
  * even when the SDK runs items in parallel.
  *
  * Internally, the resolved per-item state is a `DbBranchLease` (see
- * replayContext.ts) — that's the SDK ↔ server protocol term. We expose
+ * replayContext.ts) - that's the SDK ↔ server protocol term. We expose
  * its useful fields directly here so customer code never sees the word.
  */
 interface ReplayEnvironmentSnapshot {
@@ -576,7 +576,7 @@ declare class ReplayEnvironment {
     snapshot(): ReplayEnvironmentSnapshot | null;
     /**
      * Record on the replay context that customer code obtained the branch
-     * URL. Only `databaseUrl` and `snapshot()` count — `active`, `readOnly`
+     * URL. Only `databaseUrl` and `snapshot()` count - `active`, `readOnly`
      * and friends inspect the lease without exposing the connection string,
      * so they don't prove the replayed code could have connected to the
      * branch.
@@ -623,9 +623,9 @@ interface ReplayOptions {
     codeChangeFiles?: CodeChangeFile[];
     /**
      * Mock strategy for child spans during replay.
-     * - "none": everything runs real code (default)
+     * - "marked": only spans tagged with { mockOnReplay: true } in SpanOptions are mocked (default)
+     * - "none": everything runs real code
      * - "all": every child withSpan returns historical output
-     * - "marked": only spans tagged with { mockOnReplay: true } in SpanOptions are mocked
      */
     mock?: MockStrategy;
     /**
@@ -968,7 +968,7 @@ interface CurrentSpan {
  * A detached handle to a previously-created trace, looked up by its
  * caller-supplied id (the same id passed when the trace was started).
  *
- * Unlike `getCurrentTrace()`, this handle is not tied to AsyncLocalStorage —
+ * Unlike `getCurrentTrace()`, this handle is not tied to AsyncLocalStorage -
  * each method sends to the server immediately. Useful for adding context
  * to a trace from a different process, request, or thread (e.g. a forked
  * agent that wants to annotate the original conversation's trace).
@@ -1102,7 +1102,7 @@ interface SpanOptions {
      * ignored outside replay and under the "all"/"none" strategies.
      *
      * Use this for child spans that are expensive (paid LLM/API calls),
-     * slow, or non-deterministic — the root function still runs real code,
+     * slow, or non-deterministic - the root function still runs real code,
      * only the marked descendants return their recorded output.
      */
     mockOnReplay?: boolean;
@@ -1324,7 +1324,7 @@ declare class Bitfab {
      * Wrap a BAML client method to automatically capture prompt and LLM metadata.
      *
      * Creates a BAML Collector, calls the method through a tracked client,
-     * then extracts rendered messages and token usage — calling setPrompt()
+     * then extracts rendered messages and token usage - calling setPrompt()
      * and addContext() on the current span automatically.
      *
      * The BAML client can be provided in the constructor or passed explicitly:
@@ -1385,7 +1385,7 @@ declare class Bitfab {
      * Get a detached handle to a previously-created trace, looked up by the
      * caller-supplied id (the same id passed at trace creation).
      *
-     * The returned handle is not tied to AsyncLocalStorage — each method sends
+     * The returned handle is not tied to AsyncLocalStorage - each method sends
      * to the server immediately. Useful for adding context to a trace from a
      * different process or thread than the one that created it.
      *
@@ -1443,7 +1443,7 @@ declare class Bitfab {
      * plain callable: plain callables are wrapped internally so each replayed
      * invocation records a trace tied to the test run. The plain-callable form
      * is how handler-instrumented workflows (LangGraph/LangChain, Claude Agent
-     * SDK) replay — those record traces under a key with no `withSpan`-wrapped
+     * SDK) replay - those record traces under a key with no `withSpan`-wrapped
      * root in the app.
      *
      * @param traceFunctionKey - The trace function key to replay
@@ -1571,7 +1571,7 @@ declare class BitfabFunction {
      */
     getLangGraphCallbackHandler(): BitfabLangGraphCallbackHandler;
     /**
-     * Alias of {@link getLangGraphCallbackHandler} — LangChain and LangGraph
+     * Alias of {@link getLangGraphCallbackHandler} - LangChain and LangGraph
      * share one callback system, so the same bound handler serves both.
      *
      * @returns A LangChain callback handler for this client and key
@@ -1584,7 +1584,7 @@ declare class BitfabFunction {
      * Unlike the other methods on this handle, `wrapBAML` does NOT use the bound
      * key: it opens no span of its own. It enriches the *current* span (via
      * `getCurrentSpan().setPrompt()` / `addContext()`), so call it inside a
-     * function wrapped by this handle's `withSpan` — the bound key keys that
+     * function wrapped by this handle's `withSpan` - the bound key keys that
      * wrapper, and the BAML prompt/metadata attach to it.
      *
      * @param methodOrClient - Either a BAML method (uses constructor bamlClient) or the BAML client instance
@@ -1603,7 +1603,7 @@ declare class BitfabFunction {
 /**
  * SDK version from package.json (injected at build time)
  */
-declare const __version__ = "0.28.2";
+declare const __version__ = "0.28.4";
 
