josephgoksu/ osprey-narrator

1 Download Updated yesterday

ollama run josephgoksu/osprey-narrator

curl http://localhost:11434/api/chat \
  -d '{
    "model": "josephgoksu/osprey-narrator",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='josephgoksu/osprey-narrator',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'josephgoksu/osprey-narrator',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size

Context

Input

osprey-narrator:latest

2.5GB · 256K context window · Text · yesterday

osprey-narrator:latest

2.5GB

256K

Text

Readme

Osprey Narrator

A fine-tuned LLM that generates SAR-style compliance narratives from transaction monitoring alerts.

Built by Joseph Goksu at opensource.finance — open-source infrastructure for the next generation of fintech.

What It Does

Osprey Narrator takes raw JSON output from the Osprey transaction monitoring engine and produces structured, analyst-ready compliance narratives — replacing hours of manual report writing with seconds of inference.

Input: Osprey evaluation JSON (rules, scores, typologies, transaction details) Output: Complete SAR-style narrative with risk assessment, FATF references, and recommended actions

Quick Start

ollama run josephgoksu/osprey-narrator

Or pipe an Osprey evaluation directly:

cat evaluation.json | ollama run josephgoksu/osprey-narrator

Model Details

Property	Value
Base Model	Qwen3-4B-Instruct-2507
Method	LoRA fine-tuning
Quantization	Q4_K_M
Size	2.5 GB
Test Perplexity	2.65
License	Apache 2.0

Coverage

12 FATF Rules — High-value transactions, structuring, geographic risk, shell companies, PEP, cross-border wires, and more
6 FATF Typologies — Smurfing, trade-based ML, shell company layering, PEP corruption, funnel accounts, currency exchange laundering

Output Sections

Each generated narrative includes:

Alert Summary — Alert ID, transaction type, amount, parties, decision
Transaction Details — Contextual description
Risk Assessment — Overall risk analysis with score interpretation
Rules Triggered — Table of all 12 FATF rules with scores
Typology Analysis — Table of 6 typologies with match status
Narrative — Human-readable explanation connecting the dots
Recommended Actions — Next steps for the compliance team

Links

HuggingFace: josephgoksu/osprey-narrator-v0.1
Osprey Engine: opensource-finance/osprey
Organization: opensource.finance

About

Part of the opensource.finance ecosystem. We’re engineers who built fraud detection systems for national payment infrastructure — now making that same rigor accessible to everyone.

Created by Joseph Goksu — GitHub · opensource.finance

# Osprey Narrator

**A fine-tuned LLM that generates SAR-style compliance narratives from transaction monitoring alerts.**

Built by [Joseph Goksu](https://github.com/josephgoksu) at [opensource.finance](https://opensource.finance) — open-source infrastructure for the next generation of fintech.

## What It Does

Osprey Narrator takes raw JSON output from the [Osprey](https://github.com/opensource-finance/osprey) transaction monitoring engine and produces structured, analyst-ready compliance narratives — replacing hours of manual report writing with seconds of inference.

**Input**: Osprey evaluation JSON (rules, scores, typologies, transaction details)
**Output**: Complete SAR-style narrative with risk assessment, FATF references, and recommended actions

## Quick Start

```bash
ollama run josephgoksu/osprey-narrator
```

Or pipe an Osprey evaluation directly:

```bash
cat evaluation.json | ollama run josephgoksu/osprey-narrator
```

## Model Details

| Property | Value |
|----------|-------|
| **Base Model** | Qwen3-4B-Instruct-2507 |
| **Method** | LoRA fine-tuning |
| **Quantization** | Q4_K_M |
| **Size** | 2.5 GB |
| **Test Perplexity** | 2.65 |
| **License** | Apache 2.0 |

## Coverage

- **12 FATF Rules** — High-value transactions, structuring, geographic risk, shell companies, PEP, cross-border wires, and more
- **6 FATF Typologies** — Smurfing, trade-based ML, shell company layering, PEP corruption, funnel accounts, currency exchange laundering

## Output Sections

Each generated narrative includes:

1. **Alert Summary** — Alert ID, transaction type, amount, parties, decision
2. **Transaction Details** — Contextual description
3. **Risk Assessment** — Overall risk analysis with score interpretation
4. **Rules Triggered** — Table of all 12 FATF rules with scores
5. **Typology Analysis** — Table of 6 typologies with match status
6. **Narrative** — Human-readable explanation connecting the dots
7. **Recommended Actions** — Next steps for the compliance team

## Links

- **HuggingFace**: [josephgoksu/osprey-narrator-v0.1](https://huggingface.co/josephgoksu/osprey-narrator-v0.1)
- **Osprey Engine**: [opensource-finance/osprey](https://github.com/opensource-finance/osprey)
- **Organization**: [opensource.finance](https://opensource.finance)

## About

Part of the [opensource.finance](https://opensource.finance) ecosystem. We're engineers who built fraud detection systems for national payment infrastructure — now making that same rigor accessible to everyone.

Created by **Joseph Goksu** — [GitHub](https://github.com/josephgoksu) · [opensource.finance](https://opensource.finance)

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)