Local AI

LiteLLM

Call Anthropic, OpenAI, Google and 100+ models through one consistent API, with cost tracking and fallbacks.

Who it is for

Anyone building on more than one model who wants to avoid lock-in and track spend.

Install it

pip install litellm
# run the proxy:
litellm --model claude-3-5-sonnet

Before production

Lets you swap models without rewriting code. Set per-key budgets so costs do not surprise you.

Where Blash AI comes in

We stand up the gateway, set budgets and fallbacks, and route each workflow to the right model for the job.

Run it, then wire it in

When you want this running on your real stack, that is the engagement

Open-source workflow automation. Connect apps, APIs and AI steps in one flow. A self-hosted Zapier you own.

Run open language models on your own machine or server. Keeps data in-house for sensitive work.

An open vector database that powers retrieval, so a model can answer from your documents with citations.