-
Notifications
You must be signed in to change notification settings - Fork 5.1k
Open
Description
Hello, I tried to used CLI version or main.py version with different local LLM like llama3.2:3b, qwen3:8b, qwen3:8b, gpt-oss:20b with Ollama
All models tested return error when they call get_news
CLI Error:
[Tool Call] get_news(end_date=2025-11-07, start_date=2025-11-07, ticker=VOO)
11:28:29 [Reasoning] { "Information": "Invalid inputs. Please refer to the API documentation https://www.alphavantage.co/documentation#newsapi and try again." }
11:28:31 [Reasoning] It seems that I made an error in my previous tool call. The get_news tool requires specific parameters such as ticker symbol, start date, and end date. Let me try again with the correct parameters. Since we are analyzing VOO, which is a total stock market index fund, I will use a broad range of dates to capture recent news and sentiment. {"name": "get_news", "parameters": {"end_date":"2025-11-07","start_date":"2022-01-01","ticker":"VOO"}}
Main.py Error:
================================== Ai Message ==================================
Tool Calls:
get_news (call_mcyzyi99)
Call ID: call_mcyzyi99
Args:
end_date: 2025-11-07
start_date: 2025-10-31
ticker: VOO
DEBUG: get_news - Primary: [alpha_vantage] | Full fallback order: [alpha_vantage → openai → google → local]
DEBUG: Attempting PRIMARY vendor 'alpha_vantage' for get_news (attempt #1)
DEBUG: Calling get_news from vendor 'alpha_vantage'...
SUCCESS: get_news from vendor 'alpha_vantage' completed successfully
SUCCESS: Vendor 'alpha_vantage' succeeded - Got 1 result(s)
DEBUG: Stopping after successful vendor 'alpha_vantage' (single-vendor config)
FINAL: Method 'get_news' completed with 1 result(s) from 1 vendor attempt(s)
================================= Tool Message =================================
No errors were encountered when calling other tools, including those using alpha_vantage functionality.
I tested calling the get and get functions without using an agent, and they work correctly.
Is this solely related to the use of local models?
Many thanks for your help and for your amazing work
Metadata
Metadata
Assignees
Labels
No labels