feat(stream-text)!: re-implement #44
Conversation
exoticknight
commented
Feb 15, 2025
exoticknight
commented
- onChunk
- onStepFinish
- onFinish
- chunkStream
- stepStream
- textStream
exoticknight
changed the title
| let usage: undefined | Usage | ||
| // state | ||
| const steps: StreamTextStep[] = [] | ||
| const stepOne = async (options: StreamTextOptions): RecursivePromise<void> => { |
There was a problem hiding this comment.
While having a closure wrapper here to manage the state implemented well and perfect, but IMO,
- https://github.com/moeru-ai/xsai/pull/44/files#diff-7636fbbf51755bf087c9dcf9f6610de6a2413d69ccf7e6bc91a96f9ad60b8031R138
- https://github.com/moeru-ai/xsai/pull/44/files#diff-7636fbbf51755bf087c9dcf9f6610de6a2413d69ccf7e6bc91a96f9ad60b8031R91
- https://github.com/moeru-ai/xsai/pull/44/files#diff-7636fbbf51755bf087c9dcf9f6610de6a2413d69ccf7e6bc91a96f9ad60b8031R241
- https://github.com/moeru-ai/xsai/pull/44/files#diff-7636fbbf51755bf087c9dcf9f6610de6a2413d69ccf7e6bc91a96f9ad60b8031R303
A lot of things are happening here, there is no comments or diagrams to explain the workflow, or could we have a split version of those functions for better readabilities? What do you think @kwaa ?
There was a problem hiding this comment.
A lot of things are happening here, there is no comments or diagrams to explain the workflow, or could we have a split version of those functions for better readabilities? What do you think @kwaa ?
I'm in favor of splitting functions.
There was a problem hiding this comment.
The processing of messages primarily follows the workflow of OpenAI's SDK. Multi-step conversations are implemented with reference to generateText. Each step represents a single dialogue flow, which may output multiple chunks during that step.
Co-authored-by: Neko <neko@ayaka.moe>
Co-authored-by: Neko <neko@ayaka.moe>
Co-authored-by: Neko <neko@ayaka.moe>
Co-authored-by: Neko <neko@ayaka.moe>
kwaa
changed the title
