Update app.py from m-ric's repo, using litellm.dattw

app.py

@@ -5,13 +5,15 @@ import re
 import shutil
 import threading
 from typing import Optional
+from loguru import logger
 
 import gradio as gr
 from dotenv import load_dotenv
-from huggingface_hub import login
+# from huggingface_hub import login
 from smolagents import (
     CodeAgent,
     HfApiModel,
+    LiteLLMModel,
     Tool,
     GoogleSearchTool
 )
@@ -36,9 +38,10 @@ from scripts.text_web_browser import (
 from scripts.visual_qa import visualizer
 
 
-web_search = GoogleSearchTool(provider="serper")
+# web_search = GoogleSearchTool(provider="serper")
 
-print(web_search(query="Donald Trump news"))
+# print(web_search(query="Donald Trump news"))
+# TODO fix ValueError: {'message': 'Unauthorized.', 'statusCode': 403}
 
 # quit()
 AUTHORIZED_IMPORTS = [
@@ -66,7 +69,7 @@ AUTHORIZED_IMPORTS = [
     "csv",
 ]
 load_dotenv(override=True)
-login(os.getenv("HF_TOKEN"))
+# login(os.getenv("HF_TOKEN"))  # this is not necessary if env var HF_TOKEN is set
 
 append_answer_lock = threading.Lock()
 
@@ -86,9 +89,24 @@ BROWSER_CONFIG = {
 
 os.makedirs(f"./{BROWSER_CONFIG['downloads_folder']}", exist_ok=True)
 
-model = HfApiModel(
-    custom_role_conversions=custom_role_conversions,
-)
+model_id = os.getenv("MODEL_ID", "deepseek-ai/DeepSeek-V3")
+_ = "" if os.getenv("OPENAI_API_KEY") is None else os.getenv("OPENAI_API_KEY")[:8] + "..."
+
+if os.getenv("MODEL_ID") and os.getenv("OPENAI_API_BASE"):
+    logger.debug(f"""using LiteLLMModel: {model_id=}, {os.getenv("OPENAI_API_BASE")=}, os.getenv("OPENAI_API_BASE")={_}""")
+    model = LiteLLMModel(
+        # "gpt-4o",
+        # os.getenv("MODEL_ID", "gpt-4o-mini"),
+        model_id,
+        custom_role_conversions=custom_role_conversions,
+        api_base=os.getenv("OPENAI_API_BASE"),
+        api_key=os.getenv("OPENAI_API_KEY"),
+    )
+else:
+    logger.debug(f"""using LiteLLMModel: HfApiModel default model_id=Qwen/Qwen2.5-Coder-32B-Instruct""")
+    model = HfApiModel(
+        custom_role_conversions=custom_role_conversions,
+    )
 
 text_limit = 20000
 ti_tool = TextInspectorTool(model, text_limit)
@@ -96,7 +114,7 @@ ti_tool = TextInspectorTool(model, text_limit)
 browser = SimpleTextBrowser(**BROWSER_CONFIG)
 
 WEB_TOOLS = [
-    web_search,
+    # web_search,  # TODO
     VisitTool(browser),
     PageUpTool(browser),
     PageDownTool(browser),
@@ -268,11 +286,11 @@ class GradioUI:
                         with gr.Sidebar():
                             gr.Markdown("""# open Deep Research - free the AI agents!
 
-                OpenAI just published [Deep Research](https://openai.com/index/introducing-deep-research/), an amazing assistant that can perform deep searches on the web to answer user questions.
+                OpenAI just (February 2, 2025) published [Deep Research](https://openai.com/index/introducing-deep-research/), an amazing assistant that can perform deep searches on the web to answer user questions.
 
-                However, their agent has a huge downside: it's not open. So we've started a 24-hour rush to replicate and open-source it. Our resulting [open-Deep-Research agent](https://github.com/huggingface/smolagents/tree/main/examples/open_deep_research) took the #1 rank of any open submission on the GAIA leaderboard! ✨
+                However, their agent has a huge downside: it's not open. So we've started a 24-hour rush to replicate and open-source it. Our (Huggingface's) resulting [open-Deep-Research agent](https://github.com/huggingface/smolagents/tree/main/examples/open_deep_research) took the #1 rank of any open submission on the GAIA leaderboard! ✨
 
-                You can try a simplified version here that uses `Qwen-Coder-32B` instead of `o1`.<br><br>""")
+                You can try a simplified version here that uses `Qwen-Coder-32B` (via smolagnet.HfApiModel) instead of `o1`. Modified: if you set MODEL_ID, OPENAI_API_BASE and OPENAI_API_KEY in the .env or env vars (in hf space these can be set in settings, .env will override env vars), the correspoding model will be used. N.B. if you see errors, it might be because whatever quota is exceeded, clone this space and plug in your own resources and run your own deep-research.<br><br>""")
                             with gr.Group():
                                 gr.Markdown("**Your request**", container=True)
                                 text_input = gr.Textbox(

requirements.txt

@@ -47,4 +47,5 @@ litellm
 google-api-python-client
 google-auth-httplib2
 google-auth-oauthlib
-gradio
+gradio
+loguru

scripts/text_web_browser.py

@@ -11,8 +11,8 @@ from urllib.parse import unquote, urljoin, urlparse
 
 import pathvalidate
 import requests
-# from serpapi import GoogleSearch
-from serpapi.google_search import GoogleSearch
+from serpapi import GoogleSearch
+# from serpapi.google_search import GoogleSearch
 
 from smolagents import Tool
 

set-env.bat

@@ -1,11 +1,11 @@
-set OPENAI_BASE_URL=h=tps://api.siliconflow.cn/v1
-set OPENAI_API_BASE=https://api.siliconflow.cn/v1
-set OPENAI_API_KEY=%SILICONFLOW_API_KEY%
-set MODEL_ID=deepseek-ai/DeepSeek-V3
+REM siliconflow rate limited, wont work for deep-research
+REM set OPENAI_API_BASE=https://api.siliconflow.cn/v1
+REM set OPENAI_API_KEY=%SILICONFLOW_API_KEY%
+REM set MODEL_ID=openai/deepseek-ai/DeepSeek-V3
 
-set OPENAI_BASE_URL=https://litellm.dattw.eu.org/v1
 set OPENAI_API_BASE=https://litellm.dattw.eu.org/v1
 set OPENAI_API_KEY=%LITELLM_API_KEY%
 set MODEL_ID=gpt-4o-mini
 
-REM set SERPAPI_API_KEY b84...
+REM set SERPAPI_API_KEY=b84...
+REM set HF_TOKEN=hf_yC...