opik 1.10.13 → 1.10.14

package/dist/index.d.ts

@@ -3001,6 +3001,8 @@ interface DatasetItem {
     source: DatasetItemSource;
     data: JsonNode;
     tags?: string[];
+    evaluators?: EvaluatorItem[];
+    executionPolicy?: ExecutionPolicy;
     experimentItems?: ExperimentItem[];
     datasetId?: string;
     createdAt?: Date;
@@ -3019,6 +3021,8 @@ interface DatasetItemCompare {
     source: DatasetItemCompareSource;
     data: JsonNode;
     tags?: string[];
+    evaluators?: EvaluatorItemCompare[];
+    executionPolicy?: ExecutionPolicyCompare;
     experimentItems?: ExperimentItemCompare[];
     datasetId?: string;
     createdAt?: Date;
@@ -3087,6 +3091,8 @@ interface DatasetItemPublic {
     source: DatasetItemPublicSource;
     data: JsonNode;
     tags?: string[];
+    evaluators?: EvaluatorItemPublic[];
+    executionPolicy?: ExecutionPolicyPublic;
     experimentItems?: ExperimentItemPublic[];
     datasetId?: string;
     createdAt?: Date;
@@ -3123,6 +3129,9 @@ interface DatasetItemUpdate {
     data?: JsonNode;
     /** Tags */
     tags?: string[];
+    /** Evaluators */
+    evaluators?: EvaluatorItem[];
+    executionPolicy?: ExecutionPolicy;
 }
 
 interface DatasetItemWrite {
@@ -3132,6 +3141,8 @@ interface DatasetItemWrite {
     source: DatasetItemWriteSource;
     data: JsonNode;
     tags?: string[];
+    evaluators?: EvaluatorItemWrite[];
+    executionPolicy?: ExecutionPolicyWrite;
 }
 
 declare const DatasetItemWriteSource: {
@@ -3299,6 +3310,74 @@ interface ErrorInfoWrite {
     traceback: string;
 }
 
+interface EvaluatorItem {
+    name: string;
+    type: EvaluatorItemType;
+    config: JsonNode;
+}
+
+interface EvaluatorItemCompare {
+    name: string;
+    type: EvaluatorItemCompareType;
+    config: JsonNodeCompare;
+}
+
+declare const EvaluatorItemCompareType: {
+    readonly LlmJudge: "llm_judge";
+    readonly CodeMetric: "code_metric";
+};
+type EvaluatorItemCompareType = (typeof EvaluatorItemCompareType)[keyof typeof EvaluatorItemCompareType];
+
+interface EvaluatorItemPublic {
+    name: string;
+    type: EvaluatorItemPublicType;
+    config: JsonNodePublic;
+}
+
+declare const EvaluatorItemPublicType: {
+    readonly LlmJudge: "llm_judge";
+    readonly CodeMetric: "code_metric";
+};
+type EvaluatorItemPublicType = (typeof EvaluatorItemPublicType)[keyof typeof EvaluatorItemPublicType];
+
+declare const EvaluatorItemType: {
+    readonly LlmJudge: "llm_judge";
+    readonly CodeMetric: "code_metric";
+};
+type EvaluatorItemType = (typeof EvaluatorItemType)[keyof typeof EvaluatorItemType];
+
+interface EvaluatorItemWrite {
+    name: string;
+    type: EvaluatorItemWriteType;
+    config: JsonNodeWrite;
+}
+
+declare const EvaluatorItemWriteType: {
+    readonly LlmJudge: "llm_judge";
+    readonly CodeMetric: "code_metric";
+};
+type EvaluatorItemWriteType = (typeof EvaluatorItemWriteType)[keyof typeof EvaluatorItemWriteType];
+
+interface ExecutionPolicy {
+    runsPerItem?: number;
+    passThreshold?: number;
+}
+
+interface ExecutionPolicyCompare {
+    runsPerItem?: number;
+    passThreshold?: number;
+}
+
+interface ExecutionPolicyPublic {
+    runsPerItem?: number;
+    passThreshold?: number;
+}
+
+interface ExecutionPolicyWrite {
+    runsPerItem?: number;
+    passThreshold?: number;
+}
+
 interface ExperimentGroupAggregationsResponse {
     content?: Record<string, GroupContentWithAggregations>;
 }
@@ -3313,6 +3392,8 @@ interface ExperimentItem {
     experimentId: string;
     datasetItemId: string;
     traceId: string;
+    projectId?: string;
+    projectName?: string;
     input?: JsonListString;
     output?: JsonListString;
     feedbackScores?: FeedbackScore[];
@@ -3340,6 +3421,7 @@ interface ExperimentItemCompare {
     experimentId: string;
     datasetItemId: string;
     traceId: string;
+    projectId?: string;
     input?: JsonListStringCompare;
     output?: JsonListStringCompare;
     feedbackScores?: FeedbackScoreCompare[];
@@ -3365,6 +3447,7 @@ interface ExperimentItemPublic {
     experimentId: string;
     datasetItemId: string;
     traceId: string;
+    projectId?: string;
     createdAt?: Date;
     lastUpdatedAt?: Date;
     createdBy?: string;
@@ -3841,6 +3924,8 @@ type JsonListStringWrite = Record<string, unknown> | Record<string, unknown>[] |
 
