beddel 0.2.2 → 0.3.0

package/dist/runtime/declarativeAgentRuntime.js

@@ -2,6 +2,8 @@
 /**
  * Declarative Agent Runtime - YAML Interpreter for Beddel Declarative Protocol
  * Safely interprets declarative YAML agent definitions without dynamic code execution
+ *
+ * Phase 3 Refactored: Delegates to individual agent handlers via workflowExecutor
  */
 var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
     if (k2 === undefined) k2 = k;
@@ -38,22 +40,21 @@ var __importStar = (this && this.__importStar) || (function () {
 })();
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.declarativeInterpreter = exports.DeclarativeAgentInterpreter = void 0;
+require("server-only");
 const yaml = __importStar(require("js-yaml"));
-const ai_1 = require("ai");
-const google_1 = require("@ai-sdk/google");
-const agentRegistry_1 = require("../agents/agentRegistry");
+const registry_1 = require("../agents/registry");
 const schemaCompiler_1 = require("./schemaCompiler");
+// Import handlers from workflowExecutor
+const workflowExecutor_1 = require("./workflowExecutor");
 /**
  * Safe declarative YAML interpreter - no dynamic code execution
+ * Delegates execution to individual agent handlers
  */
 class DeclarativeAgentInterpreter {
     constructor() {
-        this.MAX_VARIABLE_SIZE = 1024; // 1KB max variable size
-        this.MAX_WORKFLOW_STEPS = 100; // Prevent infinite loops
-        this.MAX_OUTPUT_SIZE = 5 * 1024 * 1024; // 5MB max output to accommodate image payloads
-        this.GEMINI_MODEL = "models/gemini-2.5-flash";
-        this.GEMINI_IMAGE_MODEL = "imagen-4.0-fast-generate-001";
-        this.SUPPORTED_TRANSLATION_LANGUAGES = ["pt", "en", "es", "fr"];
+        this.MAX_VARIABLE_SIZE = 1024;
+        this.MAX_WORKFLOW_STEPS = 100;
+        this.MAX_OUTPUT_SIZE = 5 * 1024 * 1024;
         this.schemaCompiler = new schemaCompiler_1.DeclarativeSchemaCompiler();
     }
     /**
@@ -62,45 +63,35 @@ class DeclarativeAgentInterpreter {
     async interpret(options) {
         const startTime = Date.now();
         try {
-            // Parse and validate YAML
             const agent = this.parseYaml(options.yamlContent);
             this.validateAgentDefinition(agent);
-            // Compile schemas and validate input up front
             const schemas = this.buildSchemaSet(agent);
-            const validatedInput = this.validateAgainstSchema(options.input, schemas.input, "input", options.context);
+            const validatedInput = this.validateAgainstSchema(options.input, schemas.input, 'input', options.context);
             const executionOptions = {
                 ...options,
                 input: validatedInput,
             };
-            // Execute declarative logic
             const result = await this.executeWorkflow(agent, executionOptions);
-            // Validate output
-            const validatedOutput = this.validateAgainstSchema(result, schemas.output, "output", options.context);
+            const validatedOutput = this.validateAgainstSchema(result, schemas.output, 'output', options.context);
             this.enforceOutputSize(validatedOutput);
             const executionTime = Date.now() - startTime;
             options.context.log(`Declarative agent executed in ${executionTime}ms`);
             return validatedOutput;
         }
         catch (error) {
-            const executionTime = Date.now() - startTime;
             options.context.log(`Declarative agent execution failed: ${error}`);
-            options.context.setError(error instanceof Error
-                ? error.message
-                : "Unknown declarative agent error");
+            options.context.setError(error instanceof Error ? error.message : 'Unknown declarative agent error');
             throw error;
         }
     }
-    /**
-     * Parse and validate YAML content
-     */
     parseYaml(yamlContent) {
         try {
             const parsed = yaml.load(yamlContent);
-            if (!parsed || typeof parsed !== "object") {
-                throw new Error("Invalid YAML: expected object");
+            if (!parsed || typeof parsed !== 'object') {
+                throw new Error('Invalid YAML: expected object');
             }
             if (!parsed.agent || !parsed.logic || !parsed.schema) {
-                throw new Error("Invalid agent definition: missing required sections");
+                throw new Error('Invalid agent definition: missing required sections');
             }
             return parsed;
         }
@@ -108,22 +99,15 @@ class DeclarativeAgentInterpreter {
             throw new Error(`YAML parsing failed: ${error}`);
         }
     }
-    /**
-     * Validate agent definition structure
-     */
     validateAgentDefinition(agent) {
-        // Validate protocol version
-        if (agent.agent.protocol !== "beddel-declarative-protocol/v2.0") {
+        if (agent.agent.protocol !== 'beddel-declarative-protocol/v2.0') {
             throw new Error(`Unsupported protocol: ${agent.agent.protocol}`);
         }
-        // Validate schema
         if (!agent.schema.input || !agent.schema.output) {
