Fix agent.run, fallbacks somehow dont work

.python-version

@@ -0,0 +1 @@
+3.12

.ruff.toml

@@ -0,0 +1,20 @@
+# Assume Python 3.12.
+target-version = "py312"
+# Decrease the maximum line length to 79 characters.
+line-length = 300
+
+# pyflakes, pycodestyle, isort
+# flake8 YTT, pydocstyle D, pylint PLC
+lint.select = ["F", "E", "W", "I001", "YTT", "D", "PLC"]
+# select = ["ALL"]
+
+# D100 Missing docstring in public module
+# D101 Missing docstring in public class
+# D102 D102 Missing docstring in public method
+# D103 Missing docstring in public function
+# D107 Missing docstring in `__init__`
+# `multi-line-summary-first-line` (D212)
+# `one-blank-line-before-class` (D203)
+lint.extend-ignore = ["D100", "D101", "D102", "D103", "D107", "D212", "D203"]
+
+exclude = [".venv"]

basic_agent.py

@@ -9,9 +9,13 @@ import pandas as pd
 import requests
 import rich
 import smolagents
-from get_model import get_model
+import wikipediaapi
 from loguru import logger
-from smolagents import CodeAgent, DuckDuckGoSearchTool, FinalAnswerTool, HfApiModel, VisitWebpageTool
+from smolagents import CodeAgent, DuckDuckGoSearchTool, FinalAnswerTool, HfApiModel, Tool, VisitWebpageTool
+
+from get_model import get_model
+from litellm_model import litellm_model
+from openai_model import openai_model
 
 print = rich.get_console().print
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
@@ -45,10 +49,33 @@ AUTHORIZED_IMPORTS = [
 ]
 
 
+class WikipediaSearchTool(Tool):
+    name = "wikipedia_search"
+    description = "Fetches a summary of a Wikipedia page based on a given search query (only one word or group of words)."
+
+    inputs = {
+        "query": {"type": "string", "description": "The search term for the Wikipedia page (only one word or group of words)."}
+    }
+
+    output_type = "string"
+
+    def __init__(self, lang="en"):
+        super().__init__()
+        self.wiki = wikipediaapi.Wikipedia(
+            language=lang, user_agent="MinimalAgent/1.0")
+
+    def forward(self, query: str):
+        page = self.wiki.page(query)
+        if not page.exists():
+            return "No Wikipedia page found."
+        return page.summary[:1000]
+
+
 @dataclass
 class BasicAgent:
     model: smolagents.models.Model = HfApiModel()
     tools: list = field(default_factory=lambda: [])
+    verbosity_level: int = 0
     # def __init__(self):
     def __post_init__(self):
         """Run post_init."""
@@ -56,7 +83,7 @@ class BasicAgent:
         self.agent = CodeAgent(
             tools=self.tools,
             model=self.model,
-            verbosity_level=3,
+            verbosity_level=self.verbosity_level,
             additional_authorized_imports=AUTHORIZED_IMPORTS,
             planning_interval=4,
         )
@@ -66,17 +93,17 @@ class BasicAgent:
 
     def __call__(self, question: str) -> str:
         # print(f"Agent received question (first 50 chars): {question[:50]}...")
-        print(f"Agent received question: {question}...")
+        # print(f"Agent received question: {question}...")
 
         # fixed_answer = "This is a default answer."
         # print(f"Agent returning fixed answer: {fixed_answer}")
         # return fixed_answer
         try:
             # answer = self.get_answer(question)
-            answer = self.agent(question)
+            answer = self.agent.run(question)
         except Exception as e:
             logger.error(e)
-            answer = str(e)[:10] + "..."
+            answer = str(e)[:110] + "..."
 
         return answer
 
@@ -93,7 +120,28 @@ def main():
 
     space_id = f"{username}/{repo_name}"
 
