@agentmark-ai/shared-utils 0.3.3 → 0.4.0

package/dist/index.d.mts

@@ -558,4 +558,43 @@ declare function normalizeSpan(resource: OtelResource, scope: OtelScope, span: O
  */
 declare function normalizeOtlpSpans(resourceSpans: OtlpResourceSpans[]): NormalizedSpan[];
 
-export { AGENTMARK_SCOPE_NAME, AgentMarkTransformer, type AgentmarkConfig, type AgentmarkModelConfig, type AgentmarkModelSchema, type AgentmarkModelSettingsConfig, type AgentmarkModelSettingsSchema, AiSdkTransformer, type AiSdkVersion, type AttributeExtractor, AgentMarkTransformer as ClaudeAgentTransformer, type GenerateTypesLanguage, MastraTransformer, type McpServerConfig, type McpServers, type McpStdioServerConfig, type McpUrlServerConfig, type Message, type ModelSettingsTypeAspectRatio, type ModelSettingsTypeImageSize, type ModelSettingsTypeSelect, type ModelSettingsTypeSlider, type NormalizedSpan, type OtelEvent, OtelGenAiTransformer, type OtelLink, type OtelResource, type OtelScope, type OtelSpan, type OtlpAttribute, type OtlpAttributeValue, type OtlpEvent, type OtlpLink, type OtlpResource, type OtlpResourceSpans, type OtlpScope, type OtlpScopeSpans, type OtlpSpan, SEMANTIC_KINDS, type ScopeTransformer, type SemanticKind, SpanType, type StandardMessageContent, type StandardTextContent, type StandardToolCallContent, type StandardToolResultContent, type ToolCall, TransformerRegistry, TypeClassifier, convertOtlpAttributes, createSignature, detectVersion, extractCustomMetadata, extractReasoningFromProviderMetadata, extractResourceScopeSpan, fetchPromptsFrontmatter, findPromptFiles, generateTypeDefinitions, generateUnique8CharString, normalizeOtlpSpans, normalizeSpan, parseAgentMarkAttributes, parseMetadata, parseTokens, registry, resolveSemanticKind, toFrontMatter, typeClassifier, verifySignature };
+/**
+ * Compute a stable identifier for a dataset row, derived from its input.
+ *
+ * The identifier survives row reordering, additions, and deletions in the
+ * dataset — anything that doesn't change the row's *content* keeps the same
+ * id. This is the property regression-vs-baseline comparisons need: the
+ * same logical test case in two different runs must produce the same name
+ * so the baseline lookup can match them.
+ *
+ * Format: first 12 hex characters of the MD5 digest of the row's input,
+ * canonicalised with sorted keys at every level. This matches the Python
+ * runner's implementation byte-for-byte so cross-runtime comparisons
+ * (TS-emitted baseline vs. Python-emitted PR run, or vice versa) produce
+ * matching identifiers.
+ *
+ * Falls back to the positional index as a string when the row has no
+ * input — in that case there is nothing stable to hash, so positional is
+ * the best we can do.
+ *
+ * Note: we use MD5 here for compatibility with the existing Python
+ * implementation. MD5 is not used as a security primitive — it's used as a
+ * non-cryptographic content fingerprint, which is its appropriate role.
+ */
+declare function computeDatasetItemName(input: unknown, fallbackIndex: number): string;
+/**
+ * Stringify a value as canonical JSON with object keys sorted recursively.
+ *
+ * Differs from `JSON.stringify`:
+ * - Object keys are sorted lexicographically at every level of nesting, so
+ *   `{a: 1, b: 2}` and `{b: 2, a: 1}` produce the same output.
+ * - Non-serializable values (functions, symbols, bigints, undefined) are
+ *   coerced to strings via `String(value)`, matching Python's `default=str`
+ *   behavior in `json.dumps`.
+ *
+ * Exported for testing and for callers that need the same canonicalisation
+ * for purposes other than item-name hashing.
+ */
+declare function canonicalJsonStringify(value: unknown): string;
+
+export { AGENTMARK_SCOPE_NAME, AgentMarkTransformer, type AgentmarkConfig, type AgentmarkModelConfig, type AgentmarkModelSchema, type AgentmarkModelSettingsConfig, type AgentmarkModelSettingsSchema, AiSdkTransformer, type AiSdkVersion, type AttributeExtractor, AgentMarkTransformer as ClaudeAgentTransformer, type GenerateTypesLanguage, MastraTransformer, type McpServerConfig, type McpServers, type McpStdioServerConfig, type McpUrlServerConfig, type Message, type ModelSettingsTypeAspectRatio, type ModelSettingsTypeImageSize, type ModelSettingsTypeSelect, type ModelSettingsTypeSlider, type NormalizedSpan, type OtelEvent, OtelGenAiTransformer, type OtelLink, type OtelResource, type OtelScope, type OtelSpan, type OtlpAttribute, type OtlpAttributeValue, type OtlpEvent, type OtlpLink, type OtlpResource, type OtlpResourceSpans, type OtlpScope, type OtlpScopeSpans, type OtlpSpan, SEMANTIC_KINDS, type ScopeTransformer, type SemanticKind, SpanType, type StandardMessageContent, type StandardTextContent, type StandardToolCallContent, type StandardToolResultContent, type ToolCall, TransformerRegistry, TypeClassifier, canonicalJsonStringify, computeDatasetItemName, convertOtlpAttributes, createSignature, detectVersion, extractCustomMetadata, extractReasoningFromProviderMetadata, extractResourceScopeSpan, fetchPromptsFrontmatter, findPromptFiles, generateTypeDefinitions, generateUnique8CharString, normalizeOtlpSpans, normalizeSpan, parseAgentMarkAttributes, parseMetadata, parseTokens, registry, resolveSemanticKind, toFrontMatter, typeClassifier, verifySignature };

