Feature/dynamic rag support #1377
Conversation
…enerate` methods
…ient into feature/dynamic-rag-support
|
Great to see you again! Thanks for the contribution. |
…g on LLM performance
…ient into feature/dynamic-rag-support
| @@ -18,6 +19,7 @@ | |||
| __all__ = [ | |||
| "Filter", | |||
| "GeoCoordinate", | |||
| "GenerativeProvider", | |||
There was a problem hiding this comment.
should we have this in wvc.generative.XXX?
| return _GenerativeOllama(api_endpoint=api_endpoint, model=model, temperature=temperature) | ||
|
|
||
| @staticmethod | ||
| def openai( |
There was a problem hiding this comment.
let's have openai and openai_azure
| temperature=temperature, | ||
| top_p=top_p, | ||
| is_azure=is_azure or False, | ||
| ) |
There was a problem hiding this comment.
should we have a "custom" method? 🤔 Mainly thinking about the generative-dummy module
| temperature=temperature, | ||
| top_p=top_p, | ||
| is_azure=is_azure or False, | ||
| ) |
There was a problem hiding this comment.
How would you implement image/video generation for an existing provider? XXX_image(...)?
| The prompt to use for RaG on the entire result set. | ||
| `grouped_properties` | ||
| The properties to use in the RaG on the entire result set. | ||
| `dynamic_rag` |
| `single_prompt` | ||
| The prompt to use for RaG on each object individually. | ||
| `grouped_task` | ||
| The prompt to use for RaG on the entire result set. |
There was a problem hiding this comment.
We don't really use RAG in the python client, but generative.
How about something like
The prompt for the generative query on all results
(same for the others)
| @@ -53,6 +55,8 @@ async def bm25( | |||
| The prompt to use for RaG on the entire result set. | |||
| `grouped_properties` | |||
| The properties to use in the RaG on the entire result set. | |||
There was a problem hiding this comment.
docstrings are missing here :)
| self._query = _QueryGRPC( | ||
| self._connection, | ||
| self._name, | ||
| self.__tenant, | ||
| self.__consistency_level, | ||
| validate_arguments=self._validate_arguments, | ||
| uses_125_api=self.__uses_125_api, |
There was a problem hiding this comment.
why remove this and doing the check in there again?
| self.generative_provider = generative_provider | ||
|
|
||
| def to_grpc(self, server_version: _ServerVersion) -> generative_pb2.GenerativeSearch: | ||
| if server_version.is_lower_than(1, 27, 1): # TODO: change to 1.28.0 when it drops |
There was a problem hiding this comment.
shouldn't we use the is_127_api bool here?
No description provided.