{ "edges": [ { "to": "1174cddb-9bf5-4582-a7ba-aed412336b7f", "from": "f89f5698-5d3c-4771-8b0b-511b32b9fc33", "toPort": "0", "fromPort": "0" }, { "to": "f89f5698-5d3c-4771-8b0b-511b32b9fc33", "from": "0612aeb7-a3a2-4215-9910-4cd077e06586", "toPort": "0", "fromPort": "0" }, { "to": "48988700-46e2-4464-87a5-99054d6e9cbc", "from": "0612aeb7-a3a2-4215-9910-4cd077e06586", "toPort": "0", "fromPort": "1" }, { "to": "6517c215-19c5-4644-97f6-26d650c65540", "from": "69339a87-95fa-4827-a768-a6a9aa1def9e", "toPort": "0", "fromPort": "0" }, { "to": "0612aeb7-a3a2-4215-9910-4cd077e06586", "from": "6517c215-19c5-4644-97f6-26d650c65540", "toPort": "0", "fromPort": "0" }, { "to": "49b80e52-27d3-455d-8082-e3e0f2365a69", "from": "de2e5d93-df58-41ef-a22f-e7a681308a64", "toPort": "0", "fromPort": "0" }, { "to": "6517c215-19c5-4644-97f6-26d650c65540", "from": "99814f2b-ab97-49bf-8eec-43083f731dad", "toPort": "0", "fromPort": "ai_languageModel" }, { "to": "40df551b-4ca7-4268-a1fe-4ebb01fd304f", "from": "1174cddb-9bf5-4582-a7ba-aed412336b7f", "toPort": "0", "fromPort": "0" }, { "to": "5d940fc3-8727-4dc4-9a58-133ab5180f08", "from": "1174cddb-9bf5-4582-a7ba-aed412336b7f", "toPort": "0", "fromPort": "0" }, { "to": "de2e5d93-df58-41ef-a22f-e7a681308a64", "from": "1174cddb-9bf5-4582-a7ba-aed412336b7f", "toPort": "0", "fromPort": "0" }, { "to": "69339a87-95fa-4827-a768-a6a9aa1def9e", "from": "9e228e13-31c4-4f40-8bc1-83ffc0c0df21", "toPort": "0", "fromPort": "0" } ], "nodes": [ { "id": "9e228e13-31c4-4f40-8bc1-83ffc0c0df21", "name": "When clicking ‘Execute workflow’", "type": "manualTrigger", "params": {}, "position": [ -176, 16 ] }, { "id": "69339a87-95fa-4827-a768-a6a9aa1def9e", "name": "Fetch Raw Feedback Data", "type": "integration", "params": { "options": {}, "service": "googleSheets", "operation": "read", "sheetName": { "__rl": true, "mode": "list", "value": 315277036, "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit#gid=315277036", "cachedResultName": "Raw_Feedback" }, "documentId": { "__rl": true, "mode": "list", "value": "1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit?usp=drivesdk", "cachedResultName": "Interviewer Brief Pack " } }, "position": [ 32, 16 ] }, { "id": "99814f2b-ab97-49bf-8eec-43083f731dad", "name": "AI Quality Evaluator (GPT-4o1", "type": "llm", "params": { "model": "gpt-4o-mini", "options": {}, "provider": "openai" }, "position": [ 272, 192 ] }, { "id": "6517c215-19c5-4644-97f6-26d650c65540", "name": "Analyze Feedback Quality", "type": "llm", "params": { "text": "=You are an Interview Feedback Quality Auditor.\n\nYour task is to evaluate interviewer feedback notes and score them across 5 dimensions:\n- specificity (1–5)\n- structure_STAR (1–5)\n- bias_free_language (1–5)\n- actionability (1–5)\n- depth (1–5)\n\n⚖️ Scoring Guidelines:\n- 5 = Excellent: Clear, detailed, STAR format (Situation, Task, Action, Result) explicitly used or strongly implied, with evidence/examples.\n- 4 = Good: Mostly structured, some detail, minor gaps, still useful for decisions.\n- 3 = Adequate: Some relevant info but mixed with vagueness, missing STAR elements.\n- 2 = Poor: Mostly vague or generic, no clear evidence, over-reliant on subjective phrasing.\n- 1 = Unusable: Purely subjective (“great guy”, “nice energy”), no actionable details.\n\nBias-free language: Score low if feedback references gender, looks, personality, or irrelevant traits.\n\nActionability: Score higher if the feedback directly helps in making a decision (e.g., “passed all test cases under time constraint” vs “seems smart”).\n\nDepth: Score higher if multiple competencies or dimensions are covered, lower if only 1 vague point.