 /**
  * Constants for the Bitfab SDK.

package/dist/index.d.ts

@@ -137,7 +137,7 @@ interface ActiveSpanContext$2 {
  * Bitfab spans with proper parent-child hierarchy.
  *
  * The TypeScript Claude Agent SDK exposes a single `query()` entry point (there
- * is no `ClaudeSDKClient` class — that exists only in the Python SDK). Wrap the
+ * is no `ClaudeSDKClient` class - that exists only in the Python SDK). Wrap the
  * `query()` async iterator with `wrapQuery`; tool and subagent spans come from
  * the hooks injected by `instrumentOptions`.
  *
@@ -221,7 +221,7 @@ declare class BitfabClaudeAgentHandler {
      * SDK's `wrapResponse` (which wraps `ClaudeSDKClient.receiveResponse()`);
      * in TypeScript, prefer `wrapQuery` around `query()`.
      *
-     * Pass `{ input }` (the prompt) to record a replayable root span — see
+     * Pass `{ input }` (the prompt) to record a replayable root span - see
      * `wrapQuery`.
      */
     wrapResponse(stream: AsyncIterable<unknown>, opts?: {
@@ -233,8 +233,8 @@ declare class BitfabClaudeAgentHandler {
      * Tool and subagent spans are captured separately via the hooks injected
      * by `instrumentOptions` into the `options` passed to `query()`.
      *
-     * Pass `{ input }` — the prompt (or the serializable args that produced it)
-     * — to make a handler-only run replayable: the handler records a root `agent`
+     * Pass `{ input }` - the prompt (or the serializable args that produced it)
+     * - to make a handler-only run replayable: the handler records a root `agent`
      * span with that input, so `replay(key, fn)` can re-feed it. Omit it only
      * when an enclosing `withSpan` already supplies the replayable root.
      *
@@ -445,7 +445,7 @@ declare class BitfabLangGraphCallbackHandler {
  *
  *   1. A process-wide `TracingProcessor` (see `getOpenAiTracingProcessor` /
  *      `BitfabOpenAITracingProcessor`) registered once with `addTraceProcessor`.
- *      It captures everything *inside* a run — LLM calls, tool calls, handoffs —
+ *      It captures everything *inside* a run - LLM calls, tool calls, handoffs -
  *      as Bitfab spans.
  *   2. This handler's `wrapRun`, which owns the *root*. The processor never sees
  *      the caller's input (the SDK's trace events don't carry it), so a
@@ -516,7 +516,7 @@ declare class BitfabOpenAIAgentHandler {
      * argument, so `replay(key, fn)` re-feeds it), and the run's `finalOutput`
      * is recorded as the root output. For streaming runs (`{ stream: true }`),
      * the result is handed back immediately and the final output is recorded
-     * once the stream completes — first-byte latency is untouched.
+     * once the stream completes - first-byte latency is untouched.
      *
      * The process-wide tracing processor (`getOpenAiTracingProcessor`) must still
      * be registered: it captures the LLM/tool/handoff spans that nest beneath
@@ -542,7 +542,7 @@ declare class BitfabOpenAIAgentHandler {
  * even when the SDK runs items in parallel.
  *
  * Internally, the resolved per-item state is a `DbBranchLease` (see
- * replayContext.ts) — that's the SDK ↔ server protocol term. We expose
+ * replayContext.ts) - that's the SDK ↔ server protocol term. We expose
  * its useful fields directly here so customer code never sees the word.
  */
 interface ReplayEnvironmentSnapshot {
@@ -576,7 +576,7 @@ declare class ReplayEnvironment {
     snapshot(): ReplayEnvironmentSnapshot | null;
     /**
      * Record on the replay context that customer code obtained the branch
-     * URL. Only `databaseUrl` and `snapshot()` count — `active`, `readOnly`
+     * URL. Only `databaseUrl` and `snapshot()` count - `active`, `readOnly`
      * and friends inspect the lease without exposing the connection string,
      * so they don't prove the replayed code could have connected to the
      * branch.
@@ -623,9 +623,9 @@ interface ReplayOptions {
     codeChangeFiles?: CodeChangeFile[];
     /**
      * Mock strategy for child spans during replay.
-     * - "none": everything runs real code (default)
+     * - "marked": only spans tagged with { mockOnReplay: true } in SpanOptions are mocked (default)
