@wix/evalforge-evaluator 0.146.0 → 0.148.0

package/build/types/run-scenario/agents/claude-code/types.d.ts

@@ -4,7 +4,7 @@
  * These types are internal to the Claude Code adapter and extend or
  * specialize the shared types from @wix/evalforge-types.
  */
-import type { MCPEntity, SubAgent, Rule } from '@wix/evalforge-types';
+import type { MCPEntity, SubAgent, Rule, EffortLevel } from '@wix/evalforge-types';
 export type { TraceContext } from '@wix/evalforge-types';
 /**
  * Options for Claude Agent SDK execution.
@@ -23,14 +23,24 @@ export interface ClaudeCodeExecutionOptions {
     rules?: Rule[];
     /** Model to use for generation */
     model?: string;
-    /** Temperature for generation (0–1, from agent modelConfig) */
+    /** Temperature for generation (0–1, from agent config) */
     temperature?: number;
-    /** Max output tokens (from agent modelConfig) */
+    /** Max output tokens (from agent config) */
     maxTokens?: number;
     /** Maximum number of agent turns */
     maxTurns?: number;
     /** Maximum thinking tokens */
     maxThinkingTokens?: number;
+    /** Override the default allowedTools list */
+    allowedTools?: string[];
+    /** Tools to remove from the model's context entirely */
+    disallowedTools?: string[];
+    /** Controls thinking depth: low, medium, high, max */
+    effort?: EffortLevel;
+    /** Maximum USD spend per run */
+    maxBudgetUsd?: number;
+    /** Execution timeout in milliseconds (overrides the default maxTurns-based calculation) */
+    maxDurationMs?: number;
     /** AI Gateway base URL for Anthropic API */
     aiGatewayUrl?: string;
     /** AI Gateway custom headers as key-value pairs */
@@ -44,6 +54,8 @@ export interface ClaudeCodeExecutionOptions {
      * - string: custom system prompt text
      */
     systemPrompt?: string | null;
+    /** Extra config properties not in the known schema. Spread into SDK query options. */
+    extras?: Record<string, unknown>;
 }
 /**
  * Result from Claude Code execution.

package/build/types/run-scenario/agents/opencode/config.d.ts

@@ -2,11 +2,14 @@ import { type MCPEntity } from '@wix/evalforge-types';
 export interface OpenCodeConfigOptions {
     model?: string;
     temperature?: number;
+    topP?: number;
     maxTurns?: number;
     aiGatewayUrl?: string;
     aiGatewayHeaders?: Record<string, string>;
     mcps?: MCPEntity[];
     cwd: string;
+    /** Agent config bag — may contain permission overrides */
+    config?: Record<string, unknown>;
 }
 /**
  * Build environment variables for spawning `opencode run`.

package/build/types/run-scenario/agents/opencode/types.d.ts

@@ -58,11 +58,19 @@ export interface OpenCodeExecutionOptions {
     rules?: Rule[];
     model?: string;
     temperature?: number;
+    /** Nucleus sampling (0–1). Alternative to temperature. */
+    topP?: number;
     maxTurns?: number;
+    /** Thinking variant: 'high' | 'low' | 'none'. Default: 'high'. 'none' omits --thinking flag. */
+    thinkingVariant?: string;
+    /** Execution timeout override in milliseconds */
+    maxDurationMs?: number;
     aiGatewayUrl?: string;
     aiGatewayHeaders?: Record<string, string>;
     traceContext?: import('@wix/evalforge-types').TraceContext;
     systemPrompt?: string | null;
+    /** Agent config bag for passing to buildOpenCodeEnv (permission overrides, etc.) */
+    config?: Record<string, unknown>;
 }
 export interface OpenCodeExecutionResult {
     outputText: string;

package/build/types/run-scenario/agents/simple-agent/execute.d.ts

@@ -2,6 +2,7 @@ import { type LanguageModel, type StepResult, type ToolSet } from 'ai';
 import type { AgentExecutionContext, AgentExecutionResult, LLMTrace } from '@wix/evalforge-types';
 export declare function createModel(modelId: string, baseUrl: string, headers: Record<string, string>): LanguageModel;
 export declare function isClaudeModelId(modelId: string): boolean;
+export declare function isGeminiModelId(modelId: string): boolean;
 /**
  * Extract text content from a skill's latest version files.
  * Concatenates all file contents, prioritizing SKILL.md if present.

