Instructions to use srikanth235/centraid-native-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use srikanth235/centraid-native-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 srikanth235/centraid-native-gguf:Q4_0
# Run inference directly in the terminal:
llama cli -hf srikanth235/centraid-native-gguf:Q4_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf srikanth235/centraid-native-gguf:Q4_0
# Run inference directly in the terminal:
llama cli -hf srikanth235/centraid-native-gguf:Q4_0

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 srikanth235/centraid-native-gguf:Q4_0
# Run inference directly in the terminal:
./llama-cli -hf srikanth235/centraid-native-gguf:Q4_0

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 srikanth235/centraid-native-gguf:Q4_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf srikanth235/centraid-native-gguf:Q4_0

Use Docker

docker model run hf.co/srikanth235/centraid-native-gguf:Q4_0

LM Studio
Jan

vLLM

How to use srikanth235/centraid-native-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "srikanth235/centraid-native-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": "srikanth235/centraid-native-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/srikanth235/centraid-native-gguf:Q4_0

Ollama
How to use srikanth235/centraid-native-gguf with Ollama:
```
ollama run hf.co/srikanth235/centraid-native-gguf:Q4_0
```
Unsloth Desktop

How to use srikanth235/centraid-native-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf srikanth235/centraid-native-gguf:Q4_0

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": "srikanth235/centraid-native-gguf:Q4_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use srikanth235/centraid-native-gguf with Docker Model Runner:
```
docker model run hf.co/srikanth235/centraid-native-gguf:Q4_0
```

Lemonade

How to use srikanth235/centraid-native-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull srikanth235/centraid-native-gguf:Q4_0

Run and chat with the model

lemonade run user.centraid-native-gguf-Q4_0

List all available models

lemonade list

Hermes Agent

How to use srikanth235/centraid-native-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 srikanth235/centraid-native-gguf:Q4_0

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 srikanth235/centraid-native-gguf:Q4_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use srikanth235/centraid-native-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf srikanth235/centraid-native-gguf:Q4_0

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 "srikanth235/centraid-native-gguf:Q4_0" \
  --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"

centraid-native-s2 (Q4_0)

A GGUF of a fine-tune of Qwen/Qwen3.5-0.8B (Apache-2.0) for the native tool task (#1044) of Centraid: a personal, local-first superapp whose assistant runs on the phone. The model drives a personal vault through eight tools (SPEC.md): each step it writes a short think in a fixed slot format, and the app's runtime reads the think and writes the call.

This is not a general chat model. It was trained to continue the exact prompt the Centraid runtime renders (the tool list, the vault block, the history) and to write the think the runtime compiles; outside that runtime it has no use. It is published so the on-device model can be fetched, checked against its hash and reproduced.

Provenance

Base model: Qwen/Qwen3.5-0.8B.
Fine-tuned model: the private repository srikanth235/centraid-native-models, tag s2, revision 430867aaf2949475709fdf931784bbd9343b7fb9.
Data version: data-v7 of the native tool task's private dataset repository (authored, synthetic households; no personal data).
Converted with llama.cpp 26394b4e6749a41c3633db040e0987500a5f7013 (https://github.com/ggml-org/llama.cpp), the commit that llama-cpp-sys-2 0.1.158 (the engine the app links) vendors: convert_hf_to_gguf.py to F16, then llama-quantize to Q4_0. centraid-native-s2-Q4_0.gguf.json records the checkpoint's file hashes, the converter and quantizer hashes and the tool versions; train/to_gguf.py in the repository rebuilds this file byte for byte.

Evaluation

Scored on the native tool task's held-out validation set, val v7.4 of data-v7 (655 multi-turn sessions over seven synthetic households), through assist-step: the phone's own decode step and llama.cpp engine, on CPU, greedy. A session passes when every turn's effect on the vault and every answer matches the gold.

model	decode	sessions passed
this file (`Q4_0`)	the phone's (`assist-step`, llama.cpp)	557 / 655 (85.0 %, 95 % CI 82.1–87.6 %)
S2 (the fine-tuned weights before conversion)	the training harness's free decoding, its recorded outputs replayed on the same runtime and gold	542 / 655 (82.7 %)

S2's live score of 537 / 655 was on the earlier gold, val v7.3. Turns: 1,912 of 2,055 pass; 33 turns write something the gold does not. No think ran past its limit.

Only the tool task. Asked to describe a photograph through the base model's projector with a plain chat prompt, this model writes a tool-call think instead of a description: fine-tuning on the tool task alone took the base model's general chat with it. Use the base model for anything else.

Vision

A photograph is read by a separate projector the app loads beside these weights; fine-tuning left the vision tower as it was, so the projector is the base model's: mmproj-F16.gguf of unsloth/Qwen3.5-0.8B-GGUF at revision 6ab461498e2023f6e3c1baea90a8f0fe38ab64d0 (204,987,232 bytes, sha256 56e4c6cfe73b0c82e3e82bc518d7591997e61d81f723fc41a586f4fa69ea2453). It is not part of this repository.

Files

file	bytes	sha256
`centraid-native-s2-Q4_0.gguf`	563,036,224	`b1be1e45024c8e98f670d456edfdf261c786878dd8260117806e285af8e9f668`

Downloads last month: 134

GGUF

Model size

0.8B params

Architecture

qwen35

Hardware compatibility

4-bit

Model tree for srikanth235/centraid-native-gguf

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Quantized

(321)

this model