package/dist/index.d.ts

@@ -558,4 +558,43 @@ declare function normalizeSpan(resource: OtelResource, scope: OtelScope, span: O
  */
 declare function normalizeOtlpSpans(resourceSpans: OtlpResourceSpans[]): NormalizedSpan[];
 
-export { AGENTMARK_SCOPE_NAME, AgentMarkTransformer, type AgentmarkConfig, type AgentmarkModelConfig, type AgentmarkModelSchema, type AgentmarkModelSettingsConfig, type AgentmarkModelSettingsSchema, AiSdkTransformer, type AiSdkVersion, type AttributeExtractor, AgentMarkTransformer as ClaudeAgentTransformer, type GenerateTypesLanguage, MastraTransformer, type McpServerConfig, type McpServers, type McpStdioServerConfig, type McpUrlServerConfig, type Message, type ModelSettingsTypeAspectRatio, type ModelSettingsTypeImageSize, type ModelSettingsTypeSelect, type ModelSettingsTypeSlider, type NormalizedSpan, type OtelEvent, OtelGenAiTransformer, type OtelLink, type OtelResource, type OtelScope, type OtelSpan, type OtlpAttribute, type OtlpAttributeValue, type OtlpEvent, type OtlpLink, type OtlpResource, type OtlpResourceSpans, type OtlpScope, type OtlpScopeSpans, type OtlpSpan, SEMANTIC_KINDS, type ScopeTransformer, type SemanticKind, SpanType, type StandardMessageContent, type StandardTextContent, type StandardToolCallContent, type StandardToolResultContent, type ToolCall, TransformerRegistry, TypeClassifier, convertOtlpAttributes, createSignature, detectVersion, extractCustomMetadata, extractReasoningFromProviderMetadata, extractResourceScopeSpan, fetchPromptsFrontmatter, findPromptFiles, generateTypeDefinitions, generateUnique8CharString, normalizeOtlpSpans, normalizeSpan, parseAgentMarkAttributes, parseMetadata, parseTokens, registry, resolveSemanticKind, toFrontMatter, typeClassifier, verifySignature };
+/**
+ * Compute a stable identifier for a dataset row, derived from its input.
+ *
+ * The identifier survives row reordering, additions, and deletions in the
+ * dataset — anything that doesn't change the row's *content* keeps the same
+ * id. This is the property regression-vs-baseline comparisons need: the
+ * same logical test case in two different runs must produce the same name
+ * so the baseline lookup can match them.
+ *
+ * Format: first 12 hex characters of the MD5 digest of the row's input,
+ * canonicalised with sorted keys at every level. This matches the Python
+ * runner's implementation byte-for-byte so cross-runtime comparisons
+ * (TS-emitted baseline vs. Python-emitted PR run, or vice versa) produce
+ * matching identifiers.
+ *
+ * Falls back to the positional index as a string when the row has no
+ * input — in that case there is nothing stable to hash, so positional is
+ * the best we can do.
+ *
+ * Note: we use MD5 here for compatibility with the existing Python
+ * implementation. MD5 is not used as a security primitive — it's used as a
+ * non-cryptographic content fingerprint, which is its appropriate role.
+ */
+declare function computeDatasetItemName(input: unknown, fallbackIndex: number): string;
+/**
+ * Stringify a value as canonical JSON with object keys sorted recursively.
+ *
+ * Differs from `JSON.stringify`:
+ * - Object keys are sorted lexicographically at every level of nesting, so
+ *   `{a: 1, b: 2}` and `{b: 2, a: 1}` produce the same output.
+ * - Non-serializable values (functions, symbols, bigints, undefined) are
+ *   coerced to strings via `String(value)`, matching Python's `default=str`
+ *   behavior in `json.dumps`.
+ *
+ * Exported for testing and for callers that need the same canonicalisation
+ * for purposes other than item-name hashing.
+ */
+declare function canonicalJsonStringify(value: unknown): string;
+
+export { AGENTMARK_SCOPE_NAME, AgentMarkTransformer, type AgentmarkConfig, type AgentmarkModelConfig, type AgentmarkModelSchema, type AgentmarkModelSettingsConfig, type AgentmarkModelSettingsSchema, AiSdkTransformer, type AiSdkVersion, type AttributeExtractor, AgentMarkTransformer as ClaudeAgentTransformer, type GenerateTypesLanguage, MastraTransformer, type McpServerConfig, type McpServers, type McpStdioServerConfig, type McpUrlServerConfig, type Message, type ModelSettingsTypeAspectRatio, type ModelSettingsTypeImageSize, type ModelSettingsTypeSelect, type ModelSettingsTypeSlider, type NormalizedSpan, type OtelEvent, OtelGenAiTransformer, type OtelLink, type OtelResource, type OtelScope, type OtelSpan, type OtlpAttribute, type OtlpAttributeValue, type OtlpEvent, type OtlpLink, type OtlpResource, type OtlpResourceSpans, type OtlpScope, type OtlpScopeSpans, type OtlpSpan, SEMANTIC_KINDS, type ScopeTransformer, type SemanticKind, SpanType, type StandardMessageContent, type StandardTextContent, type StandardToolCallContent, type StandardToolResultContent, type ToolCall, TransformerRegistry, TypeClassifier, canonicalJsonStringify, computeDatasetItemName, convertOtlpAttributes, createSignature, detectVersion, extractCustomMetadata, extractReasoningFromProviderMetadata, extractResourceScopeSpan, fetchPromptsFrontmatter, findPromptFiles, generateTypeDefinitions, generateUnique8CharString, normalizeOtlpSpans, normalizeSpan, parseAgentMarkAttributes, parseMetadata, parseTokens, registry, resolveSemanticKind, toFrontMatter, typeClassifier, verifySignature };