-    model = get_model(cat="gemini")
+    # model = get_model(cat="gemini")
+
+    _ = (
+        "gemini-2.5-flash-preview-04-17",
+        # "https://api-proxy.me/gemini/v1beta",
+        "https://generativelanguage.googleapis.com/v1beta",
+        os.getenv("GEMINI_API_KEY"),
+    )
+
+    _ = (
+        "grok-3-beta",
+        "https://api.x.ai/v1",
+        os.getenv("XAI_API_KEY"),
+    )
+
+    # model = litellm_model(*_)
+    model = openai_model(*_)
+
+    messages = [{'role': 'user', 'content': 'Say this is a test.'}]
+    print(model(messages))
+    # raise SystemExit("By intention")
+
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
         # agent = BasicAgent()
@@ -102,6 +150,7 @@ def main():
             tools=[
                 DuckDuckGoSearchTool(),
                 VisitWebpageTool(),
+                WikipediaSearchTool(),
                 FinalAnswerTool(),
             ]
         )
@@ -142,7 +191,8 @@ def main():
     print(f"Running agent on {len(questions_data)} questions...")
 
     # for item in questions_data:
-    for item in questions_data[-1:]:
+    # for item in questions_data[-1:]:
+    for item in questions_data[14:15]:
         task_id = item.get("task_id")
         question_text = item.get("question")
         if not task_id or question_text is None:

exit_gracefully.py

@@ -0,0 +1,11 @@
+# pylint: disable=unused-argument,missing-module-docstring,missing-function-docstring
+import signal
+import sys
+
+
+def exit_gracefully(signum, frame):
+    print("Exiting...")
+    sys.exit(0)
+
+
+signal.signal(signal.SIGINT, exit_gracefully)

get_gemini_keys.py

@@ -5,10 +5,15 @@ from pathlib import Path
 import rich
 import yaml
 from dotenv import dotenv_values
+from loguru import logger
 
 
 def get_gemini_keys(file=r".env-gemini", dotenv=False):
     """Get gemini keys."""
+    if not Path(file).exists():
+        logger.debug(f"{file} does not exit, returing [] ")
+        return []
+
     if Path(file).name.startswith(".env"):
         dotenv = True
 
@@ -19,7 +24,7 @@ def get_gemini_keys(file=r".env-gemini", dotenv=False):
         try:
             keys = yaml.load(dotenv_values(file).get("GEMINI_API_KEYS"), yaml.Loader)
         except Exception as e:
-            print(e)
+            logger.error(e)
             return []
         return keys
 

get_model.py

@@ -25,7 +25,7 @@ def get_model(cat: str = "hf", provider=None, model_id=None):
         # get gemini_api_keys
         # dedup
         _ = re.findall(r"AIzaSy[A-Z][\w-]{32}", os.getenv("GEMINI_API_KEYS", ""))
-        gemini_api_keys = dict.fromkeys(get_gemini_keys() + _)
+        gemini_api_keys = [*dict.fromkeys(get_gemini_keys() + _)]
 
         # assert gemini_api_keys, "No GEMINI_API_KEYS, set env var GEMINI_API_KEYS or put them in .env-gemini and try again."
         if not gemini_api_keys:
@@ -67,23 +67,37 @@ def get_model(cat: str = "hf", provider=None, model_id=None):
                     "api_key": os.getenv("SILICONFLOW_API_KEY"),
                     "api_base": "https://api.siliconflow.cn/v1",
                 },
-            }
+            },
         ]
-
+        
+        # gemma-3-27b-it
+        llm_loadbalancer_model_list_gemma = [
+            {
+                "model_name": "model-group-3",
+                "litellm_params": {
+                    "model": f"gemini/gemma-3-27b-it",
+                    "api_key": os.getenv("GEMINI_API_KEY")                },
+            },
+        ]
+            
         fallbacks = []
         model_list = llm_loadbalancer_model_list_gemini
         if os.getenv("SILICONFLOW_API_KEY"):
             fallbacks = [{"model-group-1": "model-group-2"}]
             model_list += llm_loadbalancer_model_list_siliconflow
-
+        
+        model_list += llm_loadbalancer_model_list_gemma
+        fallbacks13 = [{"model-group-1": "model-group-3"}]
+        fallbacks31 = [{"model-group-3": "model-group-1"}]
+        
         model = LiteLLMRouterModel(
             model_id="model-group-1",
             model_list=model_list,
             client_kwargs={
                 "routing_strategy": "simple-shuffle",
                 "num_retries": 3,
-                # "retry_after": 130,   # waits min  s before retrying request
-                "fallbacks": fallbacks,
+                "retry_after": 180,   # waits min  s before retrying request
+                "fallbacks": fallbacks13,  # falllacks dont seem to work
             },
         )
 

