Credit score NLP with ollama

Anand Nerurkar
6 days ago
5 min read

We will take a one walk through where cusotmer will use copilot to query info on structural data set like "what is my credit score??". then ollama model will respond in NLP way.

Objective

-----

👉 This is a complete, compilable, enterprise-style Spring Boot GenAI POC with:

Prompt stored as code in Git
Prompt registry
Fake microservice API
Prompt injection
Local Ollama call
Swagger UI testing
Model + prompt version returned in API

Assumption:

1.We have contextual API that will return credit score based on customer id. Now this is hardcoded for POC , in real enterprise , it will give a call for cibil api and get score.

Java 17, mvn installed locally,

ollama is intalled locally

how to install ollamalocally and start server

ChatGPT said:

Perfect — I’ll give you clear, OS-specific steps to install Ollama locally and run the LLM server, especially for Windows, since your error shows you’re on Windows.

✅ 1. INSTALL OLLAMA LOCALLY (WINDOWS)

🔹 Step 1: Download Installer

Go to:

https://ollama.com

Click:

Download for Windows

This downloads:

OllamaSetup.exe

🔹 Step 2: Install

Double-click OllamaSetup.exeFollow normal Next → Next → Install

After install, RESTART YOUR PC (important)

✅ 2. VERIFY OLLAMA IS INSTALLED

Open Command Prompt (CMD) — NOT PowerShell first:

ollama --version

✅ If installed correctly, you should see:

ollama version x.x.x

❌ If you still get:

'ollama' is not recognized

Then Ollama is NOT added to PATH → I’ll show how to fix that below.

✅ 3. START OLLAMA SERVER

Ollama automatically runs as a background service on Windows.

To verify server is running:

curl http://localhost:11434/api/tags

✅ If running, you’ll get JSON with available models❌ If not running, restart Windows OR run:

ollama serve

(keep this window open)

✅ 4. DOWNLOAD A MODEL (IMPORTANT)

Since you selected Qwen, pull a LOCAL model (cloud one won’t work offline):

✅ Recommended LOCAL MODELS

Use ONE of these:

ollama pull qwen2.5:7b

ollama pull llama3:8b

OR (low RAM)

ollama pull phi3:mini

This downloads the model to your machine.

✅ 5. RUN MODEL ONCE (WARMUP)

ollama run qwen2.5:7b

Type:

Hello

If it responds, ✔️ your LLM is working.

Exit with:

/bye

✅ 6. CONFIGURE YOUR SPRING BOOT APP

Update your file:

📄 application.yml

ollama:
  url: http://localhost:11434/api/generate
  model: qwen2.5:7b

⚠️ Do NOT use:

qwen3-v1-235b-cloud

That is cloud-only and will fail locally.

✅ 7. TEST OLLAMA API DIRECTLY (IMPORTANT)

Before running your Spring app, test:

curl http://localhost:11434/api/generate ^
-d "{\"model\":\"qwen2.5:7b\",\"prompt\":\"What is my credit score?\",\"stream\":false}"

✅ You should receive JSON with:

"response": "..."

If this works → your Spring app WILL WORK.

3.codebase: https://github.com/anerurkar/chatgpt-creditscore-nlp-llm/tree/main

genai-credit-poc/

│

├── pom.xml

│

└── src/

└── main/

├── java/

│ └── com/

│ └── digitallending/

│ │

│ ├── Application.java

│ │

│ ├── controller/

│ │ ├── ChatController.java

│ │ └── FakeScoreApiController.java

│ │

│ ├── model/

│ │ ├── ChatRequest.java

│ │ ├── ChatResponse.java

│ │ └── CustomerScore.java

│ │

│ ├── prompt/

│ │ ├── PromptBuilder.java

│ │ └── PromptRegistry.java

│ │

│ └── service/

│ ├── ChatOrchestratorService.java

│ ├── CreditScoreService.java

│ └── OllamaClient.java

│

└── resources/

├── application.yml

└── prompts/

└── credit-score-nlp-v1.0.prompt

Swagger-ui http://localhost:8080/swagger-ui/index.html

Architecture Flow

===

User (Swagger UI)