 type JsonNode = Record<string, unknown>;
 
+type JsonNodeCompare = Record<string, unknown>;
+
 type JsonNodeDetail = Record<string, unknown>;
 
 type JsonNodePublic = Record<string, unknown>;
@@ -4634,6 +4719,8 @@ interface Span$1 {
     totalEstimatedCostVersion?: string;
     /** Duration in milliseconds as a decimal number to support sub-millisecond precision */
     duration?: number;
+    /** Time to first token in milliseconds */
+    ttft?: number;
 }
 
 /**
@@ -4665,6 +4752,8 @@ interface SpanExperimentItemBulkWriteView {
     lastUpdatedAt?: Date;
     totalEstimatedCost?: number;
     totalEstimatedCostVersion?: string;
+    /** Time to first token in milliseconds */
+    ttft?: number;
 }
 
 declare const SpanExperimentItemBulkWriteViewType: {
@@ -4802,6 +4891,8 @@ interface SpanPublic {
     totalEstimatedCostVersion?: string;
     /** Duration in milliseconds as a decimal number to support sub-millisecond precision */
     duration?: number;
+    /** Time to first token in milliseconds */
+    ttft?: number;
 }
 
 declare const SpanPublicType: {
@@ -4843,6 +4934,7 @@ interface SpanUpdate$1 {
     usage?: Record<string, number>;
     totalEstimatedCost?: number;
     errorInfo?: ErrorInfo;
+    ttft?: number;
 }
 
 declare const SpanUpdateType: {
@@ -4889,6 +4981,8 @@ interface SpanWrite {
     lastUpdatedAt?: Date;
     totalEstimatedCost?: number;
     totalEstimatedCostVersion?: string;
+    /** Time to first token in milliseconds */
+    ttft?: number;
 }
 
 declare const SpanWriteType: {
@@ -5017,6 +5111,8 @@ interface Trace$1 {
     spanCount?: number;
     /** Duration in milliseconds as a decimal number to support sub-millisecond precision */
     duration?: number;
+    /** Time to first token in milliseconds */
+    ttft?: number;
     threadId?: string;
     visibilityMode?: TraceVisibilityMode;
     llmSpanCount?: number;
@@ -5058,6 +5154,8 @@ interface TraceExperimentItemBulkWriteView {
     tags?: string[];
     errorInfo?: ErrorInfoExperimentItemBulkWriteView;
     lastUpdatedAt?: Date;
+    /** Time to first token in milliseconds */
+    ttft?: number;
     threadId?: string;
 }
 