-            throw new Error("Invalid schema: missing input or output definition");
+            throw new Error('Invalid schema: missing input or output definition');
         }
-        // Validate workflow
-        if (!Array.isArray(agent.logic.workflow) ||
-            agent.logic.workflow.length === 0) {
-            throw new Error("Invalid workflow: must be non-empty array");
+        if (!Array.isArray(agent.logic.workflow) || agent.logic.workflow.length === 0) {
+            throw new Error('Invalid workflow: must be non-empty array');
         }
         if (agent.logic.workflow.length > this.MAX_WORKFLOW_STEPS) {
             throw new Error(`Workflow too complex: max ${this.MAX_WORKFLOW_STEPS} steps allowed`);
@@ -131,8 +115,8 @@ class DeclarativeAgentInterpreter {
     }
     buildSchemaSet(agent) {
         return {
-            input: this.schemaCompiler.compile(agent.schema.input, "schema.input"),
-            output: this.schemaCompiler.compile(agent.schema.output, "schema.output"),
+            input: this.schemaCompiler.compile(agent.schema.input, 'schema.input'),
+            output: this.schemaCompiler.compile(agent.schema.output, 'schema.output'),
         };
     }
     validateAgainstSchema(data, schema, phase, context) {
@@ -140,9 +124,9 @@ class DeclarativeAgentInterpreter {
         if (!validationResult.success) {
             const issues = validationResult.error.issues;
             const issueSummary = issues
-                .map((issue) => `${issue.path.join(".") || "root"}: ${issue.message}`)
-                .join("; ");
-            const label = phase === "input" ? "Input" : "Output";
+                .map((issue) => `${issue.path.join('.') || 'root'}: ${issue.message}`)
+                .join('; ');
+            const label = phase === 'input' ? 'Input' : 'Output';
             const message = `${label} validation failed: ${issueSummary}`;
             context.setError(message);
             throw new schemaCompiler_1.DeclarativeSchemaValidationError(message, phase, issues);
@@ -155,13 +139,9 @@ class DeclarativeAgentInterpreter {
             throw new Error(`Output size exceeds maximum allowed: ${outputSize} > ${this.MAX_OUTPUT_SIZE}`);
         }
     }
-    /**
-     * Execute declarative workflow
-     */
     async executeWorkflow(agent, options) {
         const variables = new Map();
         let output = undefined;
-        // Initialize variables
         if (agent.logic.variables) {
             for (const variable of agent.logic.variables) {
                 this.validateVariable(variable);
@@ -169,51 +149,62 @@ class DeclarativeAgentInterpreter {
                 variables.set(variable.name, value);
             }
         }
-        // Execute workflow steps
         for (const step of agent.logic.workflow) {
             output = await this.executeWorkflowStep(step, variables, options);
         }
         return output;
     }
-    /**
-     * Execute single workflow step
-     */
     async executeWorkflowStep(step, variables, options) {
         options.context.log(`Executing workflow step: ${step.name} (${step.type})`);
         switch (step.type) {
-            case "output-generator":
+            case 'output-generator':
                 return this.executeOutputGenerator(step, variables, options);
-            case "genkit-joke":
-                return this.executeGenkitJoke(step, variables, options);
-            case "genkit-translation":
-                return this.executeGenkitTranslation(step, variables, options);
-            case "genkit-image":
-                return this.executeGenkitImage(step, variables, options);
-            case "custom-action":
+            // Preferred step types
+            case 'joke':
+            case 'genkit-joke':
+                return this.executeJoke(step, variables, options);
+            case 'translation':
+            case 'genkit-translation':
+                return this.executeTranslation(step, variables, options);
+            case 'image':
+            case 'genkit-image':
+                return this.executeImage(step, variables, options);
+            case 'vectorize':
+            case 'gemini-vectorize':
+                return this.executeVectorize(step, variables, options);
+            case 'custom-action':
                 return this.executeCustomAction(step, variables, options);
+            case 'mcp-tool':
+                return this.executeMcpTool(step, variables, options);
+            case 'chromadb':
+                return this.executeChromaDB(step, variables, options);
+            case 'gitmcp':
+                return this.executeGitMcp(step, variables, options);
+            case 'rag':
+                return this.executeRag(step, variables, options);
+            case 'chat':
+                return this.executeChat(step, variables, options);
+            case 'builtin-agent':
+                return this.executeBuiltinAgent(step, variables, options);
             default:
                 throw new Error(`Unsupported workflow step type: ${step.type}`);
         }
     }
-    /**
-     * Execute output generator step
