Model Catalog Migration 7/7: Guides & Tutorials

[roh26it]

Model Catalog Migration - Guides & Tutorials

What's changing

Step-by-step guides and tutorials updated for Model Catalog flow.

Files changed (24)

Getting Started (Major):

File	Changes
getting-started-with-ai-gateway.mdx	+584/-998 (Core onboarding)
101-on-portkey-s-gateway-configs.mdx	+40/-99
image-generation.mdx	+34/-112
llama-3-on-groq.mdx	+27/-67
tackling-rate-limiting.mdx	+25/-49

Integration Guides (6):

• groq.mdx, vercel-ai.mdx, arize-portkey.mdx
• langchain.mdx, llama-3-on-portkey-+-together-ai.mdx, mixtral-8x22b.mdx

Use Cases (10):

• Fallbacks, JSON schema, few-shot prompting, DeepSeek R1
• Supabase + PGVector, cost tracking, and more

Prompts (2):

• build-a-chatbot-using-portkeys-prompt-templates.mdx
• llama-prompts.mdx

API Reference (1):

• config-object.mdx

Review checklist

• Step numbers match actual UI flow
• Code snippets use new @provider/model format
• Links to Model Catalog are correct
• User journey makes sense for newcomers

Risk: 🟡 Medium

User-facing tutorials - important for onboarding.

Part of

Model Catalog Migration - Split from #597 for easier review

---

🎉 This is the final PR in the migration series!

[siddharthsambharia-portkey]

in all thse files wherever we are passing a dummy value for API key

client = OpenAI(
    api_key="X",
    base_url=PORTKEY_GATEWAY_URL,
    default_headers=createHeaders(
        api_key=userdata.get("PORTKEY_API_KEY"),
        config=config
    )
)

like this case above

it shoud be

client = OpenAI(
    api_key=userdata.get("PORTKEY_API_KEY"),
    base_url=PORTKEY_GATEWAY_URL,
    default_headers=createHeaders(
        config=config
    )
)

Portkey works iwth OpeAI client natively and we don't need to pass the API Key as a separete header

make sur ethis is different from

client = OpenAI(
    api_key=userdata.get('GROQ_API_KEY'),  # replace with your Groq API key
    base_url=PORTKEY_GATEWAY_URL,
    default_headers=createHeaders(
        provider="groq",
        api_key=userdata.get('PORTKEY_API_KEY')  # replace with your Portkey API key
    )
)

in this case we need portkey API key separetely because we are using provider api key natively without and AI provider slug from portkey

so we do need thte default header separately

[siddharthsambharia-portkey]

Addressed the API key pattern issue across all files:

Fixed files:

• guides/use-cases/fallback-from-sdxl-to-dall-e-3.mdx
• guides/use-cases/comparing-top10-lmsys-models-with-portkey.mdx
• guides/integrations/langchain.mdx
• guides/integrations/groq.mdx
• guides/getting-started/image-generation.mdx

Changes made:
When using configs or provider slugs from Model Catalog, the Portkey API key is now passed directly to the api_key parameter instead of in headers:

client = OpenAI(
    api_key=PORTKEY_API_KEY,  # Portkey API key goes here
    base_url=PORTKEY_GATEWAY_URL,
    default_headers=createHeaders(
        config=config  # No api_key needed in headers
    )
)

Note: Files using provider API keys directly (like the Groq example with GROQ_API_KEY) still correctly keep the Portkey API key in headers.