Why Does My Openrouter API Key Fail in a Cron-Triggered Workflow but Succeed in Test Mode?

I created an AI agent using Openrouter, and it works flawlessly in test mode. However, when I activate the workflow and trigger it via cron, the API key fails with the following error:

Authorization failed - please check your credentials

Incorrect API key provided: sk-or-v1************************fe11. You can find your API key at https://platform.openai.com/account/api-keys.

I suspect the issue might be related to how credentials are stored or loaded in the active workflow compared to test mode. In test mode, n8n might be using environment variables or temporary settings, while the active (cron-triggered) run uses the saved credentials, which might include extra characters or whitespace. Additionally, ensure the correct credential (openrouter - hacheng1) is attached to the OpenAI Chat Model node.

Below is the workflow in JSON format for reference:

JSONCopied
{
  "nodes": [
    {
      "parameters": {
        "databaseId": 1,
        "tableId": 3,
        "returnAll": true,
        "additionalOptions": {}
      },
      "type": "n8n-nodes-base.baserow",
      "typeVersion": 1,
      "position": [
        -140,
        -80
      ],
      "id": "a58c6468-9f43-487b-8903-aa93b7c3579a",
      "name": "Baserow",
      "credentials": {
        "baserowApi": {
          "id": "b7Y4I4TJZb33ejAF",
          "name": "Baserow account - contentsprout"
        }
      }
    },
    {
      "parameters": {
        "model": {
          "__rl": true,
          "value": "deepseek/deepseek-r1-distill-qwen-14b",
          "mode": "list",
          "cachedResultName": "deepseek/deepseek-r1-distill-qwen-14b"
        },
        "options": {}
      },
      "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
      "typeVersion": 1.2,
      "position": [
        60,
        220
      ],
      "id": "e833dd3e-641d-402f-9685-36edca838805",
      "name": "OpenAI Chat Model",
      "credentials": {
        "openAiApi": {
          "id": "lj7opvoCxbMSqKLn",
          "name": "openrouter - hacheng1"
        }
      }
    },
    {
      "parameters": {
        "operation": "update",
        "databaseId": 1,
        "tableId": 3,
        "rowId": "={{ $('Baserow').item.json.id }}",
        "fieldsUi": {
          "fieldValues": [
            {
              "fieldId": 32,
              "fieldValue": "={{ $json.score }}"
            }
          ]
        }
      },
      "type": "n8n-nodes-base.baserow",
      "typeVersion": 1,
      "position": [
        1240,
        -60
      ],
      "id": "ad7ef737-8d71-4fed-928a-225f6c602009",
      "name": "Baserow1",
      "credentials": {
        "baserowApi": {
          "id": "b7Y4I4TJZb33ejAF",
          "name": "Baserow account - contentsprout"
        }
      }
    },
    {
      "parameters": {
        "assignments": {
          "assignments": [
            {
              "id": "1c73adc0-b753-4f29-9113-9021647392a2",
              "name": "score",
              "value": "={{ $json.score }}",
              "type": "number"
            }
          ]
        },
        "options": {}
      },
      "type": "n8n-nodes-base.set",
      "typeVersion": 3.4,
      "position": [
        1020,
        -100
      ],
      "id": "1e39b435-3f6c-4257-88da-454b598caedd",
      "name": "Edit Fields"
    },
    {
      "parameters": {
        "promptType": "define",
        "text": "=You are an expert lead scorer for an AI development and automation agency. \nWe focus on companies making over $1M annually that might hire us for AI or automation solutions.\n\nAnalyze the following LinkedIn headline and assign it a lead score according to these rules:\n\nScoring Guidelines (0–10):\n\n0: Not Relevant\n- Student, intern, or trainee roles\n- Competitor companies (AI/automation agencies)\n- Clearly not an industry fit or no capacity for AI/automation investment\n\n1–3: Low Priority\n- Entry-level / individual contributor\n- Company size or role suggests < $1M revenue\n- No mention of technology, growth, or budget authority\n\n4–5: Medium Priority\n- Mid-level managers in potentially $1M+ companies\n- Some interest in efficiency or analytics\n- Not clearly a decision-maker but in a relevant industry\n\n6–7: High Priority\n- Senior leaders (Director, VP) in $1M+ revenue firms\n- Likely to adopt modern tech or shows partial AI/automation interest\n- Some budget or decision-making ability implied\n\n8–10: Perfect Leads\n- C-level, Founder, Board Member at $1M+ revenue companies\n- Clear references to AI, automation, or major digital transformations\n- Strong authority over budget and company direction\n\nOutput Requirements:\n1. Output exactly one JSON object with a single key: `score`.\n2. The value must be a float between 0 and 10, with no more than two decimal places (e.g., 8.50).\n3. Do not output arrays, code fences, extra text, or additional keys.\n4. No explanations or disclaimers—only the JSON object.\n\n**Example** of correct output:\n\n{\n“score”: 7.50\n}\n\nLinkedIn Headline to Score: {{ $json.headline }}",
        "hasOutputParser": true,
        "options": {
          "systemMessage": "You are an AI that **must** respond with exactly one JSON object.\nThis JSON object has exactly one key: \"score\".\nThe value must be a float between 0 and 10 with at most two decimals.\nDo not output anything else—no code fences, no arrays, no explanations.\n\nExample of the only acceptable output:\n{\n  \"score\": 7.50\n}"
        }
      },
      "type": "@n8n/n8n-nodes-langchain.agent",
      "typeVersion": 1.7,
      "position": [
        120,
        -140
      ],
      "id": "af518d1a-8336-4e44-bd9b-634c00db0d9a",
      "name": "AI Agent1"
    },
    {
      "parameters": {
        "jsCode": "/**\n * This Code node will:\n * 1) Remove triple-backticks or other extraneous text around the JSON.\n * 2) Extract the JSON object from the string.\n * 3) Return items with the parsed \"score\".\n */\nreturn items.map(item => {
  // 1) Retrieve the string from item.json.output
  let rawOutput = item.json.output;
  