-     */
+    // ============================================================================
+    // Output Generator (kept inline - orchestration logic)
+    // ============================================================================
     executeOutputGenerator(step, variables, options) {
-        if (step.action?.type !== "generate" || !step.action.output) {
-            throw new Error("Invalid output generator configuration");
+        if (step.action?.type !== 'generate' || !step.action.output) {
+            throw new Error('Invalid output generator configuration');
         }
-        // Build output object
         const output = {};
-        // Debug: Log available variables
-        options.context.log(`Output generator: Available variables: ${Array.from(variables.keys()).join(", ")}`);
+        options.context.log(`Output generator: Available variables: ${Array.from(variables.keys()).join(', ')}`);
         for (const [key, valueExpr] of Object.entries(step.action.output)) {
-            if (typeof valueExpr === "string" && valueExpr.startsWith("$")) {
+            if (typeof valueExpr === 'string' && valueExpr.startsWith('$')) {
                 try {
                     const reference = valueExpr.substring(1);
-                    options.context.log(`Output generator: Resolving reference ${valueExpr} -> ${reference}`);
                     const resolved = this.resolveReference(reference, variables);
                     output[key] = resolved;
-                    options.context.log(`Output generator: Resolved ${key} = ${typeof resolved === "string" ? resolved.substring(0, 50) + "..." : JSON.stringify(resolved).substring(0, 100)}`);
                 }
                 catch (error) {
                     options.context.log(`Output generator: Failed to resolve ${valueExpr}: ${error instanceof Error ? error.message : String(error)}`);
@@ -224,109 +215,177 @@ class DeclarativeAgentInterpreter {
                 output[key] = valueExpr;
             }
         }
-        options.context.log(`Output generator: Final output keys: ${Object.keys(output).join(", ")}`);
         return output;
     }
-    /**
-     * Execute Gemini Flash text helper for the joke agent
-     */
-    async executeGenkitJoke(step, variables, options) {
-        const prompt = typeof step.action?.prompt === "string" && step.action.prompt.trim().length
+    // ============================================================================
+    // Delegated Handlers - Using extracted agent handlers
+    // ============================================================================
+    async executeJoke(step, variables, options) {
+        const prompt = typeof step.action?.prompt === 'string' && step.action.prompt.trim().length
             ? step.action.prompt.trim()
-            : "Conte uma piada curta e original em português.";
-        const temperature = typeof step.action?.temperature === "number"
-            ? step.action.temperature
-            : 0.8;
-        const maxTokens = typeof step.action?.maxTokens === "number"
-            ? step.action.maxTokens
-            : undefined;
-        const resultVar = typeof step.action?.result === "string" && step.action.result.length > 0
+            : 'Tell a short and original joke that works for any audience.';
+        const temperature = typeof step.action?.temperature === 'number' ? step.action.temperature : 0.8;
+        const maxTokens = typeof step.action?.maxTokens === 'number' ? step.action.maxTokens : undefined;
+        const resultVar = typeof step.action?.result === 'string' && step.action.result.length > 0
             ? step.action.result
-            : "jokerResult";
-        const joke = await this.callGeminiFlashText({ prompt, temperature, maxTokens }, options.props, options.context);
-        variables.set(resultVar, joke);
-        return joke;
+            : 'jokerResult';
+        const result = await (0, workflowExecutor_1.executeJokeHandler)({ prompt, temperature, maxTokens }, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
     }
-    /**
-     * Execute translation step backed by Gemini Flash
-     */
-    async executeGenkitTranslation(step, variables, options) {
-        const texto = options.input?.texto;
-        const idiomaOrigem = options.input?.idioma_origem;
-        const idiomaDestino = options.input?.idioma_destino;
-        const resultVar = typeof step.action?.result === "string" && step.action.result.length > 0
+    async executeTranslation(step, variables, options) {
+        const text = options.input?.texto || options.input?.text;
+        const sourceLanguage = options.input?.idioma_origem || options.input?.source_language;
+        const targetLanguage = options.input?.idioma_destino || options.input?.target_language;
+        const resultVar = typeof step.action?.result === 'string' && step.action.result.length > 0
             ? step.action.result
-            : "translationResult";
-        const translation = await this.callGeminiFlashTranslation({
-            texto,
-            idioma_origem: idiomaOrigem,
