Not ideal result from assistant (in comparison with Phind)

Ishan512

Not sure if I should be reporting one off assistant response quality issues, but here is one use case.

I had asked whether Claude supports function calling, like OpenAi api. For some reason the sources pulled in assistant aren't the tool function calling docs (which do show if I just search for it). And guessing because of that the response seems to be highlighting the bedrock integration. Phind on the other hand has a nicer direct answer.

Answer similar to what Phind gave

Vlad

Ishan512

Answer similar to what Phind gave

Can you show a screnshot of that answer?

Luis

Ishan512 To clarify, you're suggesting the Assistant's response is too verbose?

Note: based on your screenshot it doesn't look like Phind used Anthropic's tool use docs either.

Ishan512

Vlad Should be the second screenshot I'd attached. Are there any more details you'd like me to share?

Vlad

Ishan512 May bad i didnt notice we went to screenshot gallery

Ishan512

Luis Not necessarily the verbosity (which also would be nice if it was reduced), but it seems to over emphasize the AWS bedrock approach and didn't clearly call out the official 'tools' approach from Anthropic. While Phind (even though they didn't like the Tools link in their sources) clearly mentions it's called "tool use" and gave me the exact parameters that each tool needs etc (which seems correct based on the docs).

(Btw by and large I like assistant more than Phind now. I was using Phind for 1.5 years and recently switched to Kagi ultimate, so loving all the improvements 🙏. Just thought I'd list out any that seem off).

Ishan512

Another case where assistant response isn't great compared to Phind.