@@ -5163,6 +5261,8 @@ interface TracePublic {
     spanCount?: number;
     /** Duration in milliseconds as a decimal number to support sub-millisecond precision */
     duration?: number;
+    /** Time to first token in milliseconds */
+    ttft?: number;
     threadId?: string;
     visibilityMode?: TracePublicVisibilityMode;
     llmSpanCount?: number;
@@ -5337,6 +5437,7 @@ interface TraceUpdate {
     tags?: string[];
     errorInfo?: ErrorInfo;
     threadId?: string;
+    ttft?: number;
 }
 
 declare const TraceVisibilityMode: {
@@ -5358,6 +5459,8 @@ interface TraceWrite {
     tags?: string[];
     errorInfo?: ErrorInfoWrite;
     lastUpdatedAt?: Date;
+    /** Time to first token in milliseconds */
+    ttft?: number;
     threadId?: string;
 }
 
@@ -9782,9 +9885,11 @@ declare class Dataset<T extends DatasetItemData = DatasetItemData> {
 declare class ExperimentItemReferences {
     readonly datasetItemId: string;
     readonly traceId: string;
+    readonly projectName?: string;
     constructor(params: {
         datasetItemId: string;
         traceId: string;
+        projectName?: string;
     });
 }
 /**
@@ -10331,15 +10436,18 @@ declare class OpikClient {
      * Supported OQL format: `<COLUMN> <OPERATOR> <VALUE> [AND <COLUMN> <OPERATOR> <VALUE>]*`
      *
      * Supported columns:
-     * - `id`, `name`: String fields
+     * - `id`, `name`, `description`: String fields
+     * - `created_by`, `last_updated_by`: String fields
+     * - `template_structure`: String field (e.g., "text" or "chat")
+     * - `created_at`, `last_updated_at`: Date/time fields (ISO 8601 format)
      * - `tags`: List field (use "contains" operator only)
-     * - `created_by`: String field
+     * - `version_count`: Number field
      *
      * Supported operators by column:
-     * - `id`: =, !=, contains, not_contains, starts_with, ends_with, >, <
-     * - `name`: =, !=, contains, not_contains, starts_with, ends_with, >, <
-     * - `created_by`: =, !=, contains, not_contains, starts_with, ends_with, >, <
-     * - `tags`: contains (only)
+     * - String fields (`id`, `name`, `description`, `created_by`, `last_updated_by`, `template_structure`): =, !=, contains, not_contains, starts_with, ends_with, >, <
+     * - Date/time fields (`created_at`, `last_updated_at`): =, >, <, >=, <=
+     * - Number fields (`version_count`): =, !=, >, <, >=, <=
+     * - List fields (`tags`): contains
      *
      * @returns Promise resolving to array of matching latest prompt versions
      * @throws Error if OQL filter syntax is invalid
@@ -10359,6 +10467,15 @@ declare class OpikClient {
      *
      * // Filter by creator
      * const prompts = await client.searchPrompts('created_by = "user@example.com"');
+     *
+     * // Filter by template structure
+     * const chatPrompts = await client.searchPrompts('template_structure = "chat"');
+     *
+     * // Filter by date range
+     * const recentPrompts = await client.searchPrompts('created_at >= "2024-01-01T00:00:00Z"');
+     *
+     * // Filter by version count
+     * const multiVersion = await client.searchPrompts('version_count > 5');
      * ```
      */
     searchPrompts: (filterString?: string) => Promise<(Prompt | ChatPrompt)[]>;
@@ -10465,6 +10582,63 @@ declare class OpikClient {
         waitForAtLeast?: number;
         waitForTimeout?: number;
     }) => Promise<TraceThread[]>;
+    /**
+     * Search for spans in a project with optional filtering.
+     *
+     * Spans represent individual operations or steps within traces, such as LLM calls or function executions.
+     * This method allows you to search and filter spans using Opik Query Language (OQL).
+     *
+     * @param options - Search options
+     * @param options.projectName - Name of the project to search in. Defaults to the client's configured project.
+     * @param options.filterString - Filter string using Opik Query Language (OQL).
+     *   Supports filtering by: model, provider, type, metadata, feedback_scores, usage, duration, etc.
+     *   Examples: 'model = "gpt-4"', 'provider = "openai"', 'type = "llm"', 'metadata.version = "1.0"'
+     * @param options.maxResults - Maximum number of spans to return (default: 1000)
+     * @param options.truncate - Whether to truncate large fields in the response (default: true)