-            idioma_destino: idiomaDestino,
-            promptTemplate: typeof step.action?.promptTemplate === "string"
+            : 'translationResult';
+        const result = await (0, workflowExecutor_1.executeTranslationHandler)({
+            text,
+            source_language: sourceLanguage,
+            target_language: targetLanguage,
+            promptTemplate: typeof step.action?.promptTemplate === 'string'
                 ? step.action.promptTemplate
                 : undefined,
         }, options.props, options.context);
-        variables.set(resultVar, translation);
-        return translation;
+        variables.set(resultVar, result);
+        return result;
     }
-    /**
-     * Execute image generation step backed by Gemini Flash
-     */
-    async executeGenkitImage(step, variables, options) {
-        const descricao = typeof options.input?.descricao === "string"
+    async executeImage(step, variables, options) {
+        const description = typeof options.input?.descricao === 'string'
             ? options.input.descricao.trim()
-            : "";
-        const estilo = typeof options.input?.estilo === "string"
+            : typeof options.input?.description === 'string'
+                ? options.input.description.trim()
+                : '';
+        const style = typeof options.input?.estilo === 'string'
             ? options.input.estilo.trim()
-            : "";
-        const resolucaoInput = typeof options.input?.resolucao === "string"
+            : typeof options.input?.style === 'string'
+                ? options.input.style.trim()
+                : '';
+        const resolution = typeof options.input?.resolucao === 'string'
             ? options.input.resolucao.trim()
-            : "";
-        if (!descricao) {
-            throw new Error("Missing required image input: descricao");
+            : typeof options.input?.resolution === 'string'
+                ? options.input.resolution.trim()
+                : '';
+        if (!description) {
+            throw new Error('Missing required image input: description');
         }
-        if (!estilo) {
-            throw new Error("Missing required image input: estilo");
+        if (!style) {
+            throw new Error('Missing required image input: style');
         }
-        if (!resolucaoInput) {
-            throw new Error("Missing required image input: resolucao");
+        if (!resolution) {
+            throw new Error('Missing required image input: resolution');
         }
-        const promptTemplate = typeof step.action?.promptTemplate === "string" &&
+        const promptTemplate = typeof step.action?.promptTemplate === 'string' &&
             step.action.promptTemplate.trim().length > 0
             ? step.action.promptTemplate
-            : "Gere uma imagem detalhada no estilo {{estilo}} baseada na descrição: {{descricao}}";
-        const prompt = promptTemplate
-            .replace(/{{descricao}}/g, descricao)
-            .replace(/{{estilo}}/g, estilo)
-            .trim();
-        const resultVar = typeof step.action?.result === "string" && step.action.result.length > 0
+            : 'Create a detailed image in {{style}} style focusing on: {{description}}';
+        const resultVar = typeof step.action?.result === 'string' && step.action.result.length > 0
             ? step.action.result
-            : "imageResult";
-        const imageResult = await this.callGeminiFlashImage({
-            prompt,
-            estilo,
-            resolucao: resolucaoInput,
+            : 'imageResult';
+        const result = await (0, workflowExecutor_1.executeImageHandler)({
+            description,
+            style,
+            resolution,
+            promptTemplate,
         }, options.props, options.context);
-        variables.set(resultVar, imageResult);
-        options.context.log(`Image generator: Saved result in variable '${resultVar}' with keys: ${Object.keys(imageResult).join(", ")}`);
-        options.context.log(`Image generator: imageResult.image_url exists: ${!!imageResult?.image_url}`);
-        return imageResult;
+        variables.set(resultVar, result);
+        return result;
     }
-    /**
-     * Execute custom action backed by TypeScript implementation
-     */
+    async executeMcpTool(step, variables, options) {
+        const serverUrl = this.resolveInputValue(step.action?.server_url, options.input, variables);
+        const toolName = this.resolveInputValue(step.action?.tool_name, options.input, variables);
+        const toolArguments = this.resolveInputValue(step.action?.tool_arguments, options.input, variables) || {};
+        const resultVar = step.action?.result || 'mcpResult';
+        const result = await (0, workflowExecutor_1.executeMcpToolHandler)({
+            server_url: serverUrl,
+            tool_name: toolName,
+            tool_arguments: toolArguments,
+        }, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
+    }
+    async executeVectorize(step, variables, options) {
+        const action = step.action?.action || 'embedSingle';
+        const resultVar = step.action?.result || 'vectorizeResult';
