For the same code, I run it on Google Colab and it does not time out, but in local Python it prompts that the llm call timed out and reports an error:

[egtai]

For the same code, I run it on Google Colab and it does not time out, but in local Python it prompts that the llm call timed out and reports an error:
nfo: request_timeout: None
timeout: None. Received Model Group=gpt-4o-2024-11-20
Available Model Group Fallbacks=None LiteLLM Retried: 2 times, LiteLLM Max Retries: 3

my code:
`from paperqa import Settings, ask, agent_query
from paperqa.settings import AgentSettings
import os
import openai

OPENAI_API_KEY = "sk-......."
os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY

openai.api_key = OPENAI_API_KEY
print("✅ OpenAI API key configured successfully!")

answer = ask(
"what is PaperQA2?",
settings=Settings(
llm="gpt-4o-2024-11-20",
summary_llm="gpt-4o-2024-11-20",

    agent=AgentSettings(
        agent_llm="gpt-4o-2024-11-20",
        search_count=6,
        timeout=600.0,
    ),

    embedding="text-embedding-3-large",

    answer=dict(
        evidence_k=8,
        answer_max_sources=4,
        evidence_summary_length="about 80 words",
        answer_length="about 150 words, but can be longer",
        max_concurrent_requests=2,
    ),

    parsing=dict(
        chunk_size=4000,
        overlap=200,
    ),

    verbosity=1,
)
    
)`

ps:My test code and paperqa2.pdf are in the same folder

[dosubot]

Hi @egtai! I'm Dosu and I’m helping the paper-qa team.

Timeouts in local Python but not in Colab are often caused by differences in network, package versions, or configuration. Here are the most likely causes and fixes:

1. Make sure your OpenAI API key and project have access to the gpt-4o-2024-11-20 model, and that you use the same key in both environments. If your local project lacks access, LiteLLM will retry and eventually time out, as seen in similar issues example.
2. Align the versions of paperqa, openai, and LiteLLM between Colab and local. Version mismatches can cause subtle bugs or timeouts.
3. Set concurrency parameters lower for local runs: set index.concurrency (in IndexSettings) and max_concurrent_requests (in AnswerSettings) to 1. This prevents local overload and rate limit issues example.
4. Check your local network and firewall settings to ensure nothing is blocking or slowing requests to OpenAI’s API. Local environments are more likely to have restrictive firewalls or unstable connections, which can cause timeouts example.
5. If you see more detailed error logs or stack traces, review them for clues about network or configuration problems.

If you’re running this as a script (not in a notebook), consider using the asynchronous API, as synchronous calls can behave differently in local scripts versus Colab notebooks example.

If you still see timeouts after these checks, please share any additional error logs or your exact package versions for more targeted help.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}