package/dist/index.js

@@ -40,6 +40,8 @@ __export(index_exports, {
   SpanType: () => SpanType,
   TransformerRegistry: () => TransformerRegistry,
   TypeClassifier: () => TypeClassifier,
+  canonicalJsonStringify: () => canonicalJsonStringify,
+  computeDatasetItemName: () => computeDatasetItemName,
   convertOtlpAttributes: () => convertOtlpAttributes,
   createSignature: () => createSignature,
   detectVersion: () => detectVersion,
@@ -2320,6 +2322,36 @@ function normalizeOtlpSpans(resourceSpans) {
   }
   return normalizedSpans;
 }
+
+// src/dataset-item-name.ts
+var import_node_crypto = require("crypto");
+function computeDatasetItemName(input, fallbackIndex) {
+  if (input === void 0 || input === null) {
+    return String(fallbackIndex);
+  }
+  const canonical = canonicalJsonStringify(input);
+  return (0, import_node_crypto.createHash)("md5").update(canonical).digest("hex").slice(0, 12);
+}
+function canonicalJsonStringify(value) {
+  if (value === null) return "null";
+  if (value === void 0) return JSON.stringify(String(void 0));
+  const t = typeof value;
+  if (t === "string" || t === "number" || t === "boolean") {
+    return JSON.stringify(value);
+  }
+  if (t === "bigint" || t === "function" || t === "symbol") {
+    return JSON.stringify(String(value));
+  }
+  if (Array.isArray(value)) {
+    return "[" + value.map(canonicalJsonStringify).join(",") + "]";
+  }
+  const obj = value;
+  const keys = Object.keys(obj).sort();
+  const parts = keys.map((k) => {
+    return JSON.stringify(k) + ":" + canonicalJsonStringify(obj[k]);
+  });
+  return "{" + parts.join(",") + "}";
+}
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   AGENTMARK_SCOPE_NAME,
@@ -2332,6 +2364,8 @@ function normalizeOtlpSpans(resourceSpans) {
   SpanType,
   TransformerRegistry,
   TypeClassifier,
+  canonicalJsonStringify,
+  computeDatasetItemName,
   convertOtlpAttributes,
   createSignature,
   detectVersion,