ChatController

ChatOrchestratorService

+--> CreditScoreService → FakeScoreApi

+--> PromptRegistry → PromptBuilder (Prompt as Code)

+--> OllamaClient → Local LLM Server

GenAI NLP Response

Folder Fiels

====

Project & Config
- pom.xml
- src/main/resources/application.yml
- src/main/resources/prompts/credit-score-nlp-v1.0.prompt
Main App
- Application.java
Controllers
- ChatController.java (Swagger entry point)
- FakeScoreApiController.java (fake internal score API)
Models
Prompt as Code
- PromptRegistry.java
- PromptBuilder.java
Services

✅ Mapping of Files → Folders (So You Don’t Misplace Anything)

File	Exact Location
pom.xml	/genai-credit-poc/pom.xml
Application.java	/src/main/java/com/digitallending/Application.java
ChatController.java	/src/main/java/com/digitallending/controller/ChatController.java
FakeScoreApiController.java	/src/main/java/com/digitallending/controller/FakeScoreApiController.java
CustomerScore.java	/src/main/java/com/digitallending/model/CustomerScore.java
ChatRequest.java	/src/main/java/com/digitallending/model/ChatRequest.java
ChatResponse.java	/src/main/java/com/digitallending/model/ChatResponse.java
PromptRegistry.java	/src/main/java/com/digitallending/prompt/PromptRegistry.java
PromptBuilder.java	/src/main/java/com/digitallending/prompt/PromptBuilder.java
CreditScoreService.java	/src/main/java/com/digitallending/service/CreditScoreService.java
OllamaClient.java	/src/main/java/com/digitallending/service/OllamaClient.java
ChatOrchestratorService.java	/src/main/java/com/digitallending/service/ChatOrchestratorService.java
application.yml	/src/main/resources/application.yml
credit-score-nlp-v1.0.prompt	/src/main/resources/prompts/credit-score-nlp-v1.0.prompt

✅ ROOT

📄 genai-credit-poc/pom.xml

<project xmlns="http://maven.apache.org/POM/4.0.0">
  <modelVersion>4.0.0</modelVersion>

  <groupId>com.digitallending</groupId>
  <artifactId>genai-credit-poc</artifactId>
  <version>1.0.0</version>

  <parent>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-parent</artifactId>
    <version>3.2.2</version>
  </parent>

  <properties>
    <java.version>17</java.version>
  </properties>

  <dependencies>
    <!-- Spring Web -->
    <dependency>
      <groupId>org.springframework.boot</groupId>
      <artifactId>spring-boot-starter-web</artifactId>
    </dependency>

    <!-- Swagger UI -->
    <dependency>
      <groupId>org.springdoc</groupId>
      <artifactId>springdoc-openapi-starter-webmvc-ui</artifactId>
      <version>2.3.0</version>
    </dependency>

    <!-- JSON -->
    <dependency>
      <groupId>com.fasterxml.jackson.core</groupId>
      <artifactId>jackson-databind</artifactId>
    </dependency>
  </dependencies>
</project>

✅ src/main/resources

📄 src/main/resources/application.yml

server:
  port: 8080

ollama:
  url: http://localhost:11434/api/generate
  model: qwen3-v1-235b-cloud

📄 src/main/resources/prompts/credit-score-nlp-v1.0.prompt

You are a professional digital banking assistant.

Convert the following structured financial data into a polite,
customer-friendly explanation.