+     * - "none": everything runs real code
      * - "all": every child withSpan returns historical output
-     * - "marked": only spans tagged with { mockOnReplay: true } in SpanOptions are mocked
      */
     mock?: MockStrategy;
     /**
@@ -968,7 +968,7 @@ interface CurrentSpan {
  * A detached handle to a previously-created trace, looked up by its
  * caller-supplied id (the same id passed when the trace was started).
  *
- * Unlike `getCurrentTrace()`, this handle is not tied to AsyncLocalStorage —
+ * Unlike `getCurrentTrace()`, this handle is not tied to AsyncLocalStorage -
  * each method sends to the server immediately. Useful for adding context
  * to a trace from a different process, request, or thread (e.g. a forked
  * agent that wants to annotate the original conversation's trace).
@@ -1102,7 +1102,7 @@ interface SpanOptions {
      * ignored outside replay and under the "all"/"none" strategies.
      *
      * Use this for child spans that are expensive (paid LLM/API calls),
-     * slow, or non-deterministic — the root function still runs real code,
+     * slow, or non-deterministic - the root function still runs real code,
      * only the marked descendants return their recorded output.
      */
     mockOnReplay?: boolean;
@@ -1324,7 +1324,7 @@ declare class Bitfab {
      * Wrap a BAML client method to automatically capture prompt and LLM metadata.
      *
      * Creates a BAML Collector, calls the method through a tracked client,
-     * then extracts rendered messages and token usage — calling setPrompt()
+     * then extracts rendered messages and token usage - calling setPrompt()
      * and addContext() on the current span automatically.
      *
      * The BAML client can be provided in the constructor or passed explicitly:
@@ -1385,7 +1385,7 @@ declare class Bitfab {
      * Get a detached handle to a previously-created trace, looked up by the
      * caller-supplied id (the same id passed at trace creation).
      *
-     * The returned handle is not tied to AsyncLocalStorage — each method sends
+     * The returned handle is not tied to AsyncLocalStorage - each method sends
      * to the server immediately. Useful for adding context to a trace from a
      * different process or thread than the one that created it.
      *
@@ -1443,7 +1443,7 @@ declare class Bitfab {
      * plain callable: plain callables are wrapped internally so each replayed
      * invocation records a trace tied to the test run. The plain-callable form
      * is how handler-instrumented workflows (LangGraph/LangChain, Claude Agent
-     * SDK) replay — those record traces under a key with no `withSpan`-wrapped
+     * SDK) replay - those record traces under a key with no `withSpan`-wrapped
      * root in the app.
      *
      * @param traceFunctionKey - The trace function key to replay
@@ -1571,7 +1571,7 @@ declare class BitfabFunction {
      */
     getLangGraphCallbackHandler(): BitfabLangGraphCallbackHandler;
     /**
-     * Alias of {@link getLangGraphCallbackHandler} — LangChain and LangGraph
+     * Alias of {@link getLangGraphCallbackHandler} - LangChain and LangGraph
      * share one callback system, so the same bound handler serves both.
      *
      * @returns A LangChain callback handler for this client and key
@@ -1584,7 +1584,7 @@ declare class BitfabFunction {
      * Unlike the other methods on this handle, `wrapBAML` does NOT use the bound
      * key: it opens no span of its own. It enriches the *current* span (via
      * `getCurrentSpan().setPrompt()` / `addContext()`), so call it inside a
-     * function wrapped by this handle's `withSpan` — the bound key keys that
+     * function wrapped by this handle's `withSpan` - the bound key keys that
      * wrapper, and the BAML prompt/metadata attach to it.
      *
      * @param methodOrClient - Either a BAML method (uses constructor bamlClient) or the BAML client instance
@@ -1603,7 +1603,7 @@ declare class BitfabFunction {
 /**
  * SDK version from package.json (injected at build time)
  */
-declare const __version__ = "0.28.2";
+declare const __version__ = "0.28.4";
 
 /**
  * Constants for the Bitfab SDK.

package/dist/index.js

@@ -14,12 +14,12 @@ import {
   flushTraces,
   getCurrentSpan,
   getCurrentTrace
-} from "./chunk-NZFPJRD3.js";
+} from "./chunk-3F46BMPM.js";
 import {
   BITFAB_PROGRESS_PREFIX,
   BitfabError,
   reportReplayProgress
-} from "./chunk-3PCOUZZP.js";
+} from "./chunk-SAGZ674W.js";
 export {
   BITFAB_PROGRESS_PREFIX,
   Bitfab,

package/dist/node.cjs

@@ -40,7 +40,7 @@ function registerAsyncLocalStorageClass(cls) {
 function assertAsyncStorageRegistered() {
   if (!AsyncLocalStorageClass) {
     console.warn(
-      "Bitfab: AsyncLocalStorage not available \u2014 nested span context will not propagate."
+      "Bitfab: AsyncLocalStorage not available - nested span context will not propagate."
     );
   }
 }
@@ -398,10 +398,25 @@ async function processItem(httpClient, serverItem, fn, testRunId, mockStrategy,
     }
     let mockTree;
     if (mockStrategy === "all" || mockStrategy === "marked") {
-      const treeResponse = await httpClient.getSpanTree(
-        serverItem.externalSpanId
-      );
-      mockTree = buildMockTree(treeResponse.root);
+      try {
+        const treeResponse = await httpClient.getSpanTree(
+          serverItem.externalSpanId
+        );
+        if (treeResponse.root) {
+          mockTree = buildMockTree(treeResponse.root);
+        } else if (mockStrategy === "all") {
+          throw new BitfabError(
+            `Replay mock strategy "all" requires a span tree root for source span ${serverItem.externalSpanId}.`
+          );
+        } else {
+          mockTree = void 0;
+        }
+      } catch (e) {
+        if (mockStrategy === "all") {
+          throw e;
+        }
+        mockTree = void 0;
+      }
     }
     const maybePromise = runWithReplayContext(
       {
@@ -507,7 +522,7 @@ async function replay(httpClient, serviceUrl, traceFunctionKey, fn, options) {
     options?.experimentGroupId,
     options?.datasetId
   );
-  const mockStrategy = options?.mock ?? "none";
+  const mockStrategy = options?.mock ?? "marked";
   const maxConcurrency = options?.maxConcurrency ?? 10;
   const tasks = serverItems.map(
     (serverItem) => () => processItem(
@@ -658,7 +673,7 @@ registerAsyncLocalStorageClass(
 );
 
 // src/version.generated.ts
-var __version__ = "0.28.2";
+var __version__ = "0.28.4";
 