+     * @param options.waitForAtLeast - If specified, polls until at least this many spans are found
+     * @param options.waitForTimeout - Timeout in seconds when using waitForAtLeast (default: 60)
+     * @returns Promise resolving to an array of spans
+     * @throws {SearchTimeoutError} If waitForAtLeast is specified and timeout is reached
+     *
+     * @example
+     * ```typescript
+     * // Get all spans in a project
+     * const spans = await client.searchSpans({ projectName: "My Project" });
+     *
+     * // Filter by model
+     * const gpt4Spans = await client.searchSpans({
+     *   projectName: "My Project",
+     *   filterString: 'model = "gpt-4"'
+     * });
+     *
+     * // Filter by provider and type
+     * const openaiLLMSpans = await client.searchSpans({
+     *   projectName: "My Project",
+     *   filterString: 'provider = "openai" and type = "llm"'
+     * });
+     *
+     * // Filter by metadata
+     * const prodSpans = await client.searchSpans({
+     *   projectName: "My Project",
+     *   filterString: 'metadata.environment = "production"'
+     * });
+     *
+     * // Wait for at least 5 spans
+     * const spans = await client.searchSpans({
+     *   projectName: "My Project",
+     *   waitForAtLeast: 5,
+     *   waitForTimeout: 30
+     * });
+     * ```
+     */
+    searchSpans: (options?: {
+        projectName?: string;
+        filterString?: string;
+        maxResults?: number;
+        truncate?: boolean;
+        waitForAtLeast?: number;
+        waitForTimeout?: number;
+    }) => Promise<SpanPublic[]>;
     private logFeedbackScores;
     /**
      * Log feedback scores to existing traces in batch.
@@ -10748,7 +10922,7 @@ declare abstract class OpikBaseModel {
     /**
      * Creates a new model instance.
      *
-     * @param modelName - The name of the model (e.g., 'gpt-4o', 'claude-3-opus')
+     * @param modelName - The name of the model (e.g., 'gpt-5-nano', 'claude-3-opus')
      */
     constructor(modelName: string);
     /**
@@ -10763,7 +10937,7 @@ declare abstract class OpikBaseModel {
      *
      * @example
      * ```typescript
-     * const model = new VercelAIChatModel('gpt-4o');
+     * const model = new VercelAIChatModel('gpt-5-nano');
      *
      * // Simple text generation
      * const response = await model.generateString('What is 2+2?');
@@ -10791,7 +10965,7 @@ declare abstract class OpikBaseModel {
      *
      * @example
      * ```typescript
-     * const model = new VercelAIChatModel('gpt-4o');
+     * const model = new VercelAIChatModel('gpt-5-nano');
      * const response = await model.generateProviderResponse([
      *   { role: 'user', content: 'Hello!' }
      * ]);
@@ -10807,7 +10981,7 @@ declare abstract class OpikBaseModel {
  * @example
  * ```typescript
  * // Valid model IDs
- * const model1: SupportedModelId = "gpt-4o";
+ * const model1: SupportedModelId = "gpt-5-nano";
  * const model2: SupportedModelId = "claude-3-5-sonnet-latest";
  * const model3: SupportedModelId = "gemini-2.0-flash";
  * ```
@@ -10848,7 +11022,7 @@ type AllProviderOptions = OpenAIProviderOptions | AnthropicProviderOptions | Goo
  * @example
  * ```typescript
  * // TypeScript infers OpenAIProviderOptions
- * const options1: ProviderOptionsForModel<"gpt-4o"> = {
+ * const options1: ProviderOptionsForModel<"gpt-5-nano"> = {
  *   apiKey: "sk-...",
  *   organization: "org-123" // ✅ Valid OpenAI option
  * };
@@ -10864,7 +11038,7 @@ type ProviderOptionsForModel<T extends SupportedModelId> = T extends OpenAIChatM
  * Detects the provider from the model ID and creates the appropriate provider instance.
  * Uses pattern matching to automatically determine which provider to use.
  *
- * @param modelId - Model ID (e.g., "gpt-4o", "claude-3-5-sonnet-latest", "gemini-2.0-flash")
+ * @param modelId - Model ID (e.g., "gpt-5-nano", "claude-3-5-sonnet-latest", "gemini-2.0-flash")
  * @param options - Provider-specific configuration options
  * @returns Provider-specific model instance ready for use with Vercel AI SDK
  *
@@ -10873,7 +11047,7 @@ type ProviderOptionsForModel<T extends SupportedModelId> = T extends OpenAIChatM
  * @example
  * ```typescript
  * // OpenAI with organization
- * const openaiModel = detectProvider("gpt-4o", {
+ * const openaiModel = detectProvider("gpt-5-nano", {
  *   apiKey: "sk-...",
  *   organization: "org-123"