package/build/types/run-scenario/agents/simple-agent/mcp-tools.d.ts

@@ -12,8 +12,6 @@ interface McpToolsResult {
  * Each `MCPEntity.config` is a keyed record matching the `.mcp.json` shape,
  * e.g. `{ "server-name": { "type": "stdio", "command": "node", "args": ["server.js"] } }`.
  * Each entry becomes a live MCP client whose tools are merged into the result.
- *
- * Tool names are namespaced as `${serverName}__${toolName}` to avoid collisions.
  */
 export declare function buildMcpTools(mcps: MCPEntity[], cwd: string): Promise<McpToolsResult>;
 export declare function closeMcpClients(clients: Array<{

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wix/evalforge-evaluator",
-  "version": "0.146.0",
+  "version": "0.148.0",
   "description": "EvalForge Evaluator",
   "bin": "./build/index.js",
   "files": [
@@ -8,8 +8,8 @@
   ],
   "scripts": {
     "clean": "rm -rf build",
-    "build:cjs": "esbuild src/index.ts --bundle --platform=node --outfile=build/index.js --format=cjs --sourcemap --external:@wix/evalforge-types --external:@wix/evalforge-github-client --external:@wix/eval-assertions --external:@anthropic-ai/claude-agent-sdk --external:@ai-sdk/mcp --external:@ai-sdk/anthropic --external:@ai-sdk/openai --external:ai",
-    "build:esm": "esbuild src/index.ts --bundle --platform=node --outfile=build/index.mjs --format=esm --sourcemap --external:@wix/evalforge-types --external:@wix/evalforge-github-client --external:@wix/eval-assertions --external:@anthropic-ai/claude-agent-sdk --external:@ai-sdk/mcp --external:@ai-sdk/anthropic --external:@ai-sdk/openai --external:ai",
+    "build:cjs": "esbuild src/index.ts --bundle --platform=node --outfile=build/index.js --format=cjs --sourcemap --external:@wix/evalforge-types --external:@wix/evalforge-github-client --external:@wix/eval-assertions --external:@anthropic-ai/claude-agent-sdk --external:@ai-sdk/mcp --external:@ai-sdk/anthropic --external:@ai-sdk/google --external:@ai-sdk/openai --external:ai",
+    "build:esm": "esbuild src/index.ts --bundle --platform=node --outfile=build/index.mjs --format=esm --sourcemap --external:@wix/evalforge-types --external:@wix/evalforge-github-client --external:@wix/eval-assertions --external:@anthropic-ai/claude-agent-sdk --external:@ai-sdk/mcp --external:@ai-sdk/anthropic --external:@ai-sdk/google --external:@ai-sdk/openai --external:ai",
     "build:types": "tsc --emitDeclarationOnly --outDir ./build/types",
     "build": "yarn run clean && yarn run build:cjs && yarn run build:esm && yarn run build:types",
     "lint": "eslint .",
@@ -17,13 +17,14 @@
   },
   "dependencies": {
     "@ai-sdk/anthropic": "^3.0.53",
+    "@ai-sdk/google": "^3.0.53",
     "@ai-sdk/mcp": "^1.0.23",
     "@ai-sdk/openai": "^3.0.39",
     "@anthropic-ai/claude-agent-sdk": "^0.2.63",
     "@anthropic-ai/claude-code": "^2.1.63",
-    "@wix/eval-assertions": "0.48.0",
-    "@wix/evalforge-github-client": "0.52.0",
-    "@wix/evalforge-types": "0.77.0",
+    "@wix/eval-assertions": "0.50.0",
+    "@wix/evalforge-github-client": "0.54.0",
+    "@wix/evalforge-types": "0.79.0",
     "ai": "^6.0.107",
     "diff": "^7.0.0",
     "tar": "^7.5.3",
@@ -62,5 +63,5 @@
       "artifactId": "evalforge-evaluator"
     }
   },
-  "falconPackageHash": "d059d88009317e706f5ada0c0ec10ce11addfe3a507739fb93c30a82"
+  "falconPackageHash": "98c865c8379433039867350fce6148475e6f7daae8c73d9e8da29dd5"
 }