Skip to content

armgit / dspy

Public Imported from GitHub
0

DSPy: The framework for programming—not prompting—language models

Python

Merge branch 'main' into support-other-models-and-all-fields-pgvector

Joshmantovafbb90ae0145f
diff --cc dspy/retrieve/pgvector_rm.py
index 1c4ce7f35,d12eb6b05..2e87eed72
--- a/dspy/retrieve/pgvector_rm.py
+++ b/dspy/retrieve/pgvector_rm.py
@@@ -61,11 -65,11 +66,12 @@@ class PgVectorRM(dspy.Retrieve)
self,
db_url: str,
pg_table_name: str,
- openai_client: openai.OpenAI,
+ openai_client: Optional[openai.OpenAI] = None,
+ embedding_func: Optional[Callable] = None,
k: Optional[int] = 20,
embedding_field: str = "embedding",
- fields: List[str] = ['text'],
+ fields: list[str] = ['text'],
+ embedding_model: str = "text-embedding-ada-002",
):
"""
k = 20 is the number of paragraphs to retrieve
@@@ -81,9 -86,9 +89,9 @@@
super().__init__(k=k)
- def forward(self, query: str, k: Optional[int]=20):
+ def forward(self, query: str, k: Optional[int] = 20):
"""Search with PgVector for self.k top passages for query
-
+
Args:
query (str): The query to search for
k (Optional[int]): The number of top passages to retrieve. Defaults to self.k
@@@ -115,9 -116,17 +119,19 @@@
sql_query,
(query_embedding, self.k))
rows = cur.fetchall()
+ columns = [descrip[0] for descrip in cur.description]
for row in rows:
- related_paragraphs.append(dspy.Example(long_text=row[0], document_id=row[1]))
+ data = dict(zip(columns, row))
+ related_paragraphs.append(dspy.Example(**data))
# Return Prediction
- return related_paragraphs
+ return related_paragraphs
+
+ def _get_embeddings(self, query: str) -> list[float]:
+ if self.openai_client is not None:
+ return self.openai_client.embeddings.create(
- model="text-embedding-ada-002",
++ model=self.embedding_model,
+ input=query,
+ encoding_format="float",
+ ).data[0].embedding
+ else:
+ return self.embedding_func(query)