Does anyone know how to get the new OpenAI embedding model to work:


---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[29], line 1
----> 1 embeddings = embed_model.get_text_embedding("Your text here")
2 print(embeddings)

File ~/anaconda3/envs/embeddings/lib/python3.11/site-packages/llama_index/core/embeddings/base.py:207, in BaseEmbedding.get_text_embedding(self, text)
196 """
197 Embed the input text.
198
(...)
202 predefined instructions can be found in embeddings/huggingface_utils.py.
203 """
204 with self.callback_manager.event(
205 CBEventType.EMBEDDING, payload={EventPayload.SERIALIZED: self.to_dict()}
206 ) as event:
--> 207 text_embedding = self._get_text_embedding(text)
209 event.on_end(
210 payload={
211 EventPayload.CHUNKS: [text],
212 EventPayload.EMBEDDINGS: [text_embedding],
213 }
214 )
216 return text_embedding
...
134 return (
--> 135 client.embeddings.create(input=[text], model=engine, **kwargs).data[0].embedding
136 )

TypeError: Embeddings.create() got an unexpected keyword argument 'dimensions'


Does the vector store automatically use the larger number (3k vs. 1.5k) of dimensions?

For graph queries, how do we control what text (e.g. "grazing animals") is used for finding the right vectors vs. instructions for GPT (e.g. "write a paragraph")?

How do I see what chunks of text were created and the associate vectors?

Anyone know what is causing this ---------------------------------------------------------------------------
ImportError Traceback (most recent call last)
Cell In[5], line 1
----> 1 from llama_index import download_loader, GPTSimpleVectorIndex
2 from pathlib import Path

File ~/.local/lib/python3.9/site-packages/llama_index/init.py:18
15 from llama_index.embeddings.openai import OpenAIEmbedding
17 # structured
---> 18 from llama_index.indices.common.struct_store.base import SQLDocumentContextBuilder
19 from llama_index.indices.composability.graph import ComposableGraph
20 from llama_index.indices.empty import GPTEmptyIndex

File ~/.local/lib/python3.9/site-packages/llama_index/indices/common/struct_store/base.py:9
7 from llama_index.data_structs.node_v2 import Node
8 from llama_index.data_structs.table import StructDatapoint
----> 9 from llama_index.indices.response.builder import ResponseBuilder, TextChunk
10 from llama_index.indices.service_context import ServiceContext
11 from llama_index.langchain_helpers.chain_wrapper import LLMPredictor

File ~/.local/lib/python3.9/site-packages/llama_index/indices/response/builder.py:18
16 from llama_index.data_structs.node_v2 import Node, NodeWithScore
17 from llama_index.docstore_v2 import DocumentStore
---> 18 from llama_index.indices.common.tree.base import GPTTreeIndexBuilder
19 from llama_index.indices.service_context import ServiceContext
...
33 )
34 from llama_index.indices.service_context import ServiceContext
35 from llama_index.optimization.optimizer import BaseTokenUsageOptimizer

ImportError: cannot import name 'RESPONSE_TEXT_TYPE' from partially initialized module 'llama_index.indices.response.builder' (most likely due to a circular import) (/home/gregtanaka/.local/lib/python3.9/site-packages/llama_index/indices/response/builder.py)

It looks like GPTSimpleVectorIndex no longer has set_text so the code doesn't work...

Has anyone been able to get pickle to work with Llama Index? Seems like everyone gets a 'cannot pickle 'builtins.CoreBPE' object' error

@Logan M any idea how to do this? I tried several method but none worked.

Has anyone figured out how to do custom prompts with ComposableGraph.query()?