  * });
@@ -10903,20 +11077,20 @@ type VercelAIChatModelOptions = {
  * Supports:
  * - Direct LanguageModel instances for maximum flexibility
  * - Typed model IDs with automatic provider detection:
- *   - OpenAI: `"gpt-4o"`, `"gpt-4-turbo"`, `"o1"`, etc.
+ *   - OpenAI: `"gpt-5-nano"`, `"gpt-5"`, `"o1"`, etc.
  *   - Anthropic: `"claude-3-5-sonnet-latest"`, `"claude-3-opus"`, etc.
  *   - Google Gemini: `"gemini-2.0-flash"`, `"gemini-1.5-pro"`, etc.
  *
  * @example
  * ```typescript
  * // Using typed model ID
- * const model1 = new VercelAIChatModel("gpt-4o", {
+ * const model1 = new VercelAIChatModel("gpt-5-nano", {
  *   apiKey: "sk-...",
  *   organization: "org-123"
  * });
  *
  * // Using LanguageModel instance directly
- * const customModel = openai("gpt-4o");
+ * const customModel = openai("gpt-5-nano");
  * const model2 = new VercelAIChatModel(customModel);
  * ```
  */
@@ -10938,7 +11112,7 @@ declare class VercelAIChatModel extends OpikBaseModel {
     /**
      * Creates a new VercelAIChatModel instance with a typed model ID.
      *
-     * @param modelId - The model ID (e.g., 'gpt-4o', 'claude-3-5-sonnet-latest', 'gemini-2.0-flash')
+     * @param modelId - The model ID (e.g., 'gpt-5-nano', 'claude-3-5-sonnet-latest', 'gemini-2.0-flash')
      * @param options - Provider-specific configuration options
      */
     constructor(modelId: SupportedModelId, options?: VercelAIChatModelOptions);
@@ -10954,7 +11128,7 @@ declare class VercelAIChatModel extends OpikBaseModel {
      *
      * @example
      * ```typescript
-     * const model = new VercelAIChatModel("gpt-4o");
+     * const model = new VercelAIChatModel("gpt-5-nano");
      *
      * // Simple text generation
      * const response = await model.generateString("What is 2+2?");
@@ -10982,7 +11156,7 @@ declare class VercelAIChatModel extends OpikBaseModel {
      *
      * @example
      * ```typescript
-     * const model = new VercelAIChatModel("gpt-4o");
+     * const model = new VercelAIChatModel("gpt-5-nano");
      * const response = await model.generateProviderResponse([
      *   { role: 'user', content: 'Hello!' }
      * ]);
@@ -10999,14 +11173,14 @@ declare class VercelAIChatModel extends OpikBaseModel {
  * Supports multiple providers (OpenAI, Anthropic, Google Gemini) with automatic
  * provider detection based on model ID patterns.
  *
- * @param modelId - Model ID (e.g., 'gpt-4o', 'claude-3-5-sonnet-latest', 'gemini-2.0-flash')
+ * @param modelId - Model ID (e.g., 'gpt-5-nano', 'claude-3-5-sonnet-latest', 'gemini-2.0-flash')
  * @param options - Optional provider-specific configuration options
  * @returns An OpikBaseModel instance
  *
  * @example
  * ```typescript
  * // OpenAI with organization
- * const model1 = createModel('gpt-4o', {
+ * const model1 = createModel('gpt-5-nano', {
  *   apiKey: 'sk-...',
  *   organization: 'org-123'
  * });
@@ -11039,7 +11213,7 @@ declare function createModel(modelId: SupportedModelId, options?: VercelAIChatMo
  * import { anthropic } from '@ai-sdk/anthropic';
  *
  * // OpenAI with custom settings
- * const customOpenAI = openai('gpt-4o', {
+ * const customOpenAI = openai('gpt-5-nano', {
  *   structuredOutputs: true,
  * });
  * const model1 = createModelFromInstance(customOpenAI);
@@ -11056,7 +11230,7 @@ declare function createModelFromInstance(languageModel: LanguageModel, options?:
  * Resolves a model identifier to an OpikBaseModel instance.
  *
  * This function implements a resolution strategy that handles multiple input types:
- * 1. undefined/null → Creates default model (gpt-4o)
+ * 1. undefined/null → Creates default model (gpt-5-nano)
  * 2. string → Creates model from model ID
  * 3. OpikBaseModel → Returns as-is
  * 4. LanguageModel → Wraps in OpikBaseModel adapter
@@ -11074,13 +11248,13 @@ declare function createModelFromInstance(languageModel: LanguageModel, options?:
  * const model1 = resolveModel();
  *
  * // Using model ID
- * const model2 = resolveModel('gpt-4o');
+ * const model2 = resolveModel('gpt-5-nano');
  *
  * // Using OpikBaseModel instance
- * const model3 = resolveModel(new VercelAIChatModel('gpt-4o'));
+ * const model3 = resolveModel(new VercelAIChatModel('gpt-5-nano'));
  *
  * // Using LanguageModel instance
- * const model4 = resolveModel(openai('gpt-4o'));
+ * const model4 = resolveModel(openai('gpt-5-nano'));
