Model Catalog Migration 8C/9: Extended LLM Provider Integration (Remaining Providers)

[siddharthsambharia-portkey]

Model Catalog Migration - Remaining Providers

What's changing

48 additional LLM provider pages updated to Model Catalog pattern.

Files changed (48 files)

Major Providers:

• Groq, Mistral AI, Together AI, Fireworks, DeepSeek
• Ollama, vLLM, Replicate, HuggingFace
• Perplexity, OpenRouter, Cerebras, Cohere

Other Providers:

• Deepgram, Jina AI, Nomic, Voyage AI, Stability AI
• SambaNova, Lambda, Lepton, Modal, Triton
• Local AI, Snowflake Cortex, Workers AI
• And 25+ more

Review checklist

• "Add Provider" section uses 4-step flow
• Code examples use model="@provider-slug/model-name"
• Links point to /product/model-catalog
• All providers follow consistent pattern

Risk: 🟡 Medium

Multiple provider pages - but all follow same pattern as previous PRs.

Part of

Model Catalog Migration - Split from PR #637 (migration/08-llm-providers-extended)

[siddharthsambharia-portkey]

All review comments addressed (9/9 resolved)

Here's a summary of all changes made in commit f313c17:

1. cohere.mdx — Fixed whitespace in code block

Removed extra leading spaces from the re-ranking Python code block (portkey.post() call and its arguments were over-indented).

2. deepseek.mdx — Fixed all whitespace in code blocks

Fixed indentation issues in three code blocks:

• Multi-round conversation Python example (extra 4-space indent on client, round logic, etc.)
• Multi-round conversation JS example (extra indent on const client, model param)
• JSON output Python example (extra 4-space indent throughout)

3. fireworks.mdx — Added grammar mode description + whitespace fixes

• Added the formal grammars description text before the Grammar Mode section (links to Wikipedia & GGML, response_format param explanation, patient classification intro)
• Fixed whitespace across all code blocks: Image Generation, Grammar Mode, JSON Mode, Vision Models, and Embeddings sections

4. fireworks.mdx — Simplified Supported Models section

Replaced the model list (Llama, FireLLaVA, SD XL, embeddings) with a single Card linking to Fireworks documentation.

5. groq.mdx — Simplified Supported Models section

Replaced the model table (Llama, Mixtral, Gemma, Whisper, TTS) with a single Card linking to Groq documentation.

6. huggingface.mdx — Added custom host URL example

Added https://your-custom-url/v1 as an example in the Custom Host setup step for dedicated inference endpoints.

7. mistral-ai.mdx — Removed Codestral vs Mistral API Endpoint section

Removed the entire comparison section including the image and external link.

8. segmind.mdx — Replaced chat completions with image generation

Replaced Python, JS, OpenAI Py, and OpenAI JS chat completion examples with portkey.images.generate() examples. cURL example was already correct.

9. upstage.mdx — Removed Supported Models & Endpoints sections

Removed the "Supported Models" accordion and "Supported Endpoints and Parameters" table entirely.