litellm_model.py

@@ -0,0 +1,89 @@
+# ruff: noqa: F401
+
+import os
+import sys
+
+import rich
+from loguru import logger
+from smolagents import LiteLLMModel, OpenAIServerModel
+
+from exit_gracefully import exit_gracefully
+
+print = rich.get_console().print  # noqa
+
+
+def litellm_model(
+    model_id = None,
+    api_base = None,
+    api_key = None,
+    provider = "openai",
+    kwargs = None,
+):
+    kwargs = kwargs or {}
+    # default siliconflow
+    api_base = api_base or "https://api.siliconflow.cn/v1"
+    api_key = api_key or os.getenv("SILICONFLOW_API_KEY")
+    model_id = model_id or "deepseek-ai/DeepSeek-V3"
+
+    if provider.lower() in ["gemini", "xai"]:
+        return LiteLLMModel(
+            f"{provider}/{model_id}",
+            # api_base=api_base,  # must not set api_base to None or not set at all for gemini
+            api_key=api_key,
+            temperature=0.,
+            **kwargs,
+        )
+    else:
+        return LiteLLMModel(
+            f"{provider}/{model_id}",
+            api_base=api_base,
+            api_key=api_key,
+            temperature=0.,
+            **kwargs,
+        )
+
+
+def main():
+    messages = [{'role': 'user', 'content': 'Say this is a test.'}]
+    logger.debug(sys.argv)
+    if not sys.argv[1:]:
+        model = litellm_model()
+        logger.debug(model(messages))
+        return
+
+    if len(sys.argv[1:]) < 4:
+        raise SystemExit("Provide at least four args (model_id, api_base, api_key, provider)")
+
+    model_id, api_base, api_key, provider, *_ = sys.argv[1:]
+    model = litellm_model(model_id, api_base, api_key, provider)
+    try:
+        response = model(messages)
+        logger.debug(response)
+    except Exception as e:
+        logger.error(e)
+        return
+    try:
+        print(response.content)
+    except Exception as e:
+        logger.error(e)
+
+
+if __name__ == "__main__":
+    main()
+    # python litellm_model.py
+    #  deepseek-ai/DeepSeek-V3 https://api.siliconflow.cn/v1 %SILICONFLOW_API_KEY%
+
+    # python litellm_model.py  grok-3-beta https://api.x.ai/v1 %XAI_API_KEY%
+
+    # set proxy or use https://api-proxy.me/gemini/v1beta
+    #   gemini-2.5-flash-preview-04-17  https://generativelanguage.googleapis.com/v1beta %GEMINI_API_KEY%
+    # python litellm_model.py  gemini-2.5-flash-preview-04-17  https://api-proxy.me/gemini/v1beta %GEMINI_API_KEY%
+    #   gemini-2.0-flash
+    # litellmmodel
+
+    # https://api.together.ai/models/deepseek-ai/DeepSeek-V3
+    # deepseek-ai/DeepSeek-V3 https://api.together.xyz/v1 %TOGETHER_API_KEY%
+
+    #  deepseek-chat  https://litellm.dattw.eu.org/v1 %LITELLM_API_KEY%
+
+    # LLM API proxy: https://linux.do/t/topic/290871

main.py

@@ -0,0 +1,6 @@
+def main():
+    print("Hello from final-assignment-hf-agents-course!")
+
+
+if __name__ == "__main__":
+    main()