\n\n🚨 Additional Rules:\n- If text <30 words OR contains mostly emojis/placeholders → set ALL scores ≤2 and add `\"too_short\"` to vague_phrases.\n- Extract vague phrases (e.g., “good energy”, “smart guy”, “should be fine”) into `\"vague_phrases\"` array.\n\nReturn ONLY valid JSON in this schema:\n{\n \"specificity\": <1–5>,\n \"structure\": <1–5>,\n \"bias_free_language\": <1–5>,\n \"actionability\": <1–5>,\n \"depth\": <1–5>,\n \"vague_phrases\": [ ... ]\n}\n", "prompt": "=You are an Interview Feedback Quality Auditor.\n\nYour task is to evaluate interviewer feedback notes and score them across 5 dimensions:\n- specificity (1–5)\n- structure_STAR (1–5)\n- bias_free_language (1–5)\n- actionability (1–5)\n- depth (1–5)\n\n⚖️ Scoring Guidelines:\n- 5 = Excellent: Clear, detailed, STAR format (Situation, Task, Action, Result) explicitly used or strongly implied, with evidence/examples.\n- 4 = Good: Mostly structured, some detail, minor gaps, still useful for decisions.\n- 3 = Adequate: Some relevant info but mixed with vagueness, missing STAR elements.\n- 2 = Poor: Mostly vague or generic, no clear evidence, over-reliant on subjective phrasing.\n- 1 = Unusable: Purely subjective (“great guy”, “nice energy”), no actionable details.\n\nBias-free language: Score low if feedback references gender, looks, personality, or irrelevant traits.\n\nActionability: Score higher if the feedback directly helps in making a decision (e.g., “passed all test cases under time constraint” vs “seems smart”).\n\nDepth: Score higher if multiple competencies or dimensions are covered, lower if only 1 vague point.\n\n🚨 Additional Rules:\n- If text <30 words OR contains mostly emojis/placeholders → set ALL scores ≤2 and add `\"too_short\"` to vague_phrases.\n- Extract vague phrases (e.g., “good energy”, “smart guy”, “should be fine”) into `\"vague_phrases\"` array.\n\nReturn ONLY valid JSON in this schema:\n{\n \"specificity\": <1–5>,\n \"structure\": <1–5>,\n \"bias_free_language\": <1–5>,\n \"actionability\": <1–5>,\n \"depth\": <1–5>,\n \"vague_phrases\": [ ... ]\n}\n", "batching": {}, "messages": { "messageValues": [ { "message": "You are an Interview Feedback Quality Auditor. Evaluate interview feedback for specificity, structure (STAR), bias-free language, actionability, and depth. Be strict but fair. Always return **only valid JSON** that follows the given schema.Return ONLY valid JSON, no explanations, no markdown, no quotes wrapping the whole object.\n" }, { "type": "HumanMessagePromptTemplate", "message": "=Role: {{$json[\"Role\"]}} Stage: {{$json[\"Stage\"]}} Feedback: {{$json[\"Feedback_Text\"]}}" } ] }, "provider": "openai", "promptType": "define", "approximated": true }, "position": [ 288, 16 ] }, { "id": "0612aeb7-a3a2-4215-9910-4cd077e06586", "name": "Validate AI Response", "type": "if", "params": { "options": {}, "conditions": { "options": { "version": 2, "leftValue": "", "caseSensitive": true, "typeValidation": "strict" }, "combinator": "and", "conditions": [ { "id": "4901c65c-6aaf-4efe-a133-1fedfedc0bca", "operator": { "type": "string", "operation": "notEquals" }, "leftValue": "={{ $json.text }}", "rightValue": "undefined " } ] } }, "position": [ 640, 16 ] }, { "id": "f89f5698-5d3c-4771-8b0b-511b32b9fc33", "name": "Parse AI JSON Output", "type": "code", "params": { "jsCode": "// OpenAI output comes as string in $json.text\nconst raw = $json[\"text\"];\n\n// Parse safely\nlet parsed;\ntry {\n parsed = JSON.parse(raw);\n} catch (e) {\n throw new Error(\"Invalid JSON returned by OpenAI: \" + raw);\n}\n\nreturn parsed;\n" }, "position": [ 1024, 0 ] }, { "id": "1174cddb-9bf5-4582-a7ba-aed412336b7f", "name": "Calculate Weighted Quality Score", "type": "code", "params": { "jsCode": "// Input = parsed JSON from AI\nconst data = $json;\n\n// Weights (can be adjusted or moved to Config sheet later)\nconst weights = {\n specificity: 0.35,\n structure: 0.15,\n bias_free_language: 0.15,\n actionability: 0.10,\n depth: 0.25,\n};\n\n// ✅ Fallback for structure (AI might send structure or structure_star)\nconst structureValue = data.structure ?? data.structure_star ?? 