gateway

gateway copied to clipboard

### What Happened? adding sambanova as provider does not return any available models ### What Should Have Happened? when searching for models in Model Provisioning for sambanova integration the available...

snova-jorgep

fix: apply override_params to FormData request bodies

2

## Description Transform the javascript FormData object based on the supplied override_params. Modified transformToProviderRequest.ts on line 235: ``` if (requestBody instanceof FormData) { if (typeof providerOptions.overrideParams === 'undefined') { return...

AgileEduLabs

### What Happened? Hi, I'm deploying the service using the source code directly. While reviewing the caching logic, I attempted a global search for semantic but couldn’t find any actual...

seven-lv

Audio transcription endpoint does not respect Gateway's override_params

6

### What Happened? With `portkey.chat.completions.create`, we can provide `override_params` to the gateway to specify which model to use. The `model` parameter in `portkey.chat.completions.create` itself can be omitted since we specify...

AgileEduLabs

### What Happened? Our original requests to LLM uses message order as `['image_url', 'image_url', 'text']` When we use replay for the above trace -> the portkey changes the order to...

im-ajaymeena

Unable to Force Portkey to Use Bedrock Converse Endpoint

3

### What Happened? I cannot seem to figure out why Portkey sometimes decides to use the `/invoke-with-response-stream` instead of `/converse-stream` endpoint for Bedrock models . The `/invoke-with-response-stream` url does not...

dldereklee

Retry settings not respected when guardrail fails

5

### What Happened? If the guardrail checks as defined in the config fails (using output_guardrails, with deny set to true), then the portkey.chat.completions.create request immediately errors with a status of...

AgileEduLabs

[New Provider Integration] Support Cloudflare Workers AI /chat/completions route

2

### What Would You Like to See with the Gateway? more details https://docs.together.ai/reference/chat-completions-1 ### Context for your Request _No response_ ### Your Twitter/LinkedIn _No response_

narengogi

### What Would You Like to See with the Gateway? the ai21 integration is pretty dated, the existing integration makes calls to `/v1/{model}/chat` whereas the new API is `/v1/chat/completions`, the...

narengogi

### What Happened? We use Claude as our primary model. If the context window exceeds Sonnet's limit, we fall back to Gemini using the same tool definitions that work for...

umuthopeyildirim