{
  "flow": {
    "nodes": [
      {
        "width": 340,
        "height": 160,
        "id": "comment-schema",
        "type": "comment",
        "position": {
          "x": -420,
          "y": -80
        },
        "positionAbsolute": {
          "x": -420,
          "y": -80
        },
        "data": {
          "title": "Structured Output / Schema Determinism",
          "text": "Tests whether models produce valid JSON with no markdown fences, no trailing commas, and correct key names.\n\nSimple evaluator: response must start with '{' (no preamble).\nInspect node: verify field names and types manually.\n\nEnterprise relevance: downstream parsers (n8n, LangGraph, DBs) crash on malformed JSON."
        },
        "dragHandle": ".node-header"
      },
      {
        "width": 500,
        "height": 500,
        "id": "table-schema",
        "type": "table",
        "position": {
          "x": -420,
          "y": 120
        },
        "positionAbsolute": {
          "x": -420,
          "y": 120
        },
        "data": {
          "title": "Extraction Test Cases",
          "columns": [
            {
              "header": "schema_desc",
              "key": "schema_desc"
            },
            {
              "header": "input_text",
              "key": "input_text"
            },
            {
              "header": "required_keys",
              "key": "required_keys"
            }
          ],
          "rows": [
            {
              "__uid": "j1",
              "schema_desc": "Extract to JSON with keys: name (string), age (integer), skills (array of strings)",
              "input_text": "Jane is 28 years old and knows Python, SQL, and Tableau.",
              "required_keys": "name,age,skills"
            },
            {
              "__uid": "j2",
              "schema_desc": "Extract to JSON with keys: company (string), founded (integer), products (array)",
              "input_text": "Apple was founded in 1976 and makes the iPhone, iPad, and MacBook.",
              "required_keys": "company,founded,products"
            },
            {
              "__uid": "j3",
              "schema_desc": "Extract to JSON with keys: sentiment (positive/negative/neutral), confidence (0.0-1.0), keywords (array, max 3 items)",
              "input_text": "The product launch exceeded expectations \u2014 sales were up 40% and customer reviews are outstanding.",
              "required_keys": "sentiment,confidence,keywords"
            },
            {
              "__uid": "j4",
              "schema_desc": "Extract to JSON with keys: event (string), date (string), location (string), attendees (integer)",
              "input_text": "The annual AI summit will take place on March 15 at the Amsterdam RAI convention centre with an expected 2400 attendees.",
              "required_keys": "event,date,location,attendees"
            },
            {
              "__uid": "j5",
              "schema_desc": "Extract to JSON with keys: repo (string), language (string), stars (integer), open_issues (integer), license (string)",
              "input_text": "The pytorch/pytorch repository is written in Python, has 84000 stars, 14500 open issues, and uses the BSD-3 license.",
              "required_keys": "repo,language,stars,open_issues,license"
            }
          ]
        }
      },
      {
        "width": 380,
        "height": 400,
        "id": "prompt-schema",
        "type": "prompt",
        "position": {
          "x": 140,
          "y": 0
        },
        "positionAbsolute": {
          "x": 140,
          "y": 0
        },
        "data": {
          "prompt": "{schema_desc}\n\nText:\n{input_text}\n\nRespond with ONLY a valid JSON object. No markdown fences. No explanation. No text outside the JSON.",
          "n": 1,
          "llms": [
            {
              "key": "schema-hermes3",
              "name": "hermes3",
              "emoji": "H",
              "model": "hermes3:latest",
              "base_model": "openai",
              "temp": 0.0,
              "settings": {
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": {
                  "type": "text"
                },
                "tools": [],
                "stop": [],
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              },
              "formData": {
                "shortname": "hermes3",
                "model": "hermes3:latest",
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": "text",
                "tools": "",
                "stop": "",
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              }
            },
            {
              "key": "schema-phi4",
              "name": "phi4",
              "emoji": "P",
              "model": "phi4:latest",
              "base_model": "openai",
              "temp": 0.0,
              "settings": {
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": {
                  "type": "text"
                },
                "tools": [],
                "stop": [],
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              },
              "formData": {
                "shortname": "phi4",
                "model": "phi4:latest",