DATA:
Customer Id: {{customerId}}
Credit Score: {{creditScore}}
Fraud Score: {{fraudScore}}

Rules:
- Do NOT change numeric values
- Use BFSI compliant language
- Do NOT provide financial advice
- Keep response under 3 sentences

✅ src/main/java/com/digitallending

📄 Application.java

package com.digitallending;

import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;

@SpringBootApplication
public class Application {
    public static void main(String[] args) {
        SpringApplication.run(Application.class, args);
    }
}

✅ controller

📄 src/main/java/com/digitallending/controller/ChatController.java

package com.digitallending.controller;

import com.digitallending.model.ChatRequest;
import com.digitallending.model.ChatResponse;
import com.digitallending.service.ChatOrchestratorService;
import org.springframework.web.bind.annotation.*;

@RestController
@RequestMapping("/api/chat")
public class ChatController {

    private final ChatOrchestratorService orchestrator;

    public ChatController(ChatOrchestratorService orchestrator) {
        this.orchestrator = orchestrator;
    }

    @PostMapping("/message")
    public ChatResponse chat(@RequestHeader("X-Customer-Id") String customerId,
                             @RequestBody ChatRequest request) {
        return orchestrator.process(customerId, request.getMessage());
    }
}

📄 src/main/java/com/digitallending/controller/FakeScoreApiController.java

package com.digitallending.controller;

import com.digitallending.model.CustomerScore;
import org.springframework.web.bind.annotation.*;

import java.util.Map;

@RestController
@RequestMapping("/api/internal/scores")
public class FakeScoreApiController {

    private static final Map<String, CustomerScore> DATA = Map.of(
        "CUST1001", new CustomerScore("CUST1001", 720, 11),
        "CUST1002", new CustomerScore("CUST1002", 650, 28),
        "CUST1003", new CustomerScore("CUST1003", 540, 65)
    );

    @GetMapping("/{customerId}")
    public CustomerScore getScore(@PathVariable String customerId) {
        return DATA.get(customerId);
    }
}

✅ model

📄 src/main/java/com/digitallending/model/CustomerScore.java

package com.digitallending.model;

public class CustomerScore {
    private String customerId;
    private int creditScore;
    private int fraudScore;

    public CustomerScore() {}

    public CustomerScore(String customerId, int creditScore, int fraudScore) {
        this.customerId = customerId;
        this.creditScore = creditScore;
        this.fraudScore = fraudScore;
    }

    public String getCustomerId() { return customerId; }
    public int getCreditScore() { return creditScore; }
    public int getFraudScore() { return fraudScore; }
}

📄 src/main/java/com/digitallending/model/ChatRequest.java

package com.digitallending.model;

public class ChatRequest {
    private String message;

    public String getMessage() { return message; }
    public void setMessage(String message) { this.message = message; }
}

📄 src/main/java/com/digitallending/model/ChatResponse.java

package com.digitallending.model;

public class ChatResponse {
    private String reply;
    private String promptVersion;
    private String model;

    public ChatResponse(String reply, String promptVersion, String model) {
        this.reply = reply;
        this.promptVersion = promptVersion;
        this.model = model;
    }

    public String getReply() { return reply; }
    public String getPromptVersion() { return promptVersion; }
    public String getModel() { return model; }
}

✅ prompt

📄 src/main/java/com/digitallending/prompt/PromptRegistry.java

package com.digitallending.prompt;

import jakarta.annotation.PostConstruct;
import org.springframework.core.io.ClassPathResource;
import org.springframework.stereotype.Component;

import java.nio.file.Files;
import java.util.HashMap;
import java.util.Map;

@Component
public class PromptRegistry {

    private final Map<String, String> registry = new HashMap<>();

    @PostConstruct
    public void loadPrompts() {
        try {
            ClassPathResource resource =
                new ClassPathResource("prompts/credit-score-nlp-v1.0.prompt");
            String content = Files.readString(resource.getFile().toPath());
            registry.put("credit-score-nlp-v1.0", content);
        } catch (Exception e) {
            throw new RuntimeException("Failed to load prompts", e);
        }
    }

    public String getPrompt(String key) {
        return registry.get(key);
    }
}

📄 src/main/java/com/digitallending/prompt/PromptBuilder.java

package com.digitallending.prompt;

import com.digitallending.model.CustomerScore;
import org.springframework.stereotype.Component;

