Sebastian Husch Lee
Sebastian Husch Lee
Based on the feedback from @d-kleine in the discussion [here](https://github.com/deepset-ai/haystack-experimental/discussions/209#discussioncomment-12377313) it would be interesting to explore whether we could run the tools in ToolInvoker in a parallel way. This is...
Currently we only can process Image Documents by embedding them with the new ImageEmbedder. However, another approach would be to caption the image or extract the text from a scanned...
Often when working with PDF files we run into complex scenarios with regards to non-text content. For example, * The whole PDF could be a scan so it has no...
There have been more instances of newer classifiers actually utilizing the Text-Generation task in HuggingFace where the final predicted tokens made by the LLM count as the labels. For example,...
Now that we have merged https://github.com/deepset-ai/haystack/issues/9358 it would be good to double check if the logic inside of `_convert_chat_completion_chunk_to_streaming_chunk` applies well to our core-integrations that inherit from the `OpenAIChatGenerator`. Specifically,...
**Is your feature request related to a problem? Please describe.** While working on the parent issue https://github.com/deepset-ai/haystack/issues/9347 it's becoming evident that it's useful for the streaming callback function to have...
While working on https://github.com/deepset-ai/haystack/pull/9342 we found that the HuggingFaceAPIChatGenerator does not support tool parameter schemas with `$defs` field which is supported by other LLM providers like OpenAI. There is an...
It would be great if we could support breakpoints within SuperComponents and Agent. This is still in an exploratory phase so it would be good to plan out what this...
The current implementation of pipeline breakpointing only works with `Pipeline` and not (yet) with `AsyncPipeline`. We should start planning on how to do this and determine its difficulty. For example,...
We should look to support the new Responses API from OpenAI. Here is an article from OpenAI explaining the difference https://platform.openai.com/docs/guides/responses-vs-chat-completions It sounds like this will become their preferred API...