AgenticSeek: Local AI Assistant Powered by Deepseek R1 Agents.
A fully local alternative to Manus AI, a voice-enabled AI assistant that codes, explores your filesystem, and correct it's mistakes all without sending a byte of data to the cloud. The goal of the project is to create a truly Jarvis like assistant using reasoning model such as deepseek R1.
🛠️ Work in Progress – Looking for contributors! 🚀
Features:
- Privacy-first: Runs 100% locally – no data leaves your machine
- ️Voice-enabled: Speak and interact naturally
- Filesystem interaction: Use bash to interact with your filesystem.
- Coding abilities: Code in Python, C, Golang, and soon more
- Trial-and-error: If a command or code fails, the assistant retries to fixes it automatically, saving you time.
- Agent routing: Select the best agent for the task.
- Multi-agent planning: For complex tasks, divide and conquer with multiple agents
- Tools:: All agents have their respective tools ability. Basic search, flight API, files explorer, etc...
- Web browsing (Not implemented yet): Browse the web autonomously to conduct task.
- Memory: Retain only useful information, recover conversation session, remember your preferences.
Run locally
We recommend using at least Deepseek 14B, smaller models struggle with tool use and forget quickly the context.
1️⃣ Install Dependencies
pip3 install -r requirements.txt
2️⃣ Download Models
Make sure you have Ollama installed.
Download the deepseek-r1:7b model from DeepSeek
ollama pull deepseek-r1:7b
3️⃣ Run the Assistant (Ollama)
Start the ollama server
ollama serve
Change the config.ini file to set the provider_name to ollama and provider_model to deepseek-r1:7b
[MAIN]
is_local = True
provider_name = ollama
provider_model = deepseek-r1:7b
Run the assistant:
python3 main.py
Alternative: Run the LLM on your own server
1️⃣ Set up and start the server scripts
On your "server" that will run the AI model, get the ip address
ip a | grep "inet " | grep -v 127.0.0.1 | awk '{print $2}' | cut -d/ -f1
Clone the repository and then, run the script stream_llm.py in server/
python3 stream_llm.py
2️⃣ Run it
Now on your personal computer:
Clone the repository.
Change the config.ini file to set the provider_name to server and provider_model to deepseek-r1:7b.
Set the provider_server_address to the ip address of the machine that will run the model.
[MAIN]
is_local = False
provider_name = server
provider_model = deepseek-r1:14b
provider_server_address = x.x.x.x:5000
Run the assistant:
python3 main.py
Providers
The table below show the available providers:
| Provider | Local? | Description |
|---|---|---|
| Ollama | Yes | Run LLMs locally with ease using ollama as a LLM provider |
| Server | Yes | Host the model on another machine, run your local machine |
| OpenAI | No | Use ChatGPT API (non-private) |
| Deepseek | No | Deepseek API (non-private) |
| HuggingFace | No | Hugging-Face API (non-private) |
To select a provider change the config.ini:
is_local = False
provider_name = openai
provider_model = gpt-4o
provider_server_address = 127.0.0.1:5000
is_local: should be True for any locally running LLM, otherwise False.
provider_name: Select the provider to use by its name, see the provider list above.
provider_model: Set the model to use by the agent.
provider_server_address: can be set to anything if you are not using the server provider.
FAQ
Q: What hardware do I need?
For Deepseek R1 7B, we recommend a GPU with with 8GB VRAM. The 14B model can run on 12GB GPU like the rtx 3060. The 32B model needs a GPU with 24GB+ VRAM.
Q: Why Deepseek R1 over other models?
Deepseek R1 excels at reasoning and tool use for its size. We think it’s a solid fit for our needs—other models work fine, but Deepseek is our primary pick.
Q: I get an error running main.py. What do I do?
Ensure Ollama is running (ollama serve), your config.ini matches your provider, and dependencies are installed. If none work feel free to raise an issue.
Q: Can it really run 100% locally?
Yes with Ollama or Server providers, all speech to text, LLM and text to speech model run locally. Non-local options (OpenAI, Deepseek API) are optional.
Q: How come it is older than manus ?
we started this a fun side project to make a fully local, Jarvis-like AI. However, with the rise of Manus and openManus, we saw the opportunity to redirected some tasks priority to make yet another alternative.
Q: How is it better than manus or openManus ?
It's not, never will be, we just offer an alternative that is more local and enjoyable to use.
Current contributor:
Fosowl 🇫🇷 steveh8758 🇹🇼