+        let result;
+        if (action === 'embedSingle') {
+            const text = this.resolveInputValue(step.action?.text, options.input, variables);
+            result = await (0, workflowExecutor_1.executeVectorizeHandler)({ action: 'embedSingle', text }, options.props, options.context);
+        }
+        else if (action === 'embedBatch') {
+            const texts = this.resolveInputValue(step.action?.texts, options.input, variables);
+            result = await (0, workflowExecutor_1.executeVectorizeHandler)({ action: 'embedBatch', texts }, options.props, options.context);
+        }
+        else {
+            throw new Error(`Unknown vectorize action: ${action}`);
+        }
+        variables.set(resultVar, result);
+        return result;
+    }
+    async executeChromaDB(step, variables, options) {
+        const action = this.resolveInputValue(step.action?.action, options.input, variables);
+        const collectionName = this.resolveInputValue(step.action?.collection_name, options.input, variables);
+        const resultVar = step.action?.result || 'chromaResult';
+        let params = { action, collection_name: collectionName };
+        if (action === 'hasData') {
+            params.min_count =
+                this.resolveInputValue(step.action?.min_count, options.input, variables) || 1;
+        }
+        else if (action === 'store') {
+            params.ids = this.resolveInputValue(step.action?.ids, options.input, variables);
+            params.vectors = this.resolveInputValue(step.action?.vectors, options.input, variables);
+            params.documents = this.resolveInputValue(step.action?.documents, options.input, variables);
+            params.metadatas = this.resolveInputValue(step.action?.metadatas, options.input, variables);
+        }
+        else if (action === 'search') {
+            params.query_vector = this.resolveInputValue(step.action?.query_vector, options.input, variables);
+            params.limit =
+                this.resolveInputValue(step.action?.limit, options.input, variables) || 5;
+        }
+        const result = await (0, workflowExecutor_1.executeChromaDBHandler)(params, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
+    }
+    async executeGitMcp(step, variables, options) {
+        const gitmcpUrl = this.resolveInputValue(step.action?.gitmcp_url, options.input, variables);
+        const resultVar = step.action?.result || 'gitmcpResult';
+        const result = await (0, workflowExecutor_1.executeGitMcpHandler)({ gitmcp_url: gitmcpUrl }, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
+    }
+    async executeRag(step, variables, options) {
+        const query = this.resolveInputValue(step.action?.query, options.input, variables);
+        const context = this.resolveInputValue(step.action?.context, options.input, variables) ||
+            this.resolveInputValue(step.action?.documents, options.input, variables);
+        const history = this.resolveInputValue(step.action?.history, options.input, variables);
+        const resultVar = step.action?.result || 'ragResult';
+        const result = await (0, workflowExecutor_1.executeRagHandler)({ query, context, documents: context, history }, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
+    }
+    async executeChat(step, variables, options) {
+        const messages = this.resolveInputValue(step.action?.messages, options.input, variables);
+        const query = this.resolveInputValue(step.action?.query, options.input, variables);
+        const resultVar = step.action?.result || 'chatResult';
+        const result = await (0, workflowExecutor_1.executeChatHandler)({ messages, query }, options.props, options.context);
+        variables.set(resultVar, result);
+        return result;
+    }
+    // ============================================================================
+    // Custom Action & Builtin Agent (orchestration logic)
+    // ============================================================================
     async executeCustomAction(step, variables, options) {
         const functionName = step.action?.function;
         if (!functionName) {
             throw new Error("Missing 'function' in custom-action");
         }
         options.context.log(`Custom action: Looking up function '${functionName}'`);
-        // Retrieve Code Implementation
-        const customFunc = agentRegistry_1.agentRegistry.getCustomFunction(functionName);
+        const customFunc = registry_1.agentRegistry.getCustomFunction(functionName);
         if (!customFunc) {
             throw new Error(`Custom function '${functionName}' not found in registry. ` +
                 `Make sure the corresponding .ts file is in the /agents directory.`);
         }
-        // Prepare Arguments
         const args = {
             input: options.input,
             variables: Object.fromEntries(variables),
@@ -334,13 +393,10 @@ class DeclarativeAgentInterpreter {
             context: options.context,
         };
         options.context.log(`Custom action: Executing function '${functionName}'`);