 // src/constants.ts
 var DEFAULT_SERVICE_URL = "https://bitfab.ai";
@@ -971,7 +986,7 @@ var HttpClient = class {
   }
   /**
    * Start a replay session by fetching historical traces.
-   * Blocking call — creates a test run and returns lightweight item references.
+   * Blocking call - creates a test run and returns lightweight item references.
    */
   async startReplay(traceFunctionKey, limit, traceIds, name, codeChangeDescription, codeChangeFiles, includeDbBranchLease, experimentGroupId, datasetId) {
     const payload = { traceFunctionKey };
@@ -1089,7 +1104,7 @@ var HttpClient = class {
   }
   /**
    * Ask the server to materialize a per-trace DB branch lease from a
-   * captured `dbSnapshotRef`. Blocking — the resolver creates a Neon
+   * captured `dbSnapshotRef`. Blocking - the resolver creates a Neon
    * snapshot + preview branch and polls operations to readiness, which
    * can take seconds.
    */
@@ -1229,7 +1244,7 @@ var BitfabClaudeAgentHandler = class {
     // Synthetic root span (handler-only replay). When an `input` is supplied to
     // wrapQuery/wrapResponse, the handler emits a root `agent` span carrying that
     // input, so a handler-instrumented run is replayable WITHOUT an enclosing
-    // withSpan — matching the LangGraph handler, which records the graph input as
+    // withSpan - matching the LangGraph handler, which records the graph input as
     // its root. The prompt is not present anywhere in the message stream, so it
     // must be handed in explicitly.
     this.hasRootInput = false;
@@ -1503,7 +1518,7 @@ var BitfabClaudeAgentHandler = class {
    * SDK's `wrapResponse` (which wraps `ClaudeSDKClient.receiveResponse()`);
    * in TypeScript, prefer `wrapQuery` around `query()`.
    *
-   * Pass `{ input }` (the prompt) to record a replayable root span — see
+   * Pass `{ input }` (the prompt) to record a replayable root span - see
    * `wrapQuery`.
    */
   async *wrapResponse(stream, opts) {
@@ -1516,8 +1531,8 @@ var BitfabClaudeAgentHandler = class {
    * Tool and subagent spans are captured separately via the hooks injected
    * by `instrumentOptions` into the `options` passed to `query()`.
    *
-   * Pass `{ input }` — the prompt (or the serializable args that produced it)
-   * — to make a handler-only run replayable: the handler records a root `agent`
+   * Pass `{ input }` - the prompt (or the serializable args that produced it)
+   * - to make a handler-only run replayable: the handler records a root `agent`
    * span with that input, so `replay(key, fn)` can re-feed it. Omit it only
    * when an enclosing `withSpan` already supplies the replayable root.
    *
@@ -2629,7 +2644,7 @@ var BitfabOpenAIAgentHandler = class {
    * argument, so `replay(key, fn)` re-feeds it), and the run's `finalOutput`
    * is recorded as the root output. For streaming runs (`{ stream: true }`),
    * the result is handed back immediately and the final output is recorded
-   * once the stream completes — first-byte latency is untouched.
+   * once the stream completes - first-byte latency is untouched.
    *
    * The process-wide tracing processor (`getOpenAiTracingProcessor`) must still
    * be registered: it captures the LLM/tool/handoff spans that nest beneath
@@ -2722,7 +2737,7 @@ var ReplayEnvironment = class {
   }
   /**
    * Record on the replay context that customer code obtained the branch
-   * URL. Only `databaseUrl` and `snapshot()` count — `active`, `readOnly`
+   * URL. Only `databaseUrl` and `snapshot()` count - `active`, `readOnly`
    * and friends inspect the lease without exposing the connection string,
    * so they don't prove the replayed code could have connected to the
    * branch.
@@ -3447,7 +3462,7 @@ var Bitfab = class {
     if (this.explicitlyEnabled && !this.apiKeyWarned) {
       this.apiKeyWarned = true;
       console.warn(
-        "Bitfab: apiKey is empty \u2014 tracing is disabled. Provide a valid API key to enable tracing."
+        "Bitfab: apiKey is empty - tracing is disabled. Provide a valid API key to enable tracing."
       );
     }
     return void 0;
@@ -3711,7 +3726,7 @@ var Bitfab = class {
    * Wrap a BAML client method to automatically capture prompt and LLM metadata.
    *
    * Creates a BAML Collector, calls the method through a tracked client,
-   * then extracts rendered messages and token usage — calling setPrompt()
+   * then extracts rendered messages and token usage - calling setPrompt()
    * and addContext() on the current span automatically.
    *
    * The BAML client can be provided in the constructor or passed explicitly:
@@ -4098,7 +4113,7 @@ var Bitfab = class {
    * Get a detached handle to a previously-created trace, looked up by the
    * caller-supplied id (the same id passed at trace creation).
    *
-   * The returned handle is not tied to AsyncLocalStorage — each method sends
+   * The returned handle is not tied to AsyncLocalStorage - each method sends
    * to the server immediately. Useful for adding context to a trace from a
    * different process or thread than the one that created it.
    *
@@ -4279,7 +4294,7 @@ var Bitfab = class {
    * plain callable: plain callables are wrapped internally so each replayed
    * invocation records a trace tied to the test run. The plain-callable form
    * is how handler-instrumented workflows (LangGraph/LangChain, Claude Agent
-   * SDK) replay — those record traces under a key with no `withSpan`-wrapped
+   * SDK) replay - those record traces under a key with no `withSpan`-wrapped
    * root in the app.
    *
    * @param traceFunctionKey - The trace function key to replay
@@ -4431,7 +4446,7 @@ var BitfabFunction = class {
     return this.client.getLangGraphCallbackHandler(this.traceFunctionKey);
   }
   /**
-   * Alias of {@link getLangGraphCallbackHandler} — LangChain and LangGraph
+   * Alias of {@link getLangGraphCallbackHandler} - LangChain and LangGraph
    * share one callback system, so the same bound handler serves both.
    *
    * @returns A LangChain callback handler for this client and key
@@ -4446,7 +4461,7 @@ var BitfabFunction = class {
    * Unlike the other methods on this handle, `wrapBAML` does NOT use the bound
    * key: it opens no span of its own. It enriches the *current* span (via
    * `getCurrentSpan().setPrompt()` / `addContext()`), so call it inside a
-   * function wrapped by this handle's `withSpan` — the bound key keys that
+   * function wrapped by this handle's `withSpan` - the bound key keys that
    * wrapper, and the BAML prompt/metadata attach to it.
    *
    * @param methodOrClient - Either a BAML method (uses constructor bamlClient) or the BAML client instance