  * ```
  */
 declare function resolveModel(model?: SupportedModelId | LanguageModel | OpikBaseModel, options?: VercelAIChatModelOptions): OpikBaseModel;
@@ -11120,7 +11294,7 @@ declare class ModelConfigurationError extends ModelError {
 interface EvaluatePromptOptions extends Omit<EvaluateOptions, "task"> {
     /** Message templates with {{placeholders}} to be formatted with dataset variables */
     messages: OpikMessage[];
-    /** Model to use for generation. Can be model ID string, LanguageModel instance, or OpikBaseModel instance. Defaults to gpt-4o */
+    /** Model to use for generation. Can be model ID string, LanguageModel instance, or OpikBaseModel instance. Defaults to gpt-5-nano */
     model?: SupportedModelId | LanguageModel | OpikBaseModel;
     /** Template engine type for variable substitution. Defaults to mustache */
     templateType?: PromptType;
@@ -11153,7 +11327,7 @@ interface EvaluatePromptOptions extends Omit<EvaluateOptions, "task"> {
  *   messages: [
  *     { role: 'user', content: 'Translate to {{language}}: {{text}}' }
  *   ],
- *   model: 'gpt-4o', // or omit to use default model
+ *   model: 'gpt-5-nano', // or omit to use default model
  *   temperature: 0.7,
  *   seed: 42,
  *   scoringMetrics: [new Equals()],
@@ -11161,7 +11335,7 @@ interface EvaluatePromptOptions extends Omit<EvaluateOptions, "task"> {
  *
  * // Using pre-configured LanguageModel instance
  * import { openai } from '@ai-sdk/openai';
- * const customModel = openai('gpt-4o', { structuredOutputs: true });
+ * const customModel = openai('gpt-5-nano', { structuredOutputs: true });
  * const result2 = await evaluatePrompt({
  *   dataset,
  *   messages: [
@@ -11407,7 +11581,7 @@ declare abstract class BaseLLMJudgeMetric extends BaseMetric {
      *
      * @param name - The name of the metric
      * @param options - Configuration options
-     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-4o'.
+     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-5-nano'.
      * @param options.trackMetric - Whether to track the metric. Defaults to true.
      * @param options.temperature - Temperature setting (0.0-2.0). Controls randomness. Lower values make output more focused and deterministic.
      * @param options.seed - Random seed for reproducible outputs. Useful for testing and debugging.
@@ -11548,7 +11722,7 @@ type Input$4 = z.infer<typeof validationSchema$4>;
  * ```typescript
  * import { Moderation } from 'opik/evaluation/metrics';
  *
- * // Using default model (gpt-4o)
+ * // Using default model (gpt-5-nano)
  * const metric = new Moderation();
  * const result = await metric.score({ output: "Hello, how can I help you?" });
  * console.log(result.value);  // 0.0 (safe content)
@@ -11556,14 +11730,14 @@ type Input$4 = z.infer<typeof validationSchema$4>;
  *
  * // Using custom model with temperature and seed
  * const customMetric = new Moderation({
- *   model: 'gpt-4-turbo',
+ *   model: 'gpt-5',
  *   temperature: 0.3,
  *   seed: 42
  * });
  *
  * // Using custom model instance
  * import { openai } from '@ai-sdk/openai';
- * const customModel = openai('gpt-4o');
+ * const customModel = openai('gpt-5-nano');
  * const instanceMetric = new Moderation({ model: customModel });
  *
  * // With advanced settings
@@ -11583,7 +11757,7 @@ declare class Moderation extends BaseLLMJudgeMetric {
      * Creates a new Moderation metric.
      *
      * @param options - Configuration options
-     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-4o'.
+     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-5-nano'.
      * @param options.name - The name of the metric. Defaults to "moderation_metric".
      * @param options.fewShotExamples - Optional few-shot examples to guide the model
      * @param options.trackMetric - Whether to track the metric. Defaults to true.
@@ -11658,7 +11832,7 @@ type Input$3 = z.infer<typeof validationSchema$3>;
  * ```typescript
  * import { Usefulness } from 'opik/evaluation/metrics';
  *
- * // Using default model (gpt-4o)
+ * // Using default model (gpt-5-nano)
  * const metric = new Usefulness();
  * const result = await metric.score({
  *   input: "What's the capital of France?",
@@ -11669,14 +11843,14 @@ type Input$3 = z.infer<typeof validationSchema$3>;
  *
  * // Using custom model with temperature and seed
