Update Open Deep Research's README.md #763
Conversation
| ## Notes | ||
|
|
||
| - Currently, this implementation only supports **OpenAI** models. | ||
| - Only **reasoning models** like `o1` are supported at this time. If you attempt to use a non-reasoning model, you may need to modify the script. |
There was a problem hiding this comment.
@keetrap if this is true we should fix it to work with any model (maybe just remove the reasoning_effort argument if it's passed, because it did not impact results)
There was a problem hiding this comment.
@aymeric-roucher Thanks for the review.
I am going to remove the Notes section.
Should we keep reasoning_effort as an command-line argument or remove it completely? What do you think?
Co-authored-by: Aymeric Roucher <69208727+aymeric-roucher@users.noreply.github.com>
93ecda3 to
64228b2
Compare
| - **SerperAPI Key**: Open Deep Research uses SerperAPI for web browsing functionality. | ||
| ```bash | ||
| export SERPER_API_KEY="your_serper_api_key_here" | ||
| ``` |
There was a problem hiding this comment.
I have some confusion regarding this because
Currently, we are passing the SERPAPI_API_KEY in the BROWSER_CONFIG as follows:
BROWSER_CONFIG = {
"viewport_size": 1024 * 5,
"downloads_folder": "downloads_folder",
"request_kwargs": {
"headers": {"User-Agent": user_agent},
"timeout": 300,
},
"serpapi_key": os.getenv("SERPAPI_API_KEY"),
}However, in the WEB_TOOLS, the GoogleSearchTool is using the serper provider, which requires the SERPER_API_KEY:
GoogleSearchTool(provider="serper")This requires users to provide two different API keys. Is this intentional, or should we be using the same provider throughout?
|
@aymeric-roucher @albertvillanova |
Closes #760