-        // Execute Code
         try {
             const result = await customFunc(args);
-            // Save Result
             if (step.action.result) {
                 variables.set(step.action.result, result);
-                options.context.log(`Custom action: Saved result to variable '${step.action.result}'`);
             }
             return result;
         }
@@ -351,266 +407,108 @@ class DeclarativeAgentInterpreter {
             throw new Error(`Custom action execution failed: ${errorMessage}`);
         }
     }
-    /**
-     * Evaluate value expression
-     */
+    async executeBuiltinAgent(step, variables, options) {
+        const agentName = step.action?.agent;
+        const agentInput = this.resolveInputValue(step.action?.input, options.input, variables) || options.input;
+        const agentProps = step.action?.props || options.props;
+        const resultVar = step.action?.result || 'builtinResult';
+        if (!agentName) {
+            throw new Error('Missing required builtin-agent input: agent');
+        }
+        options.context.log(`[Builtin Agent] Invoking agent: ${agentName}`);
+        try {
+            const result = await registry_1.agentRegistry.executeAgent(agentName, agentInput, agentProps, options.context);
+            variables.set(resultVar, result);
+            return result;
+        }
+        catch (error) {
+            const message = error instanceof Error ? error.message : String(error);
+            options.context.log(`[Builtin Agent] Error: ${message}`);
+            throw new Error(`Builtin agent '${agentName}' execution failed: ${message}`);
+        }
+    }
+    // ============================================================================
+    // Helper Methods
+    // ============================================================================
+    resolveInputValue(value, input, variables) {
+        if (value === undefined || value === null)
+            return undefined;
+        if (typeof value === 'string' && value.startsWith('$')) {
+            const ref = value.substring(1);
+            if (ref.startsWith('input.')) {
+                const inputKey = ref.substring(6);
+                return this.getNestedValue(input, inputKey);
+            }
+            return this.resolveReference(ref, variables);
+        }
+        if (typeof value === 'string' && input[value] !== undefined) {
+            return input[value];
+        }
+        return value;
+    }
+    getNestedValue(obj, path) {
+        const parts = path.split('.');
+        let current = obj;
+        for (const part of parts) {
+            if (current == null)
+                return undefined;
+            current = current[part];
+        }
+        return current;
+    }
     evaluateValue(expr, variables) {
-        // Handle string literals
         if (expr.startsWith('"') && expr.endsWith('"')) {
             if (expr.length - 2 > this.MAX_VARIABLE_SIZE) {
-                throw new Error("Variable initialization exceeds maximum size");
+                throw new Error('Variable initialization exceeds maximum size');
             }
             return expr.slice(1, -1);
         }
         if (expr.startsWith("'") && expr.endsWith("'")) {
             if (expr.length - 2 > this.MAX_VARIABLE_SIZE) {
-                throw new Error("Variable initialization exceeds maximum size");
+                throw new Error('Variable initialization exceeds maximum size');
             }
             return expr.slice(1, -1);
         }
         if (expr.length > this.MAX_VARIABLE_SIZE) {
-            throw new Error("Variable initialization exceeds maximum size");
+            throw new Error('Variable initialization exceeds maximum size');
         }
-        // Handle boolean literals
-        if (expr === "true")
+        if (expr === 'true')
             return true;
-        if (expr === "false")
+        if (expr === 'false')
             return false;
-        // Handle null literal
-        if (expr === "null")
+        if (expr === 'null')
             return null;
-        // Handle variable references
-        if (expr.startsWith("$")) {
+        if (expr.startsWith('$')) {
             return this.resolveReference(expr.substring(1), variables);
         }
-        // Handle numbers
         if (/^-?\d+$/.test(expr))
             return parseInt(expr, 10);
         if (/^-?\d+\.\d+$/.test(expr))
             return parseFloat(expr);
         throw new Error(`Unsupported value expression: ${expr}`);
     }
-    /**
-     * Validate variable declaration
-     */
     validateVariable(variable) {
         if (!variable.name || !variable.type) {
-            throw new Error("Invalid variable declaration: missing name or type");
+            throw new Error('Invalid variable declaration: missing name or type');
         }
-        if (!["string", "number", "boolean", "object"].includes(variable.type)) {
+        if (!['string', 'number', 'boolean', 'object'].includes(variable.type)) {
             throw new Error(`Unsupported variable type: ${variable.type}`);
         }
     }
-    /**
