mcp-google-sheets

import { createAction, Property, } from '@activepieces/pieces-framework'; import { propsValidation, } from '@activepieces/pieces-common'; import { grokAuth } from '../common/auth'; import { createModelProperty, createTemperatureProperty, createTokenProperty, makeXaiRequest, validateResponse, parseJsonResponse, XaiResponse, ExtractionResult } from '../common/utils'; import { z } from 'zod'; interface ExtractDataField { fieldName: string; fieldType: 'string' | 'number' | 'boolean' | 'array' | 'object'; fieldDescription: string; required: boolean; } export const extractDataFromText = createAction({ auth: grokAuth, name: 'extract_data_from_text', displayName: 'Extract Data From Text', description: 'Extract structured data fields from unstructured text (e.g., names, addresses, dates).', props: { model: createModelProperty({ displayName: 'Model', description: 'Grok model to use for data extraction.', defaultValue: 'grok-3-beta' }), text: Property.LongText({ displayName: 'Text to Extract From', required: true, description: 'Text to extract data from.', }), extractionPrompt: Property.LongText({ displayName: 'Extraction Instructions', required: false, description: 'How to extract the data.', defaultValue: 'Extract the following structured data from the provided text. Be accurate and only extract information that is explicitly present. If a field is not found, use null.', }), fields: Property.Array({ displayName: 'Fields to Extract', required: true, description: 'Define the structured fields to extract from the text.', properties: { fieldName: Property.ShortText({ displayName: 'Field Name', required: true, description: 'Name of the field (e.g., "firstName", "email", "phoneNumber")', }), fieldType: Property.StaticDropdown({ displayName: 'Field Type', required: true, defaultValue: 'string', options: { disabled: false, options: [ { label: 'Text (String)', value: 'string' }, { label: 'Number', value: 'number' }, { label: 'True/False (Boolean)', value: 'boolean' }, { label: 'List (Array)', value: 'array' }, { label: 'Object', value: 'object' }, ], }, }), fieldDescription: Property.ShortText({ displayName: 'Field Description', required: true, description: 'What this field represents (e.g., "The person\'s first name")', }), required: Property.Checkbox({ displayName: 'Required Field', required: false, defaultValue: true, description: 'Must be present in extracted data.', }), }, }), enableContextSearch: Property.Checkbox({ displayName: 'Enable Context Search', required: false, defaultValue: false, description: 'Search for additional context to improve extraction.', }), strictExtraction: Property.Checkbox({ displayName: 'Strict Extraction', required: false, defaultValue: true, description: 'Only extract explicitly present information.', }), temperature: createTemperatureProperty(0.1), maxCompletionTokens: createTokenProperty(1000), reasoningEffort: Property.StaticDropdown({ displayName: 'Reasoning Effort', required: false, description: 'How thoroughly to analyze the text.', options: { disabled: false, options: [ { label: 'Default', value: '' }, { label: 'Low (Quick extraction)', value: 'low' }, { label: 'High (Deep analysis)', value: 'high' }, ], }, }), includeConfidence: Property.Checkbox({ displayName: 'Include Confidence Scores', required: false, defaultValue: true, description: 'Include confidence scores for extracted fields.', }), }, async run({ auth, propsValue }) { await propsValidation.validateZod(propsValue, { temperature: z.number().min(0).max(2).optional(), maxCompletionTokens: z.number().min(100).max(4000).optional(), text: z.string().min(1).max(100000), }); const { model, text, extractionPrompt, fields, enableContextSearch, strictExtraction, temperature, maxCompletionTokens, reasoningEffort, includeConfidence, } = propsValue; if (!text.trim()) { throw new Error('Text to extract from cannot be empty'); } const fieldsArray = fields as ExtractDataField[]; if (!fieldsArray || fieldsArray.length === 0) { throw new Error('At least one field must be defined for extraction'); } const extractedFieldsSchema: any = {}; const requiredFields: string[] = []; fieldsArray.forEach((field) => { extractedFieldsSchema[field.fieldName] = { type: field.fieldType === 'array' ? 'array' : field.fieldType, description: field.fieldDescription, ...(field.fieldType === 'array' ? { items: { type: 'string' } } : {}), }; if (field.required) { requiredFields.push(field.fieldName); } }); const jsonSchema = { type: "object", properties: { extracted_data: { type: "object", properties: extractedFieldsSchema, required: requiredFields, description: "The structured data extracted from the text", }, extraction_notes: { type: "string", description: "Notes about the extraction process, any ambiguities, or missing information", }, ...