@Component
public class PromptBuilder {

    private final PromptRegistry registry;

    public PromptBuilder(PromptRegistry registry) {
        this.registry = registry;
    }

    public String build(CustomerScore score) {
        String template = registry.getPrompt("credit-score-nlp-v1.0");

        return template.replace("{{customerId}}", score.getCustomerId())
            .replace("{{creditScore}}", String.valueOf(score.getCreditScore()))
            .replace("{{fraudScore}}", String.valueOf(score.getFraudScore()));
    }

    public String version() {
        return "credit-score-nlp-v1.0";
    }
}

✅ service

📄 src/main/java/com/digitallending/service/CreditScoreService.java

package com.digitallending.service;

import com.digitallending.model.CustomerScore;
import org.springframework.stereotype.Service;
import org.springframework.web.client.RestTemplate;

@Service
public class CreditScoreService {

    private final RestTemplate restTemplate = new RestTemplate();

    public CustomerScore fetch(String customerId) {
        return restTemplate.getForObject(
            "http://localhost:8080/api/internal/scores/" + customerId,
            CustomerScore.class
        );
    }
}

📄 src/main/java/com/digitallending/service/OllamaClient.java

package com.digitallending.service;

import com.fasterxml.jackson.databind.ObjectMapper;
import org.springframework.beans.factory.annotation.Value;
import org.springframework.stereotype.Component;
import org.springframework.web.client.RestTemplate;

import java.util.Map;

@Component
public class OllamaClient {

    @Value("${ollama.url}")
    private String url;

    @Value("${ollama.model}")
    private String model;

    private final RestTemplate restTemplate = new RestTemplate();
    private final ObjectMapper mapper = new ObjectMapper();

    public String call(String prompt) {
        try {
            Map<String, Object> req = Map.of(
                "model", model,
                "prompt", prompt,
                "stream", false
            );

            String response = restTemplate.postForObject(url, req, String.class);
            return mapper.readTree(response).get("response").asText();
        } catch (Exception e) {
            return "LLM Error: " + e.getMessage();
        }
    }
}

📄 src/main/java/com/digitallending/service/ChatOrchestratorService.java

package com.digitallending.service;

import com.digitallending.model.ChatResponse;
import com.digitallending.model.CustomerScore;
import com.digitallending.prompt.PromptBuilder;
import org.springframework.stereotype.Service;

@Service
public class ChatOrchestratorService {

    private final CreditScoreService creditScoreService;
    private final OllamaClient ollamaClient;
    private final PromptBuilder promptBuilder;

    public ChatOrchestratorService(CreditScoreService c,
                                   OllamaClient o,
                                   PromptBuilder p) {
        this.creditScoreService = c;
        this.ollamaClient = o;
        this.promptBuilder = p;
    }

    public ChatResponse process(String customerId, String message) {
        CustomerScore score = creditScoreService.fetch(customerId);
        String prompt = promptBuilder.build(score);
        String llmReply = ollamaClient.call(prompt);

        return new ChatResponse(
            llmReply,
            promptBuilder.version(),
            "qwen3-v1-235b-cloud"
        );
    }
}

How to Run POC

===

✅ HOW TO RUN LOCALLY

1️⃣ Start Ollama (from desktop app)

Ensure API server is running at:

http://localhost:11434

Test:

curl http://localhost:11434/api/tags

2️⃣ Run Spring Boot

cd genai-credit-poc
mvn spring-boot:run

3️⃣ Open Swagger

http://localhost:8080/swagger-ui.html

4️⃣ Test API

POST /api/chat/messageHeader

X-Customer-Id: CUST1001

Body

{
  "message": "What is my credit score?"
}

✅ You now have a complete enterprise-style GenAI POC:

Microservice pattern
Prompt registry
Prompt as code
Ollama LLM integration
Swagger testing
BFSI-safe NLP generation

Request

===

Response

===

response

===