add gemini agent

.github/workflows/main.yaml

@@ -18,4 +18,5 @@ jobs:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
           HF_USERNAME: altozachmo
           SPACE_NAME: Final_Assignment_Template
+          GEMINI_API_KEY: ${{ secrets.GOOGLE_AI_STUDIO_API_KEY }}
         run: git push https://$HF_USERNAME:[email protected]/spaces/$HF_USERNAME/$SPACE_NAME main

agents/agent.py

@@ -11,10 +11,11 @@ class MyAgent:
     def __init__(
         self,
         provider: str = "litellm",
-        model_id: str = "gemma3:12b-it-qat",
+        model_id: str = "ollama_chat/gemma3:12b-it-qat",
         api_base: str | None = None,
         api_key: str | None = None,
         planning_interval: int = 3,
+        num_ctx: int = 8192,
     ):
         """
         Initializes the agent depending on the provider and model ID.
@@ -29,11 +30,14 @@ class MyAgent:
         self.api_base = api_base
         self.api_key = api_key
         self.planning_interval = planning_interval
+        self.num_ctx = num_ctx
 
         model = LiteLLMModel(
-            model_id="ollama_chat/gemma3:12b-it-qat",
-            api_base="http://localhost:11434",
-            num_ctx=8196, # ollama default is 2048 which will fail horribly. 8192 works for easy tasks, more is better. Check https://huggingface.co/spaces/NyxKrage/LLM-Model-VRAM-Calculator to calculate how much VRAM this will need for the selected model.
+            model_id=self.model_id,
+            api_base=self.api_base,
+            api_key=self.api_key,
+            num_ctx=self.num_ctx,
+            add_base_tools=True,
         )
 
         tools = [

app.py

@@ -30,10 +30,10 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     try:
         agent = MyAgent(
             provider="litellm",
-            model_id="gemma3:12b-it-qat",
-            api_base="http://localhost:11434",
-            api_key=None,
+            model_id="gemini/gemini-2.0-flash-lite",
+            api_key=os.getenv("GEMINI_API_KEY"),
             planning_interval=3,
+            num_ctx=8192,
         )
 
     except Exception as e:

requirements.txt

@@ -1,2 +1,275 @@
-gradio
-requests
+# This file was autogenerated by uv via the following command:
+#    uv pip compile pyproject.toml -o requirements.txt
+aiofiles==24.1.0
+    # via gradio
+aiohappyeyeballs==2.6.1
+    # via aiohttp
+aiohttp==3.11.18
+    # via litellm
+aiosignal==1.3.2
+    # via aiohttp
+annotated-types==0.7.0
+    # via pydantic
+anyio==4.9.0
+    # via
+    #   gradio
+    #   httpx
+    #   openai
+    #   starlette
+attrs==25.3.0
+    # via
+    #   aiohttp
+    #   jsonschema
+    #   referencing
+authlib==1.5.2
+    # via gradio
+beautifulsoup4==4.13.4
+    # via markdownify
+certifi==2025.4.26
+    # via
+    #   httpcore
+    #   httpx
+    #   requests
+cffi==1.17.1
+    # via cryptography
+charset-normalizer==3.4.1
+    # via requests
+click==8.1.8
+    # via
+    #   duckduckgo-search
+    #   litellm
+    #   typer
+    #   uvicorn
+cryptography==44.0.2
+    # via authlib
+distro==1.9.0
+    # via openai
+duckduckgo-search==8.0.1
+    # via smolagents
+fastapi==0.115.12
+    # via gradio
+ffmpy==0.5.0
+    # via gradio
+filelock==3.18.0
+    # via huggingface-hub
+frozenlist==1.6.0
+    # via
+    #   aiohttp
+    #   aiosignal
+fsspec==2025.3.2
+    # via
+    #   gradio-client
+    #   huggingface-hub
+gradio==5.27.0
+    # via hf-agents-gaia-agent (pyproject.toml)
+gradio-client==1.9.0
+    # via gradio
+groovy==0.1.2
+    # via gradio
+h11==0.16.0
+    # via
+    #   httpcore
+    #   uvicorn
+httpcore==1.0.9
+    # via httpx
+httpx==0.28.1
+    # via
+    #   gradio
+    #   gradio-client
+    #   litellm
+    #   openai
+    #   safehttpx
+huggingface-hub==0.30.2
+    # via
+    #   gradio
+    #   gradio-client
+    #   smolagents
+    #   tokenizers
+idna==3.10
+    # via
+    #   anyio
+    #   httpx
+    #   requests
+    #   yarl
+importlib-metadata==8.7.0
+    # via litellm
+itsdangerous==2.2.0
+    # via gradio
+jinja2==3.1.6
+    # via
+    #   gradio
+    #   litellm
+    #   smolagents
+jiter==0.9.0
+    # via openai
+jsonschema==4.23.0
+    # via litellm
+jsonschema-specifications==2025.4.1
+    # via jsonschema
+litellm==1.67.1
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   smolagents
+lxml==5.4.0
+    # via duckduckgo-search
+markdown-it-py==3.0.0
+    # via rich
+markdownify==1.1.0
+    # via smolagents
+markupsafe==3.0.2
+    # via
+    #   gradio
+    #   jinja2
+mdurl==0.1.2
+    # via markdown-it-py
+multidict==6.4.3
+    # via
+    #   aiohttp
+    #   yarl
+numpy==2.2.5
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   gradio
+    #   pandas
+openai==1.76.0
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   litellm
+orjson==3.10.16
+    # via gradio
+packaging==25.0
+    # via
+    #   gradio
+    #   gradio-client
+    #   huggingface-hub
+pandas==2.2.3
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   gradio
+pillow==11.2.1
+    # via
+    #   gradio
+    #   smolagents
+primp==0.15.0
+    # via duckduckgo-search
+propcache==0.3.1
+    # via
+    #   aiohttp
+    #   yarl
+pycparser==2.22
+    # via cffi
+pydantic==2.11.3
+    # via
+    #   fastapi
+    #   gradio
+    #   litellm
+    #   openai
+pydantic-core==2.33.1
+    # via pydantic
+pydub==0.25.1
+    # via gradio
+pygments==2.19.1
+    # via rich
+python-dateutil==2.9.0.post0
+    # via pandas
+python-dotenv==1.1.0
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   litellm
+    #   smolagents
+python-multipart==0.0.20
+    # via gradio
+pytz==2025.2
+    # via pandas
+pyyaml==6.0.2
+    # via
+    #   gradio
+    #   huggingface-hub
+referencing==0.36.2
+    # via
+    #   jsonschema
+    #   jsonschema-specifications
+regex==2024.11.6
+    # via tiktoken
+requests==2.32.3
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   huggingface-hub
+    #   smolagents
+    #   tiktoken
+    #   wikipedia-api
+rich==14.0.0
+    # via
+    #   smolagents
+    #   typer
+rpds-py==0.24.0
+    # via
+    #   jsonschema
+    #   referencing
+ruff==0.11.7
+    # via gradio
+safehttpx==0.1.6
+    # via gradio
+semantic-version==2.10.0
+    # via gradio
+shellingham==1.5.4
+    # via typer
+six==1.17.0
+    # via
+    #   markdownify
+    #   python-dateutil
+smolagents==1.14.0
+    # via hf-agents-gaia-agent (pyproject.toml)
+sniffio==1.3.1
+    # via
+    #   anyio
+    #   openai
+soupsieve==2.7
+    # via beautifulsoup4
+starlette==0.46.2
+    # via
+    #   fastapi
+    #   gradio
+tiktoken==0.9.0
+    # via litellm
+tokenizers==0.21.1
+    # via litellm
+tomlkit==0.13.2
+    # via gradio
+tqdm==4.67.1
+    # via
+    #   hf-agents-gaia-agent (pyproject.toml)
+    #   huggingface-hub
+    #   openai
+typer==0.15.2
+    # via gradio
+typing-extensions==4.13.2
+    # via
+    #   anyio
+    #   beautifulsoup4
+    #   fastapi
+    #   gradio
+    #   gradio-client
+    #   huggingface-hub
+    #   openai
+    #   pydantic
+    #   pydantic-core
+    #   referencing
+    #   typer
+    #   typing-inspection
+typing-inspection==0.4.0
+    # via pydantic
+tzdata==2025.2
+    # via pandas
+urllib3==2.4.0
+    # via requests
+uvicorn==0.34.2
+    # via gradio
+websockets==15.0.1
+    # via gradio-client
+wikipedia-api==0.8.1
+    # via hf-agents-gaia-agent (pyproject.toml)
+yarl==1.20.0
+    # via aiohttp
+zipp==3.21.0
+    # via importlib-metadata

run_local_agent.py

@@ -7,20 +7,23 @@ from dotenv import load_dotenv
 load_dotenv()
 
 QUESTIONS_FILEPATH: str = os.getenv("QUESTIONS_FILEPATH", default="metadata.jsonl")
+OLLAMA_MODEL_ID: str = os.getenv("OLLAMA_MODEL_ID", default="gemma3:12b-it-qat")
 OLLAMA_API_BASE: str = os.getenv("OLLAMA_API_BASE", default="http://localhost:11434")
-OLLAMA_API_KEY: str | None = os.getenv("OLLAMA_API_KEY")
+OLLAMA_API_KEY: str | None = os.getenv("GOOGLE_AI_STUDIO_API_KEY")
+OLLAMA_NUM_CTX: int = int(os.getenv("OLLAMA_NUM_CTX", default=8192))
 
+myagent_args = {
+    "provider": "litellm",
+    "model_id": "gemini/gemini-2.0-flash-lite",
+    # "api_base": OLLAMA_API_BASE,
+    "planning_interval": 3,
+    "num_ctx": OLLAMA_NUM_CTX,
+}
 
-print(f"Using OLLAMA API base: {OLLAMA_API_BASE}")
+print(f"Using args: {myagent_args}")
 
 if __name__ == "__main__":
-    agent = MyAgent(
-        provider="litellm",
-        model_id="gemma3:12b-it-qat",
-        api_base=OLLAMA_API_BASE,
-        api_key=OLLAMA_API_KEY,
-        planning_interval=3,
-    )
+    agent = MyAgent(**myagent_args)
 
     with open(QUESTIONS_FILEPATH, "r") as f:
         questions = json.load(f)

test.py

@@ -2,7 +2,7 @@ from smolagents import LiteLLMModel, OpenAIServerModel
 from dotenv import load_dotenv
 load_dotenv()
 
-model_id = "ollama_chat/gemma3:12b-it-qat"
+model_id = "ollama_chat/mistral-small3.1:latest"
 api_base = "http://localhost:11434/v1"
 
 # model = LiteLLMModel(