sij/khoj
1
0
Fork 0
mirror of https://github.com/khoj-ai/khoj.git synced 2024-11-23 15:38:55 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

Add comments explaining each field in the search model config in DB

Browse source
This commit is contained in:
Debanjum Singh Solanky 2024-04-25 13:42:46 +05:30
parent 799efb5974
commit cf08eaf786
2 changed files with 13 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
documentation/docs/miscellaneous/advanced.md
View file

@ -10,6 +10,8 @@ For example, the [paraphrase-multilingual-MiniLM-L12-v2](https://huggingface.co/
1. Manually update the search config in server's admin settings page. Go to [the search config](http://localhost:42110/server/admin/database/searchmodelconfig/). Either create a new one, if none exists, or update the existing one. Set the bi_encoder to `sentence-transformers/multi-qa-MiniLM-L6-cos-v1` and the cross_encoder to `mixedbread-ai/mxbai-rerank-xsmall-v1`.
2. Regenerate your content index from all the relevant clients. This step is very important, as you'll need to re-encode all your content with the new model.
Note: If you use a search model that expects a prefix (e.g [mixedbread-ai/mxbai-embed-large-v1](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1)) to the query (or docs) string before encoding. Update the `bi_encoder_query_encode_config` field with `{prompt: <prefix-prompt>}`. Eg. `{prompt: "Represent this query for searching documents"}`. You can pass a valid JSON object that the SentenceTransformer `encode` function accepts
## Query Filters
Use structured query syntax to filter entries from your knowledge based used by search results or chat responses.

11
src/khoj/database/models/__init__.py
View file

@ -179,16 +179,27 @@ class SearchModelConfig(BaseModel):
class ModelType(models.TextChoices):
TEXT = "text"
# This is the model name exposed to users on their settings page
name = models.CharField(max_length=200, default="default")
# Type of content the model can generate embeddings for
model_type = models.CharField(max_length=200, choices=ModelType.choices, default=ModelType.TEXT)
# Bi-encoder model of sentence-transformer type to load from HuggingFace
bi_encoder = models.CharField(max_length=200, default="thenlper/gte-small")
# Config passed to the sentence-transformer model constructor. E.g device="cuda:0", trust_remote_server=True etc.
bi_encoder_model_config = models.JSONField(default=dict)
# Query encode configs like prompt, precision, normalize_embeddings, etc. for sentence-transformer models
bi_encoder_query_encode_config = models.JSONField(default=dict)
# Docs encode configs like prompt, precision, normalize_embeddings, etc. for sentence-transformer models
bi_encoder_docs_encode_config = models.JSONField(default=dict)
# Cross-encoder model of sentence-transformer type to load from HuggingFace
cross_encoder = models.CharField(max_length=200, default="mixedbread-ai/mxbai-rerank-xsmall-v1")
# Inference server API endpoint to use for embeddings inference. Bi-encoder model should be hosted on this server
embeddings_inference_endpoint = models.CharField(max_length=200, default=None, null=True, blank=True)
# Inference server API Key to use for embeddings inference. Bi-encoder model should be hosted on this server
embeddings_inference_endpoint_api_key = models.CharField(max_length=200, default=None, null=True, blank=True)
# Inference server API endpoint to use for embeddings inference. Cross-encoder model should be hosted on this server
cross_encoder_inference_endpoint = models.CharField(max_length=200, default=None, null=True, blank=True)
# Inference server API Key to use for embeddings inference. Cross-encoder model should be hosted on this server
cross_encoder_inference_endpoint_api_key = models.CharField(max_length=200, default=None, null=True, blank=True)