Instructions to use CodeMonkey1/gemma-superpowers-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CodeMonkey1/gemma-superpowers-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="CodeMonkey1/gemma-superpowers-gguf",
	filename="gemma-superpowers-v2.7.Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use CodeMonkey1/gemma-superpowers-gguf 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 CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf CodeMonkey1/gemma-superpowers-gguf:Q4_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 CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf CodeMonkey1/gemma-superpowers-gguf:Q4_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 CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

Use Docker

docker model run hf.co/CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use CodeMonkey1/gemma-superpowers-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CodeMonkey1/gemma-superpowers-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CodeMonkey1/gemma-superpowers-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

Ollama
How to use CodeMonkey1/gemma-superpowers-gguf with Ollama:
```
ollama run hf.co/CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
```

Unsloth Studio

How to use CodeMonkey1/gemma-superpowers-gguf 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 CodeMonkey1/gemma-superpowers-gguf 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 CodeMonkey1/gemma-superpowers-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for CodeMonkey1/gemma-superpowers-gguf to start chatting

How to use CodeMonkey1/gemma-superpowers-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf CodeMonkey1/gemma-superpowers-gguf:Q4_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": "CodeMonkey1/gemma-superpowers-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use CodeMonkey1/gemma-superpowers-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf CodeMonkey1/gemma-superpowers-gguf:Q4_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 CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use CodeMonkey1/gemma-superpowers-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf CodeMonkey1/gemma-superpowers-gguf:Q4_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 "CodeMonkey1/gemma-superpowers-gguf:Q4_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"

Docker Model Runner
How to use CodeMonkey1/gemma-superpowers-gguf with Docker Model Runner:
```
docker model run hf.co/CodeMonkey1/gemma-superpowers-gguf:Q4_K_M
```

Lemonade

How to use CodeMonkey1/gemma-superpowers-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull CodeMonkey1/gemma-superpowers-gguf:Q4_K_M

Run and chat with the model

lemonade run user.gemma-superpowers-gguf-Q4_K_M

List all available models

lemonade list

CodeMonkey1 commited on 24 days ago

Commit

4aa385f

verified ·

1 Parent(s): e24b1a3

v2.1: native tool-calling model card

Browse files

Files changed (1) hide show

README.md +20 -12

README.md CHANGED Viewed

@@ -3,14 +3,14 @@ base_model: unsloth/gemma-4-12b-it
 license: gemma
 language: [en]
 library_name: gguf
-tags: [gguf, llama.cpp, lm-studio, gemma, gemma-4, qlora, unsloth, methodology, coding-assistant, agent]
 pipeline_tag: text-generation
 quantized_by: CodeMonkey1
 ---
 # 🦾 Gemma-4-12B — Superpowers Edition
-### *A local coding companion that thinks before it types.*
 Most models hear "build me a tool" and immediately vomit code. **Not this one.** Gemma-4-12B-Superpowers has six engineering disciplines fine-tuned **into its weights** — so it reaches for the right method on its own, **with no system prompt, no jailbreak, no babysitting.**
@@ -29,31 +29,39 @@ Most models hear "build me a tool" and immediately vomit code. **Not this one.**
 …and she **won't** over-think a one-liner — ask "what's 2+2" and you get **`4`**, not a discovery meeting.
 ## 🛠️ Built for real work — WordPress, head to toe
 ~40% of her training lives in the trenches you actually work in: **PHP** (ACF, hooks, `WP_Query`, WP-CLI), **JavaScript** (Gutenberg blocks, enqueued scripts, jQuery/vanilla), and **CSS** (responsive, CLS-safe fonts, WCAG 2.2). She debugs *your* stack, not toy code.
 ## 🚀 Quick start (LM Studio)
-1. Download **`gemma-4-12b-it.Q4_K_M.gguf`** → load in LM Studio.
-2. **Leave the system prompt EMPTY.** The discipline is baked in — that's the whole point.
-3. Talk to her like a sharp junior dev. Watch her ask the right questions.
-*Want her eyes too?* Also grab **`…BF16-mmproj.gguf`** — Gemma 4's vision projector. Drop it in the same folder as the text GGUF and LM Studio pairs it automatically; image input works (confirmed describing real photos). Vision weights are stock Gemma 4 — untouched by the text fine-tune.
 ## 🔧 Under the hood
-QLoRA on **`unsloth/gemma-4-12b-it`** via [Unsloth](https://github.com/unslothai/unsloth) — r=16, ~1,134 examples (389 discipline + 145 "answer-it-straight" + ~600 general to keep her smart), 2 epochs, **no system message in training** (so the behavior is unconditioned). Methodology adapted from the open-source [superpowers](https://github.com/obra/superpowers) skills (MIT).
-## 🔁 v1.1 — serving fixes (2026-06-15)
-- **Chat template corrected** to the canonical Gemma-4-12B-it template. The earlier build shipped a stale template that made LM Studio's engine-protocol runtime throw `Failed to parse input` (HTTP 400) mid-generation. Fixed — code-gen eval now runs clean (6/6, no 400s) on q4 and q6.
-- **Vision restored.** The earlier `mmproj` was mis-converted and crashed on load; replaced with the stock (weight-compatible) Gemma-4-12B projector. Image input confirmed working.
 ## ⚠️ Status
-**v1 — experimental.** Code-precision A/B eval (q4 vs q6) shows both quants produce correct, idiomatic code across the probe set; the disciplines fire with no system prompt. It's a 12B running locally, so it follows the disciplines well but not frontier-perfectly. Inherits the [Gemma Terms of Use](https://ai.google.dev/gemma/terms).
 ---
-*Fine-tuned with stubbornness and ~$0.21 of GPU time. 🧪*

