Instructions to use soumyaprasadrana/maximo-slm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use soumyaprasadrana/maximo-slm with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-7B-Instruct-bnb-4bit")
model = PeftModel.from_pretrained(base_model, "soumyaprasadrana/maximo-slm")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use soumyaprasadrana/maximo-slm with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf soumyaprasadrana/maximo-slm:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf soumyaprasadrana/maximo-slm:Q5_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf soumyaprasadrana/maximo-slm:Q5_K_M
# Run inference directly in the terminal:
llama cli -hf soumyaprasadrana/maximo-slm:Q5_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf soumyaprasadrana/maximo-slm:Q5_K_M
# Run inference directly in the terminal:
./llama-cli -hf soumyaprasadrana/maximo-slm:Q5_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf soumyaprasadrana/maximo-slm:Q5_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf soumyaprasadrana/maximo-slm:Q5_K_M

Use Docker

docker model run hf.co/soumyaprasadrana/maximo-slm:Q5_K_M

LM Studio
Jan

vLLM

How to use soumyaprasadrana/maximo-slm with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "soumyaprasadrana/maximo-slm"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "soumyaprasadrana/maximo-slm",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/soumyaprasadrana/maximo-slm:Q5_K_M

Ollama
How to use soumyaprasadrana/maximo-slm with Ollama:
```
ollama run hf.co/soumyaprasadrana/maximo-slm:Q5_K_M
```

Unsloth Studio

How to use soumyaprasadrana/maximo-slm with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for soumyaprasadrana/maximo-slm to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for soumyaprasadrana/maximo-slm to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for soumyaprasadrana/maximo-slm to start chatting

How to use soumyaprasadrana/maximo-slm with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf soumyaprasadrana/maximo-slm:Q5_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "soumyaprasadrana/maximo-slm:Q5_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use soumyaprasadrana/maximo-slm with Docker Model Runner:
```
docker model run hf.co/soumyaprasadrana/maximo-slm:Q5_K_M
```

Lemonade

How to use soumyaprasadrana/maximo-slm with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull soumyaprasadrana/maximo-slm:Q5_K_M

Run and chat with the model

lemonade run user.maximo-slm-Q5_K_M

List all available models

lemonade list

Hermes Agent

How to use soumyaprasadrana/maximo-slm with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf soumyaprasadrana/maximo-slm:Q5_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default soumyaprasadrana/maximo-slm:Q5_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use soumyaprasadrana/maximo-slm with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf soumyaprasadrana/maximo-slm:Q5_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "soumyaprasadrana/maximo-slm:Q5_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

MaximoSLM (Phase 1, initial)

Read-only small language model for IBM Maximo. It drives the public @soumyaprasadrana/maximo-mcp-server read tools so a laptop can query live Maximo without a giant system prompt describing the whole catalog.

This is an early research preview, not a production Maximo product and not affiliated with IBM. Phase 1 can inspect metadata and query records. It cannot create, update, delete, change status, or run workflow.

Base model: Qwen/Qwen2.5-7B-Instruct (Apache-2.0). Method: QLoRA (Unsloth), attention and MLP adapters only. embed_tokens and lm_head are not trained.

What this repository contains

Path	What it is
`adapter_config.json`, `adapter_model.safetensors`	QLoRA adapters (primary artifact)
tokenizer / chat template files	Copied from the Instruct tokenizer
`SYSTEM_PROMPT.txt`	Frozen system prompt (train = serve)
`Modelfile`	Ollama recipe; `FROM` points at the GGUF in `gguf/`
`gguf/*.gguf`	Q5_K_M GGUF for Ollama (when uploaded)
`TRAINING_CARD.json`	Hyperparameters from the training run

Not uploaded: training JSONL, Maximo credentials, merged fp16 safetensors.

Intended use

A small agent loop (this repo's orchestrator/) calls the model, parses one <tool_call> JSON object per turn, and executes only these MCP tools:

maximo_get_metadata
os_query_builder
ws_load
ws_get_records
ws_get_active

Typical read pipeline:

maximo_get_metadata  ->  os_query_builder  ->  ws_load (useLean true)

The agent owns the working-set id. The model chooses the object structure, filters, select list, and sort.

Load adapters (Unsloth / GPU)

hf download YOUR_USER/maximo-slm --local-dir maximo-slm-lora

Then, from the MaximoSLM git checkout:

python test/eval_in_env.py
python test/chat.py --unsloth

ADAPTER_DIR defaults to ./maximo-slm-lora. Serve with the same SYSTEM_PROMPT.txt that is in this repo. Do not LoRA embed_tokens or lm_head on top of these weights.

Load GGUF (Ollama)

hf download YOUR_USER/maximo-slm --include "gguf/*" --local-dir maximo-slm-hf
cd maximo-slm-hf
ollama create maximo-slm -f Modelfile

Chat through MCP (needs test/.env with MAXIMO_URL and an API key, never commit that file):

python test/chat.py

Ollama injects SYSTEM from the Modelfile. Do not also send a system message.

Limits (Phase 1)

Read-only. Write and admin MCP tools are refused by the client.
7B QLoRA will mistype UUIDs and some filters. The client must bind the last working-set id and slim tool payloads.
Working sets expire in about 10 minutes; rebuild and continue.
Quote meta.totalCount for "how many", never the page length.
orderBy needs a + or - prefix.