0;\n\n// Calculate weighted score safely\nlet total = (\n (data.specificity * weights.specificity) +\n (structureValue * weights.structure) +\n (data.bias_free_language * weights.bias_free_language) +\n (data.actionability * weights.actionability) +\n (data.depth * weights.depth)\n) / (\n weights.specificity +\n weights.structure +\n weights.bias_free_language +\n weights.actionability +\n weights.depth\n);\n\n// Scale to 0–100\ntotal = Math.round(total * 20);\n\n// Flags\nconst flags = [];\nif ((data.specificity ?? 0) < 3 || (data.depth ?? 0) < 3) {\n flags.push(\"low_detail\");\n}\nif ((data.bias_free_language ?? 0) < 3) {\n flags.push(\"bias\");\n}\n\n// Format vague phrases if they exist\nlet vagueFormatted = \"\";\nif (Array.isArray(data.vague_phrases) && data.vague_phrases.length > 0) {\n vagueFormatted = data.vague_phrases.map(p => `• ${p}`).join(\"\\n\");\n}\n\n// Return clean JSON\nreturn {\n json: {\n Score: total,\n Flags: flags.join(\", \"),\n LLM_JSON: JSON.stringify(data),\n VaguePhrasesFormatted: vagueFormatted, // for Slack message\n row_number: $json.row_number, // keep tracking the row\n Role: $item(0).$node[\"Fetch Raw Feedback Data\"].json.Role,\n Stage: $item(0).$node[\"Fetch Raw Feedback Data\"].json.Stage\n }\n};\n" }, "position": [ 1232, 0 ] }, { "id": "5d940fc3-8727-4dc4-9a58-133ab5180f08", "name": "Save Scores to Spreadsheet", "type": "integration", "params": { "columns": { "value": { "Flags": "={{ $json.Flags }}", "Score": "={{ $json.Score }}", "LLM_JSON": "={{ $json.LLM_JSON }}", "row_number": "={{ $('Fetch Raw Feedback Data').item.json.row_number }}" }, "schema": [ { "id": "Timestamp", "type": "string", "display": true, "required": false, "displayName": "Timestamp", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Candidate_ID", "type": "string", "display": true, "required": false, "displayName": "Candidate_ID", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Role", "type": "string", "display": true, "required": false, "displayName": "Role", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Stage", "type": "string", "display": true, "required": false, "displayName": "Stage", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Interviewer_Email", "type": "string", "display": true, "required": false, "displayName": "Interviewer_Email", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Feedback_Text", "type": "string", "display": true, "required": false, "displayName": "Feedback_Text", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Score", "type": "string", "display": true, "required": false, "displayName": "Score", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "Flags", "type": "string", "display": true, "required": false, "displayName": "Flags", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "LLM_JSON", "type": "string", "display": true, "required": false, "displayName": "LLM_JSON", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "row_number", "type": "number", "display": true, "removed": false, "readOnly": true, "required": false, "displayName": "row_number", "defaultMatch": false, "canBeUsedToMatch": true } ], "mappingMode": "defineBelow", "matchingColumns": [ "row_number" ], "attemptToConvertTypes": false, "convertFieldsToString": false }, "options": {}, "service": "googleSheets", "operation": "update", "sheetName": { "__rl": true, "mode": "list", "value": 