Es vector store basic implementation #21
Conversation
botnim/sync.py
Outdated
| tool_resources = None | ||
| tools = None | ||
| print(f'Updating assistant: {config["name"]}') | ||
| # Load context, if necessary | ||
| if config.get('context'): | ||
| vs = VectorStoreOpenAI(config, config_dir, production, client) | ||
| if config.get('context') and replace_context: # Only runs if both conditions are true |
There was a problem hiding this comment.
This logic is wrong - we need the tools and tools_resources even if we don't replace the context.
| # Generate embedding | ||
| response = self.openai_client.embeddings.create( | ||
| input=content, | ||
| model="text-embedding-ada-002", |
There was a problem hiding this comment.
why did you choose this older model?
'text-embedding-3-small' is cheaper and more advanced.
https://platform.openai.com/docs/guides/embeddings#embedding-models
There was a problem hiding this comment.
you are very right, and actually it was already included in the configurations. Notice we might change this selection later. For now - fixed.
| while len(file_streams) > 0: | ||
| current = file_streams[:32] | ||
|
|
||
| for filename, content_file, content_type in current: |
There was a problem hiding this comment.
no need for the double loop I think.
|
|
||
| def get_or_create_vector_store(self, context, context_name, replace_context): | ||
| ret = None # return value | ||
| vs_name = self.env_name(self.config['name']).lower().replace(' ', '_') |
There was a problem hiding this comment.
we need to include the context name as well here, as we would want each context to be in a separate index.
| return 0 | ||
|
|
||
| def update_tools(self, context_, vector_store): | ||
| if len(self.tools) == 0: |
There was a problem hiding this comment.
the tool added needs to be a function, not file-search.
Something like:
{
"type": "function",
"name": "search_common_knowledge",
"description": "Semantic search the 'common-knowledge' vector store",
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "The query string to use for searching"
}
},
"required": ["query"]
}
}There was a problem hiding this comment.
thanks. I did not think of that.
| ), | ||
| )) | ||
|
|
||
| def update_tool_resources(self, context_, vector_store): |
There was a problem hiding this comment.
no internal vector store ids for elasticsearch
closes #12
closes #13
closes #16