---
language:
- en
- id
license: apache-2.0
tags:
- anoman
- cybersecurity
- instruction-tuned
- bilingual
- asean
- gguf
pipeline_tag: text-generation
---

# Anoman-Base

**Anoman-Base** is a bilingual (English / Bahasa Indonesia) general-purpose AI assistant developed by the **Anoman AI team**, optimized for ASEAN technology teams.

## Capabilities

- General reasoning, writing, summarization, and Q&A
- Code review and software development across major languages
- Foundational cybersecurity awareness (OWASP, NIST CSF, ISO 27001)
- Southeast Asian regulatory knowledge: Indonesia (OJK, BI, BSSN), Singapore (MAS), Malaysia (BNM), Philippines (BSP, NPC), Thailand (NCSA, BOT)
- Bilingual: responds in the language the user writes in (English or Bahasa Indonesia)

For deep security work — vulnerability analysis, malware analysis, red teaming — see **Anoman-Sec** (coming soon).

## Model Details

| Property | Value |
|---|---|
| Architecture | Transformer decoder |
| Parameters | 7B |
| Quantization | Q4_K_M (GGUF) |
| Context length | 32,768 tokens |
| Languages | English, Bahasa Indonesia |
| License | Apache-2.0 |

## Usage

### Ollama (easiest)

```bash
# Pull directly from HuggingFace via Ollama
ollama run hf.co/anoman-ai/anoman-base
```

Or with a custom Modelfile:

```bash
# Download the Modelfile from this repo
ollama create anoman-base -f Modelfile
ollama run anoman-base "Who are you?"
```

### llama.cpp

```bash
./llama-cli -m anoman-base-Q4_K_M.gguf \
  --system-prompt "You are Anoman, an AI assistant developed by the Anoman AI team..." \
  -p "Apa itu prinsip least privilege?"
```

### LM Studio / Jan / Open WebUI

Download `anoman-base-Q4_K_M.gguf` and load it directly — compatible with any llama.cpp-based runtime.

## System Prompt

Anoman-Base uses the following default persona when run via Ollama:

```
You are Anoman, an AI assistant developed by the Anoman AI team.

You are fluent in English and Bahasa Indonesia, and you adapt to the language your user writes in.

You help with general reasoning, writing, code review, software development, and questions 
about cybersecurity fundamentals and Southeast Asian compliance frameworks 
(Indonesia: OJK, BI, BSSN; Singapore: MAS; Malaysia: BNM; Philippines: BSP, NPC; Thailand: NCSA, BOT).

For advanced security work — vulnerability analysis, malware analysis, or red teaming — 
direct users to Anoman Sec, the security-specialist model in the Anoman family.

Always reason step by step, be honest about uncertainty, and write secure code by default.
```

## Example Outputs

**English:**
> **User:** Write a Python function to validate a JWT token.
> **Anoman:** Here is a secure JWT validation function using the `python-jose` library...

**Bahasa Indonesia:**
> **User:** Apa itu SQL Injection dan bagaimana mencegahnya?
> **Anoman:** SQL Injection adalah teknik serangan di mana penyerang menyisipkan kode SQL berbahaya...

## License

Apache-2.0. Built on an Apache-2.0 licensed foundation model. See [LICENSE](https://www.apache.org/licenses/LICENSE-2.0) for full terms.

## About Anoman AI

Anoman is an AI assistant family built for ASEAN teams — general-purpose (Base) and security-specialized (Sec). Learn more at [anoman.io](https://anoman.io).