gardel/lightrag
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Logic Optimization

Browse Source
This commit is contained in:
jin
2024-11-25 13:40:38 +08:00
parent 1dbe803521
commit 26ae240c65
8 changed files with 185 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
examples/lightrag_api_oracle_demo..py
View File

@@ -1,16 +1,14 @@
from fastapi import FastAPI, HTTPException, File, UploadFile from fastapi import FastAPI, HTTPException, File, UploadFile
from fastapi import Query from fastapi import Query
from contextlib import asynccontextmanager from contextlib import asynccontextmanager
from pydantic import BaseModel from pydantic import BaseModel
from typing import Optional, Any from typing import Optional, Any
from fastapi.responses import JSONResponse
import sys, os import sys
print(os.getcwd()) import os
from pathlib import Path from pathlib import Path
script_directory = Path(__file__).resolve().parent.parent
sys.path.append(os.path.abspath(script_directory))
import asyncio import asyncio
import nest_asyncio import nest_asyncio
@@ -18,10 +16,12 @@ from lightrag import LightRAG, QueryParam
from lightrag.llm import openai_complete_if_cache, openai_embedding from lightrag.llm import openai_complete_if_cache, openai_embedding
from lightrag.utils import EmbeddingFunc from lightrag.utils import EmbeddingFunc
import numpy as np import numpy as np
from datetime import datetime
from lightrag.kg.oracle_impl import OracleDB from lightrag.kg.oracle_impl import OracleDB
print(os.getcwd())
script_directory = Path(__file__).resolve().parent.parent
sys.path.append(os.path.abspath(script_directory))
# Apply nest_asyncