https://github.com/burnycoder/deep-research-smolagents
https://github.com/burnycoder/deep-research-smolagents
Last synced: about 1 year ago
JSON representation
- Host: GitHub
- URL: https://github.com/burnycoder/deep-research-smolagents
- Owner: BurnyCoder
- Created: 2025-02-13T20:17:35.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-02-13T20:21:04.000Z (over 1 year ago)
- Last Synced: 2025-02-13T21:27:16.850Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 84 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Open Deep Research
This is edited version of Hugginface's Open Deep Research Agent https://huggingface.co/blog/open-deep-research and dzhng's Deep Research agent https://github.com/dzhng/deep-research combined into one agent.
## Environment Variables
Create a `.env` file in the root directory with the following variables. You can use the `.envtemplate` file as a reference.
```bash
HF_TOKEN=your_huggingface_token # Required for accessing Hugging Face models and resources
SERPAPI_API_KEY=your_serpapi_key # Enables web search functionality through SerpAPI
OPENAI_API_KEY=your_openai_key # Required for using OpenAI language models
FIRECRAWL_API_KEY=your_firecrawl_key # Enables web crawling and content extraction
PORTKEY_API_BASE=https://api.portkey.ai/v1 # Base URL for Portkey API service
PORTKEY_API_KEY=your_portkey_key # Main authentication key for Portkey services
PORTKEY_VIRTUAL_KEY_GROQ=your_groq_virtual_key # Virtual key for accessing Groq models via Portkey
PORTKEY_VIRTUAL_KEY_ANTHROPIC=your_anthropic_virtual_key # Virtual key for accessing Anthropic models
PORTKEY_VIRTUAL_KEY_OPENAI=your_openai_virtual_key # Virtual key for accessing OpenAI models through Portkey
PORTKEY_VIRTUAL_KEY_GOOGLE=your_google_virtual_key # Virtual key for accessing Google AI models
OPENAI_MODEL="o3-mini" # Specifies which OpenAI model to use by default
AGENT_TYPE="code" # Determines whether to use tool-based or code-based agent
MANAGER_AGENT_SYSTEM_PROMPT=system_prompt_here # Defines the behavior and capabilities of the manager agent
```
## Installation
Clone the repository:
```bash
git clone https://github.com/BurnyCoder/deep-research-smolagents
cd deep-research-smolagents
```
Optionally create a virtual environment:
```bash
python -m venv venv
source venv/bin/activate
```
Install dependencies:
```bash
pip install -r requirements.txt
npm install
```
## Running the agent
Parameters Explained:
- `--model-id`: Specifies which AI model to use (e.g., "o3-mini", "claude-3-5-sonnet-latest")
- `--questions`: When included, enables interactive clarifying questions mode where the agent will ask questions to better understand your research query
- `--b`: Research breadth parameter (range: 3-10, default: 4)
- `--d`: Research depth parameter (range: 1-5, default: 2)
To run smolagents deep research agent with deep reserach TS agent as a tool with optional parameters:
```bash
python run_deep_research.py --questions --model-id "o3-mini" "Best practices to build AI agents"
```
- If you don't pass query in the parameter, it will prompt you to enter a question.
- If you dont pass --questions in the parameter, it will not ask clarifying questions.
- If you don't pass --model-id in the parameter, it will use o3-mini as the model.
To run deep research TS agent alone with optional parameters:
```bash
python run_deep_research_ts.py --questions --b 2 --d 2 "Best practices to build AI agents"
```
- If you don't pass query in the parameter, it will prompt you to enter a question.
- If you dont pass --questions in the parameter, it will not ask clarifying questions.
- If you don't pass --b and --d (breadth and depth) in the parameter, it will use 2 and 2 as the default values.
To run smolagents agents with serapi (smolagents's original open deep research agent):
```bash
python run_serapi.py --questions --model-id "claude-3-5-sonnet-latest" "Best practices to build AI agents"
```
To run smolagents agents with firecrawl:
```bash
python run_firecrawl.py --questions --model-id "claude-3-5-sonnet-latest" "Best practices to build AI agents"
```