 license: gemma
 language: [en]
 library_name: gguf
+tags: [gguf, llama.cpp, lm-studio, gemma, gemma-4, qlora, unsloth, methodology, coding-assistant, agent, tool-calling, function-calling]
 pipeline_tag: text-generation
 quantized_by: CodeMonkey1
 ---
 # 🦾 Gemma-4-12B — Superpowers Edition
+### *A local coding companion that thinks before it types — and now calls tools natively.*
 Most models hear "build me a tool" and immediately vomit code. **Not this one.** Gemma-4-12B-Superpowers has six engineering disciplines fine-tuned **into its weights** — so it reaches for the right method on its own, **with no system prompt, no jailbreak, no babysitting.**
 …and she **won't** over-think a one-liner — ask "what's 2+2" and you get **`4`**, not a discovery meeting.
+## 🔧 NEW in v2 — native tool-calling
+She now emits **Gemma-4's native `<|tool_call>` format**, so function/tool calling fires cleanly in LM Studio (and any engine that speaks the canonical Gemma-4 template) **with no `{action, parameters}` dispatcher and no AnythingLLM in the loop.** Trained against **real tool schemas** (74-tool catalog: filesystem, WordPress/MCP, skills) and **many-tools-at-once** menus, so she selects the right tool from a big menu and emits correct argument names instead of inventing them.
+> **⚠️ Sampling matters for tool-calling.** Structured output (JSON, tool calls) needs the model to *repeat* tokens like `"`, `{`, `:` and field names. Keep `repeat_penalty` ≈ **1.0–1.1** and `presence_penalty` off. High penalties (e.g. 2.0) shred tool calls — you'll see `<|"|>` leaks, hallucinated IDs, and half-line repeats. This is a sampling bug, not the model.
 ## 🛠️ Built for real work — WordPress, head to toe
 ~40% of her training lives in the trenches you actually work in: **PHP** (ACF, hooks, `WP_Query`, WP-CLI), **JavaScript** (Gutenberg blocks, enqueued scripts, jQuery/vanilla), and **CSS** (responsive, CLS-safe fonts, WCAG 2.2). She debugs *your* stack, not toy code.
 ## 🚀 Quick start (LM Studio)
+1. Download **`gemma-4-12b-it.Q6_K.gguf`** → load in LM Studio.
+2. **Leave the discipline system prompt EMPTY.** The behavior is baked in — that's the whole point. (A short native-tool system prompt is fine if you want to bound her thinking.)
+3. For tool work: enable LM Studio's tool/MCP integration and keep sampling sane (see the warning above).
+4. Talk to her like a sharp junior dev. Watch her ask the right questions.
+*Want her eyes too?* Also grab **`mmproj-gemma-4-12B-it-BF16.gguf`** — Gemma 4's vision projector. Drop it in the same folder as the text GGUF and LM Studio pairs it automatically; image input works (confirmed describing real photos). Vision weights are stock Gemma 4 — untouched by the text fine-tune.
 ## 🔧 Under the hood
+QLoRA on **`unsloth/gemma-4-12b-it`** via [Unsloth](https://github.com/unslothai/unsloth). v2.1: **r=32, lora_alpha=64, 3 epochs, max_seq_len=4096**, ~808 examples (553 discipline + 145 "answer-it-straight" negatives + 60 tool + 50 agentic, blended with general/Dolly data to keep her smart), **trained with the canonical Gemma-4 chat template** so what she learns is exactly what LM Studio parses — and **no system message in training** (so the discipline is unconditioned). Methodology adapted from the open-source [superpowers](https://github.com/obra/superpowers) skills (MIT).
+## 🧬 Version history
+- **v2.1 (2026-06-16) — native tool-calling.** Retrained on the canonical Gemma-4 template against real tool schemas + large tool menus. Emits native `<|tool_call>` (`has_tool_calls: True` where v1 returned a fabricated answer). Garbling traced to a serving-side sampling preset, not the weights. Ships q6 + the stock vision projector.
+- **v1.1 (2026-06-15) — serving fixes.** Corrected the chat template (the earlier build shipped a stale template that 400'd LM Studio's engine-protocol runtime mid-generation); replaced the mis-converted mmproj with the stock weight-compatible Gemma-4-12B projector.
+- **v1 — experimental.** Disciplines fired with no system prompt; tool-calling only worked in a non-native `{action, parameters}` dialect.
 ## ⚠️ Status
+**v2.1.** Native tool-calling confirmed; disciplines fire with no system prompt; code-gen runs clean. It's a 12B running locally, so it follows the disciplines well but not frontier-perfectly. Inherits the [Gemma Terms of Use](https://ai.google.dev/gemma/terms).
 ---
+*Fine-tuned with stubbornness and a few dollars of GPU time. 🧪*