openai_model.py

@@ -0,0 +1,76 @@
+# ruff: noqa: F401
+
+import os
+import sys
+
+from exit_gracefully import exit_gracefully
+from loguru import logger
+from smolagents import LiteLLMModel, OpenAIServerModel
+import rich
+
+print = rich.get_console().print  # noqa
+
+
+def openai_model(
+    model_id = None,
+    api_base = None,
+    api_key = None,
+    kwargs = None,
+):
+    kwargs = kwargs or {}
+    # default siliconflow
+    api_base = api_base or "https://api.siliconflow.cn/v1"
+    api_key = api_key or os.getenv("SILICONFLOW_API_KEY")
+    model_id = model_id or "deepseek-ai/DeepSeek-V3"
+
+    return OpenAIServerModel(
+        model_id,
+        api_base=api_base,
+        api_key=api_key,
+        temperature=0.,
+        **kwargs,
+    )
+
+def main():
+    messages = [{'role': 'user', 'content': 'Say this is a test.'}]
+    logger.debug(sys.argv)
+    if not sys.argv[1:]:
+        model = openai_model()
+        logger.debug(model(messages))
+        return
+
+    if len(sys.argv[1:]) < 3:
+        raise SystemExit("Provide at least three args (model_id, api_base, api_key)")
+
+    model_id, api_base, api_key, *_ = sys.argv[1:]
+    model = openai_model(model_id, api_base, api_key)
+    try:
+        response = model(messages)
+        logger.debug(response)
+    except Exception as e:
+        logger.error(e)
+        return
+    try:
+        print(response.content)
+    except Exception as e:
+        logger.error(e)
+
+
+if __name__ == "__main__":
+    main()
+    # python openai_model.py
+    #  deepseek-ai/DeepSeek-V3 https://api.siliconflow.cn/v1 %SILICONFLOW_API_KEY%
+
+    # python openai_model.py  grok-3-beta https://api.x.ai/v1 %XAI_API_KEY%
+
+    #   gemini-2.5-flash-preview-04-17  https://generativelanguage.googleapis.com/v1beta %GEMINI_API_KEY%
+    #   gemini-2.0-flash
+
+    # https://api.together.ai/models/deepseek-ai/DeepSeek-V3
+    # deepseek-ai/DeepSeek-V3 https://api.together.xyz/v1 %TOGETHER_API_KEY%
+
+    #  deepseek-chat  https://litellm.dattw.eu.org/v1 %LITELLM_API_KEY%
+
+    #
+
+    # LLM API proxy: https://linux.do/t/topic/290871

tasks.py

@@ -0,0 +1,51 @@
+"""
+Define invoke tasks.
+
+invoke list
+invoke build
+
+invoke --help build
+
+"""
+from invoke import task
+
+NAME_DEF = "final assignment tasks"
+
+@task(
+    default=True,
+)
+def gemini(c):
+    """Run python litellm_model.py gemini params."""
+    c.run("python litellm_model.py gemini-2.5-flash-preview-04-17  https://generativelanguage.googleapis.com/v1beta %GEMINI_API_KEY% gemini")
+
+@task
+def gemini1(c):
+    """Run python litellm_model.py gemini-as-openai params."""
+    c.run("python litellm_model.py gemini-2.5-flash-preview-04-17  https://generativelanguage.googleapis.com/v1beta %GEMINI_API_KEY% openai")
+
+@task
+def gemini2(c):
+    """Run python litellm_model.py gemini-via-proxy params."""
+    # c.run("python litellm_model.py gemini-2.5-flash-preview-04-17  https://generativelanguage.googleapis.com/v1beta %GEMINI_API_KEY%")
+    c.run("python litellm_model.py gemini-2.5-flash-preview-04-17  https://api-proxy.me/gemini/v1beta %GEMINI_API_KEY% openai")
+
+@task
+def xai(c):
+    """Run python litellm_model.py grok params."""
+    c.run("python litellm_model.py grok-3-beta https://api.x.ai/v1 %XAI_API_KEY% xai")
+
+@task
+def grok1(c):
+    """Run python litellm_model.py grok params."""
+    c.run("python litellm_model.py grok-3-beta https://api.x.ai/v1 %XAI_API_KEY% openai")
+
+@task
+def siliconflow(c):
+    """Run python litellm_model.py (siliconflow setup)."""
+    c.run("python litellm_model.py ")
+
+
+@task
+def agent(c):
+    """Run python basic_agent.py."""
+    c.run("python basic_agent.py ")

ty.toml

@@ -0,0 +1,2 @@
+[rules]
+unresolved-import = "ignore"

uv.lock

@@ -0,0 +1,8 @@
+version = 1
+revision = 2
+requires-python = ">=3.12"
+
+[[package]]
+name = "final-assignment-hf-agents-course"
+version = "0.1.0"
+source = { virtual = "." }