  // 2) Remove any ```json fences or triple backticks
  //    (this also covers cases like ```json\n{...}\n```)
  rawOutput = rawOutput
    .replace(/```json/g, '')
    .replace(/```/g, '');
  
  // 3) Match the JSON portion between the first '{' and the final '}'
  //    This allows ignoring any text before/after the JSON block.
  const jsonMatch = rawOutput.match(/\{[\s\S]*\}/);
  
  // 4) If there's no valid JSON block, default to a score of 0 (or handle how you prefer)
  if (!jsonMatch) {
    return {
      json: {
        score: 0
      }
    };
  }
  
  // 5) Parse the JSON substring
  const parsed = JSON.parse(jsonMatch[0]);
  
  // 6) Return each item with just the \"score\" in the final JSON
  return {
    json: {
      score: parsed.score
    }
  };
});"
      },
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [
        640,
        -160
      ],
      "id": "c65a115b-928f-4e39-ba88-d4b2c4b08886",
      "name": "Code"
    },
    {
      "parameters": {
        "rule": {
          "interval": [
            {}
          ]
        }
      },
      "type": "n8n-nodes-base.scheduleTrigger",
      "typeVersion": 1.2,
      "position": [
        -400,
        -80
      ],
      "id": "ff83f1e8-fe02-4a6a-8805-ff3f8929d279",
      "name": "Schedule Trigger"
    }
  ],
  "connections": {
    "Baserow": {
      "main": [
        [
          {
            "node": "AI Agent1",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "OpenAI Chat Model": {
      "ai_languageModel": [
        [
          {
            "node": "AI Agent1",
            "type": "ai_languageModel",
            "index": 0
          }
        ]
      ]
    },
    "Edit Fields": {
      "main": [
        [
          {
            "node": "Baserow1",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "AI Agent1": {
      "main": [
        [
          {
            "node": "Code",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Code": {
      "main": [
        [
          {
            "node": "Edit Fields",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Schedule Trigger": {
      "main": [
        [
          {
            "node": "Baserow",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  },
  "pinData": {},
  "meta": {
    "templateCredsSetupCompleted": true,
    "instanceId": "f91b0b942f68abc36946915b0c9cffe661493bd31ec6576b820cb389080a2205"
  }
}

Any insights on why the API key works in test mode but not when the workflow is active would be greatly appreciated.