315277036, "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit#gid=315277036", "cachedResultName": "Raw_Feedback" }, "documentId": { "__rl": true, "mode": "list", "value": "1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit?usp=drivesdk", "cachedResultName": "Interviewer Brief Pack " } }, "position": [ 1600, -208 ] }, { "id": "40df551b-4ca7-4268-a1fe-4ebb01fd304f", "name": "Send Feedback Summary to Interviewer", "type": "integration", "params": { "text": "=:mag: *Interview Feedback Audit*\n\n*Role:* {{ $json[\"Role\"] }}\n*Stage:* {{ $json[\"Stage\"] }}\n\n:bar_chart: *Score:* {{ $json[\"Score\"] }}/100 \n:warning: *Flags:* {{ $json[\"Flags\"] || \"none\" }}\n\n{{ $json[\"VaguePhrasesFormatted\"] ? \n (\"_We noticed vague or incomplete feedback. Examples:_\\n\" + $json[\"VaguePhrasesFormatted\"] + \n \"\\n\\n_To improve: try being more specific and evidence-based (e.g., STAR method)._\") \n : \n \"_✅ Great job! Your feedback was specific, structured, and bias-free._\" \n}}\n\nKeep it up — your detailed notes help us make fairer hiring decisions 🚀\n\n_Automated with this n8n workflow_\n", "user": { "__rl": true, "mode": "list", "value": "U09HMPVD466", "cachedResultName": "newscctv22" }, "select": "user", "service": "slack", "operation": "read", "otherOptions": {} }, "position": [ 1616, 0 ] }, { "id": "de2e5d93-df58-41ef-a22f-e7a681308a64", "name": "Check if Training Needed", "type": "if", "params": { "options": {}, "conditions": { "options": { "version": 2, "leftValue": "", "caseSensitive": true, "typeValidation": "loose" }, "combinator": "and", "conditions": [ { "id": "62f94225-d1b6-42a0-a3e9-7afceb9b937d", "operator": { "type": "number", "operation": "lt" }, "leftValue": "={{$json[\"Score\"]}}", "rightValue": 50 } ] }, "looseTypeValidation": true }, "position": [ 1776, 288 ] }, { "id": "49b80e52-27d3-455d-8082-e3e0f2365a69", "name": "Send Training Recommendations", "type": "integration", "params": { "text": "=:books: *Training Recommendation*\n\nYour interview feedback for **{{$json[\"Role\"]}} ({{$json[\"Stage\"]}})** was reviewed.\n\n📊 **Score:** {{$json[\"Score\"]}}/100 \n⚠️ **Flags:** {{$json[\"Flags\"] || \"none\"}} \n\nWe noticed vague or incomplete feedback. Here are some examples: \n{{ $json.VaguePhrasesFormatted }}\n\nTo improve: try using structured, evidence-based feedback (e.g., STAR method). \n\n👉 Helpful resources: \n📘 [STAR Method Guide](https://example.com/star-training) \n🎥 [Bias-Free Interviewing Video](https://example.com/interview-bias) \n\nYour detailed notes help us make fairer hiring decisions 🚀\n", "user": { "__rl": true, "mode": "list", "value": "U09HMPVD466", "cachedResultName": "newscctv22" }, "select": "user", "service": "slack", "operation": "read", "otherOptions": {} }, "position": [ 2064, 272 ] }, { "id": "48988700-46e2-4464-87a5-99054d6e9cbc", "name": "Log AI Errors", "type": "integration", "params": { "columns": { "value": {}, "schema": [], "mappingMode": "defineBelow", "matchingColumns": [], "attemptToConvertTypes": false, "convertFieldsToString": false }, "options": {}, "service": "googleSheets", "operation": "append", "sheetName": { "__rl": true, "mode": "list", "value": 1338537721, "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit#gid=1338537721", "cachedResultName": "error log sheet" }, "documentId": { "__rl": true, "mode": "list", "value": "1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1Uldk_4BxWbdZTDZxFUeohIfeBmGHHqVEl9Ogb0l6R8Y/edit?usp=drivesdk", "cachedResultName": "Interviewer Brief Pack " } }, "position": [ 944, 304 ] } ] }

Audit interview feedback & report via Slack with GPT-4o-mini and Google Sheets