From LiteLLM

This guide covers moving a LiteLLM deployment (self-hosted Proxy or LiteLLM Cloud) to HiWay2LLM. If your team is tired of maintaining the proxy config, the Redis for rate limits, the Postgres for logs, and the alerting stack — this is a fully managed equivalent. Budget ~10 minutes for the swap, plus whatever time you need to re-express your fallback / observability config in the HiWay dashboard.

Step 1: Get your HiWay API key

Open Dashboard → Keys and click New key. Copy the hw_live_ key once — we only store its SHA-256 hash.

Step 2: Swap your configuration

In LiteLLM Proxy you were pointing your app at http://your-proxy:4000 (or the LiteLLM Cloud endpoint) with a sk-litellm-... master key, and LiteLLM dispatched to the right upstream based on your config.yaml. On HiWay you point at https://app.hiway2llm.com/v1 with a hw_live_... key, and the router dispatches based on your enabled providers and routing profile — no YAML to maintain.

from openai import OpenAI

client = OpenAI(
    base_url="https://app.hiway2llm.com/v1",
    api_key="hw_live_YOUR_KEY",
)

response = client.chat.completions.create(
    model="auto",  # LiteLLM alias translation happens via "auto" + enabled providers
    messages=[{"role": "user", "content": "Summarize our Q3 report"}],
)

print(response.choices[0].message.content)

Step 3: Verify

Send one request, check X-HiWay-Routed-Model on the response, and open Dashboard → Usage to see it land. If you were piping LiteLLM logs into Langfuse / Helicone / Datadog, you can now read the same data from Dashboard → Analytics or pull it via the /v1/usage endpoint.

Gotchas specific to LiteLLM

• Self-hosted customizations are gone: if you had custom callbacks, a modified dispatcher, or patched the LiteLLM source, those don't carry over. HiWay is a managed service — the upside is no more maintenance, the downside is you trade that surface for our feature set.
• `config.yaml` → dashboard settings: provider credentials move to Settings → Providers, model aliases are replaced by model: "auto" plus enabled-provider selection, and rate limits move to per-key settings in Dashboard → Keys.
• Fallback rules: LiteLLM's fallbacks block is replaced by HiWay's automatic same-tier provider fallback (max 2 retries). If you had elaborate fallback chains with specific orderings, check that the default HiWay behavior covers them — otherwise pin model explicitly in code.
• Observability hooks: LiteLLM's success_callback / failure_callback (Langfuse, Helicone, S3) are not pluggable in HiWay. Instead, pull analytics from the HiWay API (/v1/usage, /v1/events) and forward wherever you want.
• Master key vs per-service keys: if you were using one LiteLLM master key for everything, split it now — mint one hw_live_ key per service in Dashboard → Keys so Guardian and rate limits can scope per caller.