-     * Resolve variable reference, including nested properties (e.g., foo.bar.baz)
-     */
     resolveReference(reference, variables) {
-        const [varName, ...pathSegments] = reference.split(".");
+        const [varName, ...pathSegments] = reference.split('.');
         let value = variables.get(varName);
         if (value === undefined) {
             throw new Error(`Undefined variable referenced: ${varName}`);
         }
         for (const segment of pathSegments) {
-            if (value == null || typeof value !== "object") {
+            if (value == null || typeof value !== 'object') {
                 throw new Error(`Cannot resolve path '${reference}': segment '${segment}' is invalid`);
             }
             value = value[segment];
         }
         return value;
     }
-    /**
-     * Ensure we have a Gemini API key before calling Genkit helpers
-     */
-    ensureGeminiApiKey(props) {
-        const apiKey = props?.gemini_api_key?.trim();
-        if (!apiKey) {
-            throw new Error("Missing required prop: gemini_api_key. Configure a valid Gemini API key before calling this agent.");
-        }
-        return apiKey;
-    }
-    createGeminiModel(props) {
-        const apiKey = this.ensureGeminiApiKey(props);
-        const google = (0, google_1.createGoogleGenerativeAI)({ apiKey });
-        const model = google(this.GEMINI_MODEL);
-        return model;
-    }
-    createGeminiImageModel(props) {
-        const apiKey = this.ensureGeminiApiKey(props);
-        const google = (0, google_1.createGoogleGenerativeAI)({ apiKey });
-        return google.image(this.GEMINI_IMAGE_MODEL);
-    }
-    async callGeminiFlashText(params, props, context) {
-        const prompt = params.prompt?.trim();
-        if (!prompt) {
-            throw new Error("Gemini Flash text helper requires a prompt");
-        }
-        const temperature = typeof params.temperature === "number" ? params.temperature : 0.7;
-        const maxTokens = typeof params.maxTokens === "number" ? params.maxTokens : undefined;
-        const model = this.createGeminiModel(props);
-        const startTime = Date.now();
-        try {
-            context.log(`Invoking Gemini Flash text helper (temperature=${temperature}, maxTokens=${typeof maxTokens === "number" ? maxTokens : "provider-default"})`);
-            const generationOptions = {
-                model,
-                prompt,
-                temperature,
-            };
-            if (typeof maxTokens === "number") {
-                generationOptions.maxOutputTokens = maxTokens;
-            }
-            const { text, content } = await (0, ai_1.generateText)(generationOptions);
-            const contentText = content
-                ? content
-                    .map((part) => typeof part?.text === "string" ? part.text : "")
-                    .join(" ")
-                    .trim()
-                : "";
-            const finalText = (text || "").trim() || contentText || "";
-            if (!finalText) {
-                throw new Error("Gemini Flash text helper returned empty response");
-            }
-            return {
-                texto: finalText,
-                metadados: {
-                    modelo_utilizado: this.GEMINI_MODEL,
-                    tempo_processamento: Date.now() - startTime,
-                    temperature,
-                    max_tokens: typeof maxTokens === "number" ? maxTokens : null,
-                    prompt_utilizado: prompt,
-                },
-            };
-        }
-        catch (error) {
-            const message = error instanceof Error ? error.message : "Unknown Gemini Flash error";
-            context.log(`Gemini Flash text helper failed: ${message}`);
-            context.setError(message);
-            throw error;
-        }
-    }
-    async callGeminiFlashTranslation(params, props, context) {
-        const texto = params.texto?.trim();
-        const idiomaOrigem = params.idioma_origem?.trim().toLowerCase();
-        const idiomaDestino = params.idioma_destino?.trim().toLowerCase();
-        if (!texto || !idiomaOrigem || !idiomaDestino) {
-            throw new Error("Missing required translation parameters: texto, idioma_origem, idioma_destino");
-        }
-        if (idiomaOrigem === idiomaDestino) {
-            return {
-                texto_traduzido: texto,
-                metadados: {
-                    modelo_utilizado: this.GEMINI_MODEL,
-                    tempo_processamento: 0,
-                    confianca: 1,
-                    idiomas_suportados: this.SUPPORTED_TRANSLATION_LANGUAGES,
-                    idiomas_solicitados: {
-                        origem: idiomaOrigem,
-                        destino: idiomaDestino,
-                    },
-                    prompt_utilizado: "Bypass: idiomas de origem e destino são iguais",
-                },
-            };
-        }
-        const template = params.promptTemplate && params.promptTemplate.trim().length > 0
-            ? params.promptTemplate
-            : `Traduza o texto abaixo de {{idioma_origem}} para {{idioma_destino}}.
