A fork of AgentLab that provides the
GenericAgent and AGENT_* configs used to run the Knows benchmark
(Google Workspace Docs / Sheets / Slides tasks for browser agents).
git clone https://github.com/farhanishmam/AgentLab-Knows
cd AgentLab-Knows
pip install -e .
Use AgentLab's GenericAgent with one of the prebuilt AGENT_* configs to run
the Knows benchmark:
from agentlab.agents.generic_agent import AGENT_GPT55_AXT # GenericAgent config
from agentlab.experiments.study import Study
study = Study([AGENT_GPT55_AXT], benchmark="knows_docs_1")
study.run()
Pick the AGENT_* config that matches your model and observation mode (e.g.
AGENT_GPT55_AXT, AGENT_GEMINI_31_PRO_SCREENSHOT, AGENT_DEEPSEEK_V4_PRO_BOTH).
All configs are defined in
src/agentlab/agents/generic_agent/agent_configs.py.
See README_AgentLab.md for the full upstream AgentLab docs.
Python
75.4%
Jupyter Notebook
21.5%
HTML
2.8%
A fork of AgentLab that provides the
GenericAgent and AGENT_* configs used to run the Knows benchmark
(Google Workspace Docs / Sheets / Slides tasks for browser agents).
git clone https://github.com/farhanishmam/AgentLab-Knows
cd AgentLab-Knows
pip install -e .
Use AgentLab's GenericAgent with one of the prebuilt AGENT_* configs to run
the Knows benchmark:
from agentlab.agents.generic_agent import AGENT_GPT55_AXT # GenericAgent config
from agentlab.experiments.study import Study
study = Study([AGENT_GPT55_AXT], benchmark="knows_docs_1")
study.run()
Pick the AGENT_* config that matches your model and observation mode (e.g.
AGENT_GPT55_AXT, AGENT_GEMINI_31_PRO_SCREENSHOT, AGENT_DEEPSEEK_V4_PRO_BOTH).
All configs are defined in
src/agentlab/agents/generic_agent/agent_configs.py.
See README_AgentLab.md for the full upstream AgentLab docs.
Python
75.4%
Jupyter Notebook
21.5%
HTML
2.8%