(includeConfidence ? { confidence_scores: { type: "object", properties: Object.fromEntries( fieldsArray.map(field => [ field.fieldName, { type: "number", minimum: 0, maximum: 1, description: `Confidence score for ${field.fieldName} extraction (0-1)`, } ]) ), description: "Confidence scores for each extracted field", } } : {}), extraction_success: { type: "boolean", description: "Whether the extraction was successful overall", }, }, required: ["extracted_data", "extraction_notes", "extraction_success"], additionalProperties: false, }; const fieldDescriptions = fieldsArray.map(field => `- ${field.fieldName} (${field.fieldType}${field.required ? ', required' : ', optional'}): ${field.fieldDescription}` ).join('\n'); const systemPrompt = `You are an expert data extraction system. Your task is to extract structured data fields from unstructured text with high accuracy. ${extractionPrompt} Fields to Extract: ${fieldDescriptions} Extraction Rules: - ${strictExtraction ? 'Only extract information that is explicitly present in the text. Do not infer or guess.' : 'Extract information that is present or can be reasonably inferred from context.'} - Use null for fields that cannot be found or determined - For array fields, extract all relevant items as a list - For boolean fields, use true/false based on explicit or clear implicit information - For number fields, extract numeric values (convert text numbers if needed) - ${includeConfidence ? 'Provide confidence scores (0-1) for each extracted field' : ''} - Always provide extraction notes explaining any challenges, ambiguities, or missing information - Be consistent with data types and formatting Data Quality Standards: - Ensure accuracy over completeness - Maintain consistency in formatting (e.g., phone numbers, dates) - Flag any uncertain extractions in the notes - Preserve original formatting where appropriate Response Format: You must respond with valid JSON matching the specified schema.`; const messages = [ { role: 'system', content: systemPrompt, }, { role: 'user', content: `Please extract the structured data from the following text:\n\n${text.trim()}`, }, ]; const requestBody: any = { model, messages, temperature: temperature || 0.1, max_completion_tokens: maxCompletionTokens || 1000, response_format: { type: 'json_schema', json_schema: { name: "extraction_result", schema: jsonSchema, strict: true } }, }; if (reasoningEffort && !model.includes('grok-3')) { requestBody.reasoning_effort = reasoningEffort; } if (enableContextSearch) { requestBody.search_parameters = { mode: 'auto', max_search_results: 3, return_citations: true, // Auto-enable citations when using context search }; } try { const response = await makeXaiRequest(auth, requestBody, 120000, 'Extract Data'); const { choice, content } = validateResponse(response, 'Extract Data'); const extractionResult = parseJsonResponse(content, 'Extract Data'); const extractedData = extractionResult.extracted_data || {}; const result = { extracted_data: extractedData, extraction_notes: extractionResult.extraction_notes || '', extraction_success: extractionResult.extraction_success || false, fields_extracted: Object.keys(extractedData).length, fields_requested: fieldsArray.length, completion_rate: fieldsArray.length > 0 ? Object.keys(extractedData).length / fieldsArray.length : 0, model: response.body.model, finish_reason: choice.finish_reason, }; if (includeConfidence && extractionResult.confidence_scores) { (result as any).confidence_scores = extractionResult.confidence_scores; const confidenceValues = Object.values(extractionResult.confidence_scores) as number[]; if (confidenceValues.length > 0) { (result as any).avg_confidence = confidenceValues.reduce((a, b) => a + b, 0) / confidenceValues.length; (result as any).max_confidence = Math.max(...confidenceValues); (result as any).min_confidence = Math.min(...confidenceValues); } } const requiredFieldsExtracted = requiredFields.filter(field => extractedData[field] !== null && extractedData[field] !== undefined ); (result as any).required_fields_found = requiredFieldsExtracted.length; (result as any).required_fields_missing = requiredFields.length - requiredFieldsExtracted.length; if (response.body.usage) { (result as any).usage = response.body.usage; } if (enableContextSearch && response.body.citations) { (result as any).citations = response.body.citations; } if (choice.message.reasoning_content) { (result as any).reasoning_content = choice.message.reasoning_content; } return result as ExtractionResult; } catch (error: any) { if (error.response?.status === 400) { const errorMessage = error.response?.body?.error?.message || 'Bad request'; throw new Error(`Data extraction failed: ${errorMessage}`); } if (error.response?.status === 422) { const errorMessage = error.response?.body?.error?.message || 'Validation error'; throw new Error(`Invalid extraction parameters: ${errorMessage}`); } if (error.response?.status === 429) { throw new Error('Rate limit exceeded. Please try again later.'); } if (error.response?.status === 500) { throw new Error('Extraction service temporarily unavailable. Please try again.'); } if (error.response?.status === 401) { throw new Error('Invalid API key. Please check your xAI API key.'); } if (error.response?.status === 403) { throw new Error('Access denied. Please check your API key permissions.'); } if (error.message?.includes('timeout')) { throw new Error('Data extraction timed out. Try reducing text length or simplifying field definitions.'); } throw new Error(`Data extraction failed: ${error.message || 'Unknown error occurred'}`); } }, });