Sebastian Husch Lee
Sebastian Husch Lee
As we released AsyncPipeline in 2.10, we should add run_async methods for all relevant embedders. This doesn't make sense for the SentenceTransformer Embedders since they use local PyTorch models.
Follow up from https://github.com/deepset-ai/haystack/issues/8692 From @julian-risch: > With https://github.com/deepset-ai/haystack/pull/8698 and https://github.com/deepset-ai/haystack/pull/8708 merged, the immediate issue was addressed. We should also check whether it makes sense to emit a warning when...
We have expanded `StreamingChunk` to include new fields such as ```python content: str meta: Dict[str, Any] = field(default_factory=dict, hash=False) component_info: Optional[ComponentInfo] = field(default=None) # All fields below here are new...
Refer to parent issue for more details
## Summary and motivation **ChatGenerator Motivation:** from https://github.com/deepset-ai/haystack-core-integrations/issues/1958 > When using OpenAIChatGenerator with OpenAI-compatible APIs that return additional custom parameters in the response delta (such as reasoning_content, thinking_content, or other...
We should add a `run_async` implementation for `LlamaCppChatGenerator` since we have `AsyncPipeline` available.
As a follow up to https://github.com/deepset-ai/haystack-experimental/pull/369 we should investigate if the `BreakpointConfirmationStrategy` can be used to help enable a human in the loop integration into hayhooks. As per offline discussions...
As a sibling issue to https://github.com/deepset-ai/haystack/issues/9700 after the release of Haystack 2.18.0 we should make sure to use the new `reasoning` field of `StreamingChunk` to store the reasoning content returned...
**Describe the solution you'd like** I’d like to add a new ranker component that leverages a LLM to rerank retrieved documents based on their relevance to the query. This would...