Sheila L.

Hey, thanks 🔒[private user] and 🔒[private user]. I think it's very helpful. Another quick question, what should be the format of the tool definition string? Could you provide us an example of one of the outputs for it?

how can I know what is the expected content of tool call?

Hey, Xander S.. It's me again 🙂 . Regarding to this point that LLMApp mentioned:

• Your model supports function calling,  • Your prompt template is correctly structured to yield an explanation,  • And that your DataFrame includes a properly named “tool_call” column with exactly the expected content.

do you think this could be causing the issue? in the newest version you are installing ! pip install -qq "arize-phoenix==6.1.0" "llama-index-llms-openai==0.2.2" "openai===1.57.0" gcsfs==2024.10.0 nest_asyncio==1.6.0 langchain==0.3.10 langchain-openai==0.2.11 openinference-instrumentation-langchain==0.1.29 but I am importing this: ! pip install -qq "arize-phoenix[llama-index]" "arize-phoenix[evals]>=0.14.0" "llama-index-llms-openai" "openai>=1" gcsfs nest_asyncio langchain langchain-openai

no, I am running it in local

do you know why trace_df may be empty?

thank you, so we keep in contact

this is with the old version

with the latest version I noticed the questions are not under "", that could be affecting for trace_df empty?

now that you are running on Colab, trace_df is not empty for you?