gmakstutis/cognee
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fixes to memory component

Browse source
This commit is contained in:
Vasilije 2024-02-19 12:44:53 +01:00
parent 0a38e09b3f
commit 99b1073560
3 changed files with 12 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
cognitive_architecture/database/graphdb/graph.py
View file

@ -1004,7 +1004,7 @@ async def unlink_user_from_another(
logging.error(f"Error disconnecting user nodes: {e}") logging.error(f"Error disconnecting user nodes: {e}")
raise raise
from .networkx_graph import NetworkXGraphDB from .networkx.networkx_graph import NetworkXGraphDB
class GraphDBFactory: class GraphDBFactory:

0
cognitive_architecture/database/graphdb/networkx_graph.py → cognitive_architecture/database/graphdb/networkx/networkx_graph.py
View file

21
main.py
View file

@ -461,14 +461,15 @@ async def user_context_enrichment(
# await user_query_to_graph_db(session, user_id, query) # await user_query_to_graph_db(session, user_id, query)
semantic_mem = neo4j_graph_db.retrieve_semantic_memory(user_id=user_id) semantic_mem = await neo4j_graph_db.retrieve_semantic_memory(user_id=user_id)
await neo4j_graph_db.close() await neo4j_graph_db.close()
neo4j_graph_db = Neo4jGraphDB( neo4j_graph_db = Neo4jGraphDB(
url=config.graph_database_url, url=config.graph_database_url,
username=config.graph_database_username, username=config.graph_database_username,
password=config.graph_database_password, password=config.graph_database_password,
) )
episodic_mem = neo4j_graph_db.retrieve_episodic_memory(user_id=user_id) episodic_mem = await neo4j_graph_db.retrieve_episodic_memory(user_id=user_id)
logging.info("Episodic memory is %s", episodic_mem)
await neo4j_graph_db.close() await neo4j_graph_db.close()
# public_mem = neo4j_graph_db.retrieve_public_memory(user_id=user_id) # public_mem = neo4j_graph_db.retrieve_public_memory(user_id=user_id)
@ -581,8 +582,8 @@ async def user_context_enrichment(
context = f""" You are a memory system that uses cognitive architecture to enrich the context = f""" You are a memory system that uses cognitive architecture to enrich the
LLM context and provide better query response. LLM context and provide better query response.
You have access to the following information: You have access to the following information:
EPISODIC MEMORY: {episodic_mem[:200]} EPISODIC MEMORY: {episodic_mem}
SEMANTIC MEMORY: {semantic_mem[:200]} SEMANTIC MEMORY: {semantic_mem}
PROCEDURAL MEMORY: NULL PROCEDURAL MEMORY: NULL
SEARCH CONTEXT: The following documents provided with sources they were SEARCH CONTEXT: The following documents provided with sources they were
extracted from could be used to provide an answer {search_context} extracted from could be used to provide an answer {search_context}
@ -764,8 +765,8 @@ async def main():
class GraphQLQuery(BaseModel): class GraphQLQuery(BaseModel):
query: str query: str
gg = await user_query_to_graph_db(session, user_id, "How does cognitive architecture work?") # gg = await user_query_to_graph_db(session, user_id, "How does cognitive architecture work?")
print(gg) # print(gg)
# def cypher_statement_correcting( input: str) -> str: # def cypher_statement_correcting( input: str) -> str:
# out = aclient.chat.completions.create( # out = aclient.chat.completions.create(
@ -846,10 +847,10 @@ async def main():
# print(bb) # print(bb)
# await attach_user_to_memory(user_id=user_id, labels=['sr'], topic="PublicMemory") # await attach_user_to_memory(user_id=user_id, labels=['sr'], topic="PublicMemory")
user_id = "test_user"
# return_ = await user_context_enrichment(user_id=user_id, query="Koja je minimalna širina vrata za osobe sa invaliditetom?", session=session, memory_type="PublicMemory", generative_response=True) return_ = await user_context_enrichment(user_id=user_id, query="I need to understand what did I do yesterday?", session=session, memory_type="SemanticMemory", generative_response=True)
# print(return_) print(return_)
# aa = await relevance_feedback("I need to understand how to build a staircase in an apartment building", "PublicMemory") # aa = await relevance_feedback("I need to understand what did I do yesterday", "PublicMemory")
# print(aa) # print(aa)
# document_summary = { # document_summary = {