PraisonAI + LiteLLM / Ollama / Azure — what's your setup?

[Dhivya-Bharathy]

PraisonAI agents often sit behind LiteLLM for model routing, Ollama for local inference, or Azure OpenAI for hosted enterprise workloads. The "best" combination depends heavily on your constraints — cost ceiling, latency requirements, data-residency rules, team size, and whether you need the models to run on your own hardware.

There's no single right answer, so let's compare notes.

Reply with:

1. Which stack you run — e.g. PraisonAI → LiteLLM → mix of Ollama + Azure, or straight OpenAI, or something else entirely.
2. One production lesson — something about memory, sessions, cost, or reliability that surprised you once traffic was real.
3. (Optional) A code/config snippet or a link to a public repo that shows the setup. No secrets, API keys, or tokens — sanitise before posting.

The goal is a practical reference other builders can search later, not a hype thread.

Repo: MervinPraison/PraisonAI