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": "text",
                "tools": "",
                "stop": "",
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              }
            },
            {
              "key": "schema-llama",
              "name": "llama3.1:8b",
              "emoji": "L",
              "model": "llama3.1:8b",
              "base_model": "openai",
              "temp": 0.0,
              "settings": {
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": {
                  "type": "text"
                },
                "tools": [],
                "stop": [],
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              },
              "formData": {
                "shortname": "llama3.1:8b",
                "model": "llama3.1:8b",
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": "text",
                "tools": "",
                "stop": "",
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              }
            },
            {
              "key": "schema-qwen3",
              "name": "qwen3:14b",
              "emoji": "Q",
              "model": "qwen3:14b",
              "base_model": "openai",
              "temp": 0.0,
              "settings": {
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": {
                  "type": "text"
                },
                "tools": [],
                "stop": [],
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              },
              "formData": {
                "shortname": "qwen3:14b",
                "model": "qwen3:14b",
                "system_msg": "Respond with ONLY a valid JSON object. No markdown fences, no explanation, no preamble. Output must be parseable by json.loads().",
                "temperature": 0.0,
                "base_url": "http://localhost:11434/v1",
                "api_key": "ollama",
                "response_format": "text",
                "tools": "",
                "stop": "",
                "top_p": 1,
                "presence_penalty": 0,
                "frequency_penalty": 0
              }
            }
          ]
        }
      },
      {
        "width": 387,
        "height": 220,
        "id": "eval-json-start",
        "type": "simpleval",
        "position": {
          "x": 580,
          "y": 0
        },
        "positionAbsolute": {
          "x": 580,
          "y": 0
        },
        "data": {
          "title": "PASS: response starts with { (no preamble)",
          "availableVars": [
            "schema_desc",
            "input_text",
            "required_keys"
          ],
          "input": "prompt-schema",
          "operation": "starts with",
          "textValue": "{",
          "refresh": false
        }
      },
      {
        "width": 580,
        "height": 400,
        "id": "inspect-schema",
        "type": "inspect",
        "position": {
          "x": 580,
          "y": 260
        },
        "positionAbsolute": {
          "x": 580,
          "y": 260
        },
        "data": {
          "input": "prompt-schema"
        }
      },
      {
        "width": 400,
        "height": 300,
        "id": "vis-schema",
        "type": "vis",
        "position": {
          "x": 1020,
          "y": 0
        },
        "positionAbsolute": {
          "x": 1020,
          "y": 0
        },
        "data": {}
      }
    ],
    "edges": [
      {
        "id": "e-table-schema-prompt-schema_desc",
        "source": "table-schema",
        "sourceHandle": "schema_desc",
        "target": "prompt-schema",
        "targetHandle": "schema_desc",
        "interactionWidth": 100,
        "markerEnd": {
          "height": "22px",
          "type": "arrow",
          "width": "22px"
        }
      },
      {
        "id": "e-table-schema-prompt-input_text",
        "source": "table-schema",
        "sourceHandle": "input_text",
        "target": "prompt-schema",
        "targetHandle": "input_text",
        "interactionWidth": 100,
        "markerEnd": {
          "height": "22px",
          "type": "arrow",
          "width": "22px"
        }
      },
      {
        "id": "e-prompt-schema-eval",
        "source": "prompt-schema",
        "sourceHandle": "prompt",
        "target": "eval-json-start",
        "targetHandle": "responseBatch",
        "interactionWidth": 100,
        "markerEnd": {
          "height": "22px",
          "type": "arrow",
          "width": "22px"
        }
      },
      {
        "id": "e-prompt-schema-inspect",
        "source": "prompt-schema",
        "sourceHandle": "prompt",
        "target": "inspect-schema",
        "targetHandle": "input",
        "interactionWidth": 100,
        "markerEnd": {
          "height": "22px",
          "type": "arrow",
          "width": "22px"
        }
      },
      {
        "id": "e-eval-schema-vis",
        "source": "eval-json-start",
        "sourceHandle": "output",
        "target": "vis-schema",
        "targetHandle": "input",
        "interactionWidth": 100,
        "markerEnd": {
          "height": "22px",
          "type": "arrow",
          "width": "22px"
        }
      }
    ],
    "viewport": {
      "x": 0,
      "y": 0,
      "zoom": 0.75
    }
  }
}