Instructions to use cyberalien/CySearchAssist-CLAP-1.5B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use cyberalien/CySearchAssist-CLAP-1.5B 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 cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
# Run inference directly in the terminal:
llama cli -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
# Run inference directly in the terminal:
llama cli -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_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 cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_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 cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_0

Use Docker

docker model run hf.co/cyberalien/CySearchAssist-CLAP-1.5B:Q8_0

LM Studio
Jan
Ollama
How to use cyberalien/CySearchAssist-CLAP-1.5B with Ollama:
```
ollama run hf.co/cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
```
Unsloth Desktop

How to use cyberalien/CySearchAssist-CLAP-1.5B with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_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": "cyberalien/CySearchAssist-CLAP-1.5B:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use cyberalien/CySearchAssist-CLAP-1.5B with Docker Model Runner:
```
docker model run hf.co/cyberalien/CySearchAssist-CLAP-1.5B:Q8_0
```

Lemonade

How to use cyberalien/CySearchAssist-CLAP-1.5B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull cyberalien/CySearchAssist-CLAP-1.5B:Q8_0

Run and chat with the model

lemonade run user.CySearchAssist-CLAP-1.5B-Q8_0

List all available models

lemonade list

Hermes Agent

How to use cyberalien/CySearchAssist-CLAP-1.5B with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_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 cyberalien/CySearchAssist-CLAP-1.5B:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use cyberalien/CySearchAssist-CLAP-1.5B with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf cyberalien/CySearchAssist-CLAP-1.5B:Q8_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 "cyberalien/CySearchAssist-CLAP-1.5B:Q8_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"

CySearchAssist-CLAP-1.5B

A small multilingual language model fine-tuned by MrMybal (CyberAlien) to rewrite informal sound-search requests into structured English captions for CLAP-based semantic audio search. This is not CLAP and it does not analyze audio.

Download

Download CySearchAssist-v1-Q8_0.gguf from Hugging Face. Size: 1,646,572,544 bytes (1.65 GB / 1.53 GiB).

SHA-256:

bccbc069fe26c19c44b60f7f06f6edcb9fe9883f3684af772b62f306428448dd

Weights are hosted on Hugging Face, not in the GitHub repository. manifest.json records the version, download URL, size and checksum. CyMetaSound pins a specific Hugging Face commit and verifies SHA-256 before installation. No account/token is required to download this public model. Keep LICENSE and NOTICE with the model.

Usage contract

Use the exact system prompt in prompt/system.txt and send the original search request as the user message. Use greedy decoding (temperature 0), at most 256 new tokens, and a suitable Qwen chat template (included in the GGUF metadata). The result is a JSON object with:

clap_prompt: principal English sound description to send to CLAP.
variations: three alternative descriptions.
suggested_categories and suggested_tags: optional ranking hints, not hard filters.
exclude: concepts explicitly excluded in the user's request.

Example request: je cherche un impact metallique dans un hangar, sans musique. Review the generated caption. A failed/malformed answer should fall back to the original request. Keep exclusions only when the request actually contains a negation; never add excluded words to a positive search caption. The model does not know your library, so ignore categories absent from your taxonomy.

Run the GGUF with a compatible llama.cpp runtime. A Vulkan build supports GPU offload without requiring CUDA; actual device support depends on the runtime and driver. The runtime is separate and is not included in this model release.

CyMetaSound

Updated CyMetaSound versions list CySearchAssist-CLAP-1.5B v1 (GGUF Q8) in the semantic query-refinement settings. Choose Local LLM, install the search model with the confirmation dialog, configure a Vulkan llama-server executable and select Auto or Vulkan. The installed GGUF is detected in the model cache unless an explicit GGUF override is configured. Use AI Refine in the Library.

CLAP/LAION or other embedding models remain separate optional downloads under their respective licences. CySearchAssist only prepares text for that search. Without this model, ordinary search remains available. No audio is sent to it.

Provenance and limitations

Underlying model: Qwen2.5-1.5B-Instruct.
Supplied training report identifies unsloth/Qwen2.5-1.5B-Instruct as the training distribution.
Fine-tune and adapter merge: MrMybal / CyberAlien, 2026.
This release is the merged fine-tune converted to Q8_0 GGUF using llama.cpp b11115.
Intended languages: French, English, Spanish, German and Italian.
Not a general assistant, audio generator, audio embedding model or factual authority.
The model can omit constraints or invent details, especially with severe typos and weapon/model names. Keep the rewritten query visible and editable.
The Q8 conversion has been smoke-tested, not benchmarked for broad retrieval quality. Quality statistics from the source checkpoint are not presented as measured results for this quantized release. No universal GPU/performance claim.
Original full-precision weights, LoRA adapter and training data are not part of this release. No third-party audio files or user library indexes are included.

Licence

The released fine-tuned model and repository material are licensed under Apache License 2.0, retaining the underlying Qwen licence. See LICENSE and NOTICE. Redistribution, modification and commercial use are permitted subject to that licence. This grants no rights to unrelated audio libraries, third-party models, trademarks or the proprietary CyMetaSound plugin.

Downloads last month: 9

GGUF

Model size

2B params

Architecture

qwen2

Hardware compatibility

8-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for cyberalien/CySearchAssist-CLAP-1.5B

Base model

Qwen/Qwen2.5-1.5B

Finetuned

Qwen/Qwen2.5-1.5B-Instruct

Quantized

(295)

this model