Instructions to use kimjg/recallResolver-1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kimjg/recallResolver-1B with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/llama-3.2-1b-instruct-unsloth-bnb-4bit")
model = PeftModel.from_pretrained(base_model, "kimjg/recallResolver-1B")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use kimjg/recallResolver-1B 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 kimjg/recallResolver-1B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kimjg/recallResolver-1B:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kimjg/recallResolver-1B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kimjg/recallResolver-1B: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 kimjg/recallResolver-1B:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf kimjg/recallResolver-1B: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 kimjg/recallResolver-1B:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf kimjg/recallResolver-1B:Q4_K_M

Use Docker

docker model run hf.co/kimjg/recallResolver-1B:Q4_K_M

LM Studio
Jan

vLLM

How to use kimjg/recallResolver-1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kimjg/recallResolver-1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kimjg/recallResolver-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kimjg/recallResolver-1B:Q4_K_M

Ollama
How to use kimjg/recallResolver-1B with Ollama:
```
ollama run hf.co/kimjg/recallResolver-1B:Q4_K_M
```
Unsloth Desktop

How to use kimjg/recallResolver-1B with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kimjg/recallResolver-1B: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": "kimjg/recallResolver-1B:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use kimjg/recallResolver-1B with Docker Model Runner:
```
docker model run hf.co/kimjg/recallResolver-1B:Q4_K_M
```

Lemonade

How to use kimjg/recallResolver-1B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull kimjg/recallResolver-1B:Q4_K_M

Run and chat with the model

lemonade run user.recallResolver-1B-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use kimjg/recallResolver-1B with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kimjg/recallResolver-1B: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 kimjg/recallResolver-1B:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use kimjg/recallResolver-1B with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kimjg/recallResolver-1B: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 "kimjg/recallResolver-1B: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"

recallResolver-1B

한국어 멀티턴 파이프라인의 히스토리 값 추출기 — "그거 재고 얼마나 남았어?"처럼 과거 대화의 값을 가리키는 발화가 오면, 대화 기록에서 그 값을 찾아주는 1B 모델의 QLoRA 어댑터(~45MB)입니다.

toolRouter-1B와 짝으로 설계됐습니다: 라우터가 과거 참조를 감지하면, recallResolver가 히스토리를 교환 단위 하나씩 받아 값을 추출하고, 서버가 그 값으로 발화를 보강해 라우터를 재호출합니다. 입력이 항상 "단위 1개 + 현재 발화 + needs"로 고정이라 히스토리가 아무리 길어도 컨텍스트가 자라지 않습니다.

입출력

system: (고정 지시 — 기록에서 needs에 해당하는 값을 그대로 추출, 없으면 false, 지어내기 금지)
user:
  [기록]
  user: SKU 44871 재고 몇 개 남았는지 봐줘
  call: {"name": "get_stock", "arguments": {"sku": "44871"}}
  result: {"ok": true, "sku": "44871", "qty": 120}
  [현재 요청] 그거 어느 구역에 보관돼 있는 거야?
  [needs] 상품 SKU 번호
assistant: {"found": true, "value": "44871"}

값이 기록에 없으면 {"found": false}. 출력은 xgrammar 등으로 JSON 형식을 강제하고, value가 기록 텍스트에 글자 그대로 존재하는지 서버에서 검증(verbatim 가드)하는 것을 권장합니다.

성능 (자체 평가셋 613건, 학습과 분리)

지표	수치
found 판정 정확도	96.3%
value 정확일치 (found 중)	94.2%
오탐률 (없는데 찾았다고 함)	2.5%
형식(JSON 파싱)	100%

오탐은 4회의 hard-negative 채굴 반복으로 9.9% → 2.5%까지 압축했습니다. 남은 약점은 같은 계열 식별자의 하위 유형 혼동(주문 번호 ↔ 송장 번호)으로, 서버 측 되묻기 규칙으로 보완하는 것을 권장합니다.

사용법

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = "meta-llama/Llama-3.2-1B-Instruct"
model = AutoModelForCausalLM.from_pretrained(base, device_map="auto")
model = PeftModel.from_pretrained(model, "kimjg/recallResolver-1B")
tokenizer = AutoTokenizer.from_pretrained("kimjg/recallResolver-1B")

toolRouter-1B와 같은 베이스를 쓰므로, 베이스 한 벌에 어댑터 두 개를 얹고 set_adapter로 전환하면 두 모델이 VRAM ~2.6GB(4bit) 에서 함께 동작합니다.

GGUF (llama.cpp)

머지·양자화한 단일 파일도 제공합니다: recallResolver-1B-Q4_K_M.gguf (0.8GB) / recallResolver-1B-Q8_0.gguf (1.3GB). 리졸버 기준 양자화 손실은 아직 실측 전입니다 — 같은 베이스·같은 방식의 toolRouter-1B 실측(Q8 ~-0.8%p, Q4 ~-1.6%p)이 참고치이며, 정밀도가 중요하면 Q8을 권합니다. 공유 베이스 구성(toolRouter-1B와 base GGUF 한 벌 + LoRA 2개)이 VRAM에 더 유리합니다(실측 1.5GB).

제한

추출 전용입니다 — 요약·계산·비교는 하지 않습니다 (그런 요청은 상위 모델로 escalate 권장).
결과(result)는 250자 내로 절단해 입력하세요. ID류 필드를 앞에 배치하면 잘림에 안전합니다.
greedy 디코딩 권장. base 모델의 Llama 3.2 Community License를 따릅니다.

Downloads last month: -

GGUF

Model size

1B params

Architecture

llama

Hardware compatibility

4-bit

8-bit

Model tree for kimjg/recallResolver-1B

Base model

meta-llama/Llama-3.2-1B-Instruct

Adapter

(665)

this model