  * const customMetric = new Usefulness({
- *   model: 'gpt-4-turbo',
+ *   model: 'gpt-5',
  *   temperature: 0.7,
  *   seed: 42
  * });
  *
  * // Using custom model instance
  * import { openai } from '@ai-sdk/openai';
- * const customModel = openai('gpt-4o');
+ * const customModel = openai('gpt-5-nano');
  * const instanceMetric = new Usefulness({ model: customModel });
  *
  * // With advanced settings
@@ -11695,7 +11869,7 @@ declare class Usefulness extends BaseLLMJudgeMetric {
      * Creates a new Usefulness metric.
      *
      * @param options - Configuration options
-     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-4o'.
+     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-5-nano'.
      * @param options.name - The name of the metric. Defaults to "usefulness_metric".
      * @param options.trackMetric - Whether to track the metric. Defaults to true.
      * @param options.temperature - Temperature setting (0.0-2.0). Controls randomness. Lower values make output more focused and deterministic. See https://ai-sdk.dev/docs/reference/ai-sdk-core/generate-text#temperature
@@ -11776,7 +11950,7 @@ type Input$2 = z.infer<typeof validationSchema$2>;
  * ```typescript
  * import { Hallucination } from 'opik/evaluation/metrics';
  *
- * // Using default model (gpt-4o)
+ * // Using default model (gpt-5-nano)
  * const metric = new Hallucination();
  *
  * // With context
@@ -11797,7 +11971,7 @@ type Input$2 = z.infer<typeof validationSchema$2>;
  *
  * // Using custom model with few-shot examples
  * const customMetric = new Hallucination({
- *   model: 'gpt-4-turbo',
+ *   model: 'gpt-5',
  *   temperature: 0.3,
  *   seed: 42,
  *   fewShotExamples: [
@@ -11828,7 +12002,7 @@ declare class Hallucination extends BaseLLMJudgeMetric {
      * Creates a new Hallucination metric.
      *
      * @param options - Configuration options
-     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-4o'.
+     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-5-nano'.
      * @param options.name - The name of the metric. Defaults to "hallucination_metric".
      * @param options.fewShotExamples - Optional few-shot examples to guide the model
      * @param options.trackMetric - Whether to track the metric. Defaults to true.
@@ -11924,7 +12098,7 @@ type Input$1 = z.infer<typeof validationSchema$1>;
  * ```typescript
  * import { AnswerRelevance } from 'opik/evaluation/metrics';
  *
- * // Using default model (gpt-4o)
+ * // Using default model (gpt-5-nano)
  * const metric = new AnswerRelevance();
  *
  * // With context (default behavior)
@@ -11946,7 +12120,7 @@ type Input$1 = z.infer<typeof validationSchema$1>;
  *
  * // Using custom model with few-shot examples
  * const customMetric = new AnswerRelevance({
- *   model: 'gpt-4-turbo',
+ *   model: 'gpt-5',
  *   temperature: 0.3,
  *   seed: 42,
  *   fewShotExamples: [
@@ -11981,7 +12155,7 @@ declare class AnswerRelevance extends BaseLLMJudgeMetric {
      * Creates a new AnswerRelevance metric.
      *
      * @param options - Configuration options
-     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-4o'.
+     * @param options.model - The language model to use. Can be a string (model ID), LanguageModel instance, or OpikBaseModel instance. Defaults to 'gpt-5-nano'.
      * @param options.name - The name of the metric. Defaults to "answer_relevance_metric".
      * @param options.fewShotExamples - Optional few-shot examples with context to guide the model. If not provided, default examples will be used.
      * @param options.fewShotExamplesNoContext - Optional few-shot examples without context for no-context mode. If not provided, default examples will be used.
@@ -12392,7 +12566,7 @@ declare class DatasetVersionNotFoundError extends OpikError {
 /**
  * Supported column types in OQL queries
  */
-type ColumnType = "string" | "date_time" | "dictionary" | "feedback_scores_number" | "list" | "number";
+type ColumnType = "string" | "date_time" | "dictionary" | "feedback_scores_number" | "list" | "number" | "error_container" | "enum" | "map";
 /**
  * Parsed filter expression structure
  */
@@ -12400,10 +12574,33 @@ interface FilterExpression {
     field: string;
     key?: string;
     operator: string;
-    value: string;
+    value: string | null;
     type?: ColumnType;
 }
 
