I’ve tried the action of Perplexity and Google Search, and found that Google Search provides better results with citation. The action in Perplexity is not the same as using it directly on their platform.
What are your thoughts on Pickaxe’s use case? Would you choose to use Perplexity or stick to the Google Search action?
Just tried out the perplexity. It’s perplexing ! I asked it to search latest info, it gave me 2023 with very bad results. It’s completely different from the perplexity apps. Plus you have no idea what it is doing at the back unlike the native one with “reasoning dialogue”. Is it possible for pickaxe to include the reasoning or that’s how the API works.
Thanks for sharing your experiences. I’m looking forward to see the Perplexity experience improve to match the one provided on the platform. It adds a lot of value to our users. One of the best things about Pickaxe, it is LLM agnostic, so if we can match the results on the platforms, it becomes a great selling point for us as solutions providers. We get to choose the most suitable LLM for our client’s different use cases.
Though the results do seem a bit plain when testing on Pickaxe when compared to using it on the Perplexity platform, would this be a limitation of the Pickaxe? or could I be missing something else?
Perplexity outputs look more aesthetically pleasing because they use NextJS and TypeScript which Pickaxe currently does not use. There are several users who are requesting Pickaxe implement a similar output aesthetic as Perplexity. You can always add a feature request if you want! You have my vote.
Yes, in your Pickaxe custom action, you have correctly specified Sonar Pro as the model in this section of your code:
"model": "sonar-pro"
As shown in the attached screenshot, this is the correct place to define the model you want to use for Perplexity. If you intended to use Sonar Pro, your setup is correct.
Thanks @Ned.Malki. Aesthethics is a bonus, the abillity to cite sources the way Perplexity does is the key thing I’ll put in a feature request later today.
I’m confused. Does Perplexity have more than one kind of API key? I tried to follow Ned’s instructions above, but I seem to end up back where I have already purchased a Perplexity API key. I’d also appreciate clarification on the purpose of going to “api-request.io” as Ned’s instructions state.
Does Perplexity have more than one kind of API key?
No, Perplexity has one type of API key, but the key needs to be used correctly depending on the model (e.g., sonar-pro) and the specific request settings.
Ensure that your key is active and properly configured in your environment variables or request headers.
Why does my guide mention api-request.io?
api-request.io is a tool for testing API calls before integrating them into Pickaxe.
It allows users to manually send requests, view responses, and debug authentication or formatting issues before adding them to a Pickaxe Connected Action.
If you’re facing issues, try sending a test request via api-request.io to verify that your API key works before troubleshooting Pickaxe.
If you’re still stuck, try:
Testing your API key in api-request.io to confirm it returns a response.
Ensuring your API key is set correctly in Pickaxe’s environment variables.
Checking your model selection (e.g., sonar-pro) in the API request payload.
Thanks very much, Ned! Your recommendation about using the Perplexity key correctly depending on the model only applies to custom actions, not to the ready-made Perplexity action in Pickaxe, correct?
Correct. The ready-made Perplexity action integrates with Sonar as the documentation states. My recommendation is for users who want more granular control over the Perplexity search functionality.