Naren Gogineni

Include usage object in openai streaming response by default

1

### What Happened? To get the usage object, the user has to pass `streaming_options: { include_usage: true}` This can be default behaviour ### What Should Have Happened? _No response_ ###...

This PR introduces changes for a simplified framework for instrumenting traces to portkey from the sdk. The main changes in this PR are: - introduce decorator `@watch` for tracing custom...

computer use for anthropic on bedrock

2

https://github.com/Portkey-AI/gateway/issues/1176 Testing: sample request body ```json { "model": "us.anthropic.claude-3-7-sonnet-20250219-v1:0", "anthropic_beta":"computer-use-2025-01-24", "max_tokens": 2000, "tools": [ { "type": "computer", "computer": { "name": "computer_20250124", "display_width_px": 1024, "display_height_px": 768, "display_number": 1 } } //...

### What Would You Like to See with the Gateway? anthropic specific tools are already supported for anthropic and vertex as they use the same transforms, this is a ticket...

Feature: Beta support for /v1/messages route

4

TO-DO: - [ ] model specific additional fields TO-DO (subsequent versions): - [ ] support stream concatenation for caching - [ ] fix types for bedrock Testing Done: - [...

feat: multimodal embeddings for bedrock titan and cohere

1

This PR adds support for multimodal embeddings in compliance with the portkey signature: example requests: 1. Titan text embedding: ```sh curl --location 'http://localhost:8787/v1/embeddings' \ --header 'x-portkey-provider: bedrock' \ --header 'Content-Type:...

native fetch has a default timeout of 300s

1

### What Happened? This hasn't surfaced so far because of our use of workerd environment, but native fetch from node js has a deafult 300s timeout https://github.com/nodejs/node/issues/46375 so if the...

### What Would You Like to See with the Gateway? https://docs.anthropic.com/en/docs/agents-and-tools/computer-use https://docs.anthropic.com/en/docs/agents-and-tools/tool-use/web-search-tool ### Context for your Request _No response_ ### Your Twitter/LinkedIn _No response_

simplify openai base transform for chat complete

2

noticed that the openai chat complete base transforms are missing a few parameters like `top_logprobs` and `parallel_tool_calls` I've used the openai caht complete transformation config and I've added a line...