tri.oo.ai is a local, open-source showcase demonstrating the capabilities of Tri-7B-Search-preview, an efficient LLM with built-in internet search. Developed by Trillion Labs and based on Tri-7B ππ, it features a familiar chat interface inspired by oo.ai.
tri.oo.ai replicates the functionality of oo.ai with a local-first approach and open tooling. It uses the Tri-7B-Search model as the engine behind a smart assistant that can:
- Understand natural queries
- Retrieve relevant data from your knowledge base
- Augment answers with DuckDuckGo search results
- Run locally for full control and privacy
This is a search-enabled LLM service β combining the reasoning of LLMs with the freshness of web search.
νμ΄μ»€ μ¬κ³μ½ μμ
νμ΄μ»€(μ΄μν)λ 2029λ κΉμ§ T1κ³Ό 4λ μ¬κ³μ½μ 체결ν μ¬μ€μ΄ νμΈλμμ΅λλ€. μ΄λ 2025λ 7μ 27μΌ(2025λ κΈ°μ€) 곡μ λ°νλ λ΄μ©μΌλ‘, 리그 μ€λΈ λ μ λ(LoL) eμ€ν¬μΈ μμ¬μ μ΅μ΄λ‘ 18μμ¦ μ°μ νλμ μ΄μ΄κ°κ² λμμ΅λλ€. μ£Όμ κ³μ½ 쑰건μ λ€μκ³Ό κ°μ΅λλ€:
- κ³μ½ κΈ°κ°: 2025λ 7μ 27μΌλΆν° 2029λ 6μ 30μΌκΉμ§ (4λ κ°)
- μ±κ³Ό: ν΄λΉ κΈ°κ° λμ 2022 LCK μ€νλ§ μ μΉ μ°μΉ, 2023 μμμκ²μ κΈλ©λ¬, 2023Β·2024 LoL μλ μ±νΌμΈμ 2μ°ν¨ λ± μλ μ΅κ³ μμ€μ μ±μ μ κΈ°λ‘νμ΅λλ€.
- μλ―Έ: νμ΄μ»€λ 2013λ λ°λ· μ΄λ νμ μμ§μΌλ‘ νμ½νλ©° eμ€ν¬μΈ μμ¬μ ν νμ κ·Έμ μ μλ‘ νκ°λ°κ³ μμ΅λλ€.
μΆμ²
- Clone the repository:
git clone https://github.com/trillion-labs/tri.oo.ai.git
cd tri.oo.ai
- Install UV (if not already installed):
curl -LsSf https://astral.sh/uv/install.sh | sh
- Create virtual environment and install dependencies:
uv sync
# for offline vllm usage
uv sync --group vllm
Run the interactive assistant with direct vLLM integration:
uv run main.py
Command line options:
--model
: Model name or path (default: "trillionlabs/Tri-7B-Search-preview")--tensor-parallel-size
: Number of GPUs for tensor parallelism (default: 1)--gpu-memory-utilization
: GPU memory utilization 0-1 (default: 0.9)
Example:
uv run main.py --tensor-parallel-size 2 --gpu-memory-utilization 0.8
For better flexibility and integration, you can use the OpenAI-compatible API:
- Start the vLLM server:
./serve.sh
This will start a vLLM server at http://localhost:8000
with the Tri-7B-Search-preview model.
- In a new terminal, run the API client:
uv run main_api.py
Command line options for the API client:
--base-url
: Base URL for the OpenAI-compatible API (default: "http://localhost:8000/v1")--model
: Model name (default: "trillionlabs/Tri-7B-Search-preview")
Example with custom server:
python main_api.py --base-url "http://your-server:8080/v1"
- Direct vLLM (
main.py
): Best for single-user local usage with minimal setup - API mode (
main_api.py
): Best for multi-user scenarios, integration with other tools, or when you want to separate the model server from the client
This project is licensed under the Apache 2.0 License - see the LICENSE file for details.
- Powered by vLLM for efficient inference
- Web search functionality via DuckDuckGo Search