Saw a few posts elsewhere about this. The official line is "coming soon," but we all know what that means. Meanwhile, they're happy to sell you Pro plans with the promise of future functionality.
Tried poking at the API. It's not a "connector" framework. You'd have to:
1. Extract data from your CRM (good luck with Salesforce's nonsense API).
2. Chunk it, embed it, store it yourself in a vector DB.
3. Feed relevant chunks into Perplexity as context with every query.
You're essentially building the entire RAG pipeline yourself. At that point, why pay Perplexity? Just use the open source LLM of the week and keep your data private. Their "AI" just becomes a very expensive, proprietary summarizer.
The real pitfall? Once they *do* launch official connectors, you'll be locked into their ecosystem and pricing. Classic vendor play.
—aB
You're right about the build vs buy calculation being way off here. Building the full RAG stack just to use Perplexity as the final LLM call is architecturally unsound.
The lock-in risk you mention is the real kicker. The moment you build that pipeline, you're committed to their API's stability, cost, and feature set. If they change the context window pricing tomorrow, your entire integration's TCO shifts.
Consider this: the engineering effort you described is better spent on an abstraction layer. Build *your* connector framework once, then you can pipe data to Perplexity, OpenAI, Anthropic, or a local model. That's the only way to avoid the classic vendor trap.