-Responda somente com o texto traduzido sem comentários adicionais.
-
-Texto:
-{{texto}}`;
-        const prompt = template
-            .replace(/{{texto}}/g, texto)
-            .replace(/{{idioma_origem}}/g, idiomaOrigem)
-            .replace(/{{idioma_destino}}/g, idiomaDestino)
-            .trim();
-        const model = this.createGeminiModel(props);
-        const startTime = Date.now();
-        try {
-            context.log(`Invoking Gemini Flash translation helper (${idiomaOrigem}->${idiomaDestino})`);
-            const { text, content } = await (0, ai_1.generateText)({
-                model,
-                prompt,
-                temperature: 0.2,
-            });
-            const contentText = content
-                ? content
-                    .map((part) => typeof part?.text === "string" ? part.text : "")
-                    .join(" ")
-                    .trim()
-                : "";
-            const translatedText = (text || "").trim() || contentText || "";
-            if (!translatedText) {
-                throw new Error("Gemini Flash translation returned empty response");
-            }
-            return {
-                texto_traduzido: translatedText,
-                metadados: {
-                    modelo_utilizado: this.GEMINI_MODEL,
-                    tempo_processamento: Date.now() - startTime,
-                    confianca: 0.85,
-                    idiomas_suportados: this.SUPPORTED_TRANSLATION_LANGUAGES,
-                    idiomas_solicitados: {
-                        origem: idiomaOrigem,
-                        destino: idiomaDestino,
-                    },
-                    prompt_utilizado: prompt,
-                },
-            };
-        }
-        catch (error) {
-            const message = error instanceof Error ? error.message : "Unknown translation error";
-            context.log(`Gemini Flash translation failed: ${message}`);
-            context.setError(message);
-            throw error;
-        }
-    }
-    async callGeminiFlashImage(params, props, context) {
-        const prompt = params.prompt?.trim();
-        const estilo = params.estilo?.trim();
-        const resolucao = params.resolucao?.trim();
-        if (!prompt) {
-            throw new Error("Gemini Flash image helper requires a prompt");
-        }
-        if (!estilo) {
-            throw new Error("Gemini Flash image helper requires an estilo value");
-        }
-        if (!resolucao || !/^\d+x\d+$/.test(resolucao)) {
-            throw new Error("Gemini Flash image helper requires uma resolução no formato LARGURAxALTURA (ex: 1024x1024)");
-        }
-        const model = this.createGeminiImageModel(props);
-        const startTime = Date.now();
-        try {
-            context.log(`Invoking Gemini Flash image helper (estilo=${estilo}, resolucao=${resolucao})`);
-            const result = await (0, ai_1.experimental_generateImage)({
-                model,
-                prompt,
-                size: resolucao,
-            });
-            const image = result.image;
-            if (!image?.base64 || !image.mediaType) {
-                throw new Error("Gemini Flash image helper returned an invalid file");
-            }
-            const normalizedBase64 = image.base64.replace(/\s+/g, "");
-            const imageUrl = `data:${image.mediaType};base64,${normalizedBase64}`;
-            return {
-                image_url: imageUrl,
-                image_base64: normalizedBase64,
-                media_type: image.mediaType,
-                prompt_utilizado: prompt,
-                metadados: {
-                    modelo_utilizado: this.GEMINI_IMAGE_MODEL,
-                    tempo_processamento: Date.now() - startTime,
-                    estilo,
-                    resolucao,
-                },
-            };
-        }
-        catch (error) {
-            const message = error instanceof Error ? error.message : "Unknown image error";
-            context.log(`Gemini Flash image helper failed: ${message}`);
-            context.setError(message);
-            throw error;
-        }
-    }
 }
 exports.DeclarativeAgentInterpreter = DeclarativeAgentInterpreter;
 // Singleton instance