🔬 Extract structured data from D&B company reports with GPT-4o

⚡ 232 views · 🔬 Document Extraction & Analysis

Description

Pull a Dun & Bradstreet Business Information Report (PDF) by DUNS, convert the response into a binary PDF file, extract readable text, and use OpenAI to return a clean, flat JSON with only the key fields you care about (e.g., report date, Paydex, viability score, credit limit). Includes Sticky Notes for quick setup help and guidance.

✅ What this template does

• Requests a D&B report (PDF) for a specific DUNS via HTTP
• Converts the API response into a binary PDF file
• Extracts the text from the PDF for analysis
• Uses OpenAI with a Structured Output Parser to return a flat JSON
• Designed to be extended to Sheets, databases, or CRMs

🧩 How it works (node-by-node)

1. Manual Trigger — Runs the workflow on demand (“When clicking ‘Execute workflow’”).
2. D&B Report (HTTP Request) — Calls the D&B Reports API for a Business Information Report (PDF).
3. Convert to PDF File (Convert to File) — Turns the D&B response payload into a binary PDF.
4. Extract Binary (Extract from File) — Extracts text content from the PDF.
5. OpenAI Chat Model — Provides the language model context for the analyzer.
6. Analyze PDF (AI Agent) — Reads the extracted text and applies strict rules for a flat JSON output.
7. Structured Output (AI Structured Output Parser) — Enforces a schema and validates/auto-fixes the JSON shape.
8. (Optional) Get Bearer Token (HTTP Request) — Template guidance for OAuth token retrieval (shown as disabled; included for reference if you prefer Bearer flows).

🛠️ Setup instructions (from the JSON)

1) D&B Report (HTTP Request)

• Auth: Header Auth (use an n8n HTTP Header Auth credential)

• URL:
  https://plus.dnb.com/v1/reports/duns/804735132?productId=birstd&inLanguage=en-US&reportFormat=PDF&orderReason=6332&tradeUp=hq&customerReference=customer%20reference%20text

• Headers:

  • Accept: application/json

• Credential Example: D&B (HTTP Header Auth)
  > Put your Authorization: Bearer <token> header inside this credential, not directly in the node.

2) Convert to PDF File (Convert to File)

• Operation: toBinary
• Source Property: contents[0].contentObject
  > This takes the PDF content from the D&B API response and converts it to a binary file for downstream nodes.

3) Extract Binary (Extract from File)

• Operation: pdf
  > Produces a text field with the extracted PDF content, ready for AI analysis.

4) OpenAI Model(s)

• OpenAI Chat Model
• Model: gpt-4o (as configured in the JSON)
• Credential: Your stored OpenAI API credential (do not hardcode keys)
• Wiring:
  • Connect OpenAI Chat Model as ai_languageModel to Analyze PDF
  • Connect another OpenAI Chat Model (also gpt-4o) as ai_languageModel to Structured Output

5) Analyze PDF (AI Agent)

• Prompt Type: define
• Text:
  ={{ $json.text }}
• System Message (rules):
  • You are a precision extractor. Read the provided business report PDF and return only a single flat JSON object with the fields below.
  • No arrays/lists.
  • No prose.
  • If a value is missing, output null.
  • Dates: YYYY-MM-DD.
  • Numbers: plain numerics (no commas or $).
  • Prefer most recent or highest-level overall values if multiple are shown.
  • Never include arrays, nested structures, or text outside of the JSON object.

6) Structured Output (AI Structured Output Parser)

• JSON Schema Example:

{
  "report_date": "",
  "company_name": "",
  "duns": "",
  "dnb_rating_overall": "",
  "composite_credit_appraisal": "",
  "viability_score": "",
  "portfolio_comparison_score": "",
  "paydex_3mo": "",
  "paydex_24mo": "",
  "credit_limit_conservative": ""
}

• Auto Fix: enabled
• Wiring: Connect as ai_outputParser to Analyze PDF

7) (Optional) Get Bearer Token (HTTP Request) — Disabled example

If you prefer fetching tokens dynamically:

• Auth: Basic Auth (D&B username/password)
• Method: POST
• URL: https://plus.dnb.com/v3/token
• Body Parameters:
  • grant_type = client_credentials
• Headers:
  • Accept: application/json
• Downstream usage: Set header Authorization: Bearer {{$json["access_token"]}} in subsequent calls.

> In this template, the D&B Report node uses Header Auth credential instead. Use one strategy consistently (credentials are recommended for security).

🧠 Output schema (flat JSON)

The analyzer + parser return a single flat object like:

{
  "report_date": "2024-12-31",
  "company_name": "Example Corp",
  "duns": "123456789",
  "dnb_rating_overall": "5A2",
  "composite_credit_appraisal": "Fair",
  "viability_score": "3",
  "portfolio_comparison_score": "2",
  "paydex_3mo": "80",
  "paydex_24mo": "78",
  "credit_limit_conservative": "25000"
}

🧪 Test flow

1. Click Execute workflow (Manual Trigger).
2. Confirm D&B Report returns the PDF response.
3. Check Convert to PDF File for a binary file.
4. Verify Extract from File produces a text field.
5. Inspect Analyze PDF → Structured Output for valid JSON.

🔐 Security notes

• Do not hardcode tokens in nodes; use Credentials (HTTP Header Auth or Basic Auth).
• Restrict who can execute the workflow if it’s accessible from outside your network.
• Avoid storing sensitive payloads in logs; mask tokens/headers.

🧩 Customize

• Map the structured JSON to Google Sheets, Postgres/BigQuery, or a CRM.
• Extend the schema with additional fields (e.g., number of employees, HQ address) — keep it flat.
• Add validation (Set/IF nodes) to ensure required fields exist before writing downstream.

🩹 Troubleshooting

• Missing PDF text? Ensure Convert to File source property is contents[0].contentObject.
• Unauthorized from D&B? Refresh/verify token; confirm Header Auth credential contains Authorization: Bearer <token>.
• Parser errors? Keep the agent output short and flat; the Structured Output node will auto-fix minor issues.
• Different DUNS/product? Update the D&B Report URL query params (duns, productId, etc.).

🗒️ Sticky Notes (included)

• Overview: “Fetch D&B Company Report (PDF) → Convert → Extract → Summarize to Structured JSON (n8n)”
• Setup snippets for Data Blocks (optional) and Auth flow

📬 Contact

Need help customizing this (e.g., routing the PDF to Drive, mapping JSON to your CRM, or expanding the schema)?

📧 robert@ynteractive.com
🔗 https://www.linkedin.com/in/robert-breen-29429625/
🌐 https://ynteractive.com

🔗 Nodes Used

HTTP Request, AI Agent, OpenAI Chat Model, Structured Output Parser, Convert to File, Extract from File

📥 Import

Download workflow.json and import into n8n: Workflow menu → Import from File

📖 Importing guide · 🔑 Credential setup