Instructions to use yannelli/d1-3b-code-review-decisions-ollama with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use yannelli/d1-3b-code-review-decisions-ollama 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 yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Use Docker

docker model run hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Llama new
LM Studio
Jan
Ollama
How to use yannelli/d1-3b-code-review-decisions-ollama with Ollama:
```
ollama run hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M
```
Unsloth Desktop

How to use yannelli/d1-3b-code-review-decisions-ollama with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use yannelli/d1-3b-code-review-decisions-ollama with Docker Model Runner:
```
docker model run hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M
```

Lemonade

How to use yannelli/d1-3b-code-review-decisions-ollama with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Run and chat with the model

lemonade run user.d1-3b-code-review-decisions-ollama-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use yannelli/d1-3b-code-review-decisions-ollama with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf yannelli/d1-3b-code-review-decisions-ollama: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 yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use yannelli/d1-3b-code-review-decisions-ollama with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf yannelli/d1-3b-code-review-decisions-ollama: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 "yannelli/d1-3b-code-review-decisions-ollama: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"

d1-3B Code Review Decisions for Ollama

d1-3B Code Review Decisions packaged for Ollama. Give it a JavaScript or TypeScript snippet and the task the snippet implements, and it answers 13 typed questions with probabilities: which of 10 code review flaws the code has, whether it has any flaw, which flaw matters most, and how mergeable it is.

These are the same weights as the d1-3B files in yannelli/code-review-decisions-GGUF, with the lfm2.decision.type key removed. Ollama (tested with v0.40.3) refuses completion for a GGUF that carries that key, and its /v1/systemone endpoint does not support d1. ollama_d1.py sends d1's prompt for each question and reads the log-probabilities of the answer tokens. For llama.cpp, use the GGUF repo instead, because its files work with the native /v1/systemone endpoint of llama-server, which is about 6 times faster.

Quick start

ollama pull hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M    # or :Q8_0
curl -fsSLO https://huggingface.co/yannelli/d1-3b-code-review-decisions-ollama/resolve/main/code_review_tui.sh
sh code_review_tui.sh

code_review_tui.sh is a terminal UI. You pick the model and the language, enter the task and the code, and it shows each flaw's probability, the primary flaw and the quality level. It needs python3, curl and gum (brew install gum). On first run it downloads ollama_d1.py and the 13 questions, and it pulls the model if none is installed.

Python

ollama_d1.py uses only the Python standard library.

import json
import urllib.request

from ollama_d1 import system_one  # curl -fsSLO https://huggingface.co/yannelli/d1-3b-code-review-decisions-ollama/resolve/main/ollama_d1.py

manifest = "https://huggingface.co/datasets/yannelli/js-ts-code-review-decisions/resolve/main/manifest.json"
questions = json.load(urllib.request.urlopen(manifest))["questions"]
state = {"task": "Parse a CSV upload into rows.", "language": "TypeScript", "code": "..."}

answers = system_one(state, questions, model="hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M")["answers"]
print(answers["primary_flaw"]["choice"], answers["unsafe_input"]["noul"])

A noul answer is P(yes). A choice answer has the top choice and probabilities over all options. A score answer has score (the expected level) and probabilities over levels 0 ("Should not merge.") to 4 ("Exemplary. Merge as is."). From the command line, python ollama_d1.py --model MODEL request.json answers a JSON file that holds state and questions.

Files

File	Size	Use
`d1-3b-code-review-decisions-Q4_K_M.gguf`	1.7 GB	`ollama pull hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q4_K_M`
`d1-3b-code-review-decisions-Q8_0.gguf`	2.9 GB	`ollama pull hf.co/yannelli/d1-3b-code-review-decisions-ollama:Q8_0`
`ollama_d1.py`		System One answers from the model through Ollama's `/api/generate`
`code_review_tui.sh`		Terminal UI on top of `ollama_d1.py`

Results

Test split of JS/TS Code Review Decisions, scored with its score_decisions.py. Ollama ran on one NVIDIA L40S with every layer on the GPU, on 300 test snippets (3,900 decisions).

Build	Macro F1 @ 0.5	Accuracy	Top answers matching PyTorch	p50 per snippet	p95 per snippet
Q8_0	0.930	0.944	99.9%	1.30 s	1.60 s
Q4_K_M	0.921	0.941	99.3%	1.40 s	1.60 s

On the same 300 snippets the PyTorch model scores 0.930 macro F1 and 0.944 accuracy. On all 1,200 test snippets it scores 0.917 and 0.940, and the source model card has the full results and limitations. Ollama sends one request per question, so a snippet takes 13 requests, and on 16 Neoverse V3 CPU cores Q8_0 took 67 s per snippet.

License

These files are derivatives of LiquidAI/d1-3B and are released under the base model's LFM Open License v1.0, included as LICENSE. The license allows commercial use only for entities with less than $10M in annual revenue, and LICENSE has the full terms.

The training data, JS/TS Code Review Decisions, is CC BY 4.0.

Citation

@misc{yannelli2026d13bcodereview,
  title        = {d1-3B Code Review Decisions},
  author       = {Ryan Yannelli},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/yannelli/d1-3b-code-review-decisions}}
}