+/**
+ * Abstract base class for OQL (Opik Query Language) configuration
+ */
+
+declare abstract class OQLConfig {
+    /**
+     * Map of supported fields to their types
+     */
+    abstract get columns(): Record<string, ColumnType>;
+    /**
+     * Map of fields to their supported operators
+     */
+    abstract get supportedOperators(): Record<string, readonly string[]>;
+    /**
+     * Fields that support nested key access via dot notation
+     */
+    get nestedFields(): readonly string[];
+    /**
+     * Keys supported for the usage field
+     */
+    get usageKeys(): readonly string[];
+}
+
 /**
  * This file contains the OQL parser and validator. It is currently limited in scope to only support
  * simple filters without "and" or "or" operators.
@@ -12431,7 +12628,24 @@ interface FilterExpression {
 declare class OpikQueryLanguage {
     private readonly filterExpressions;
     readonly parsedFilters: string | null;
-    constructor(queryString?: string);
+    private readonly config;
+    constructor(queryString?: string, config?: OQLConfig);
+    /**
+     * Create an OpikQueryLanguage instance for trace filtering
+     */
+    static forTraces(queryString?: string): OpikQueryLanguage;
+    /**
+     * Create an OpikQueryLanguage instance for span filtering
+     */
+    static forSpans(queryString?: string): OpikQueryLanguage;
+    /**
+     * Create an OpikQueryLanguage instance for trace thread filtering
+     */
+    static forThreads(queryString?: string): OpikQueryLanguage;
+    /**
+     * Create an OpikQueryLanguage instance for prompt filtering
+     */
+    static forPrompts(queryString?: string): OpikQueryLanguage;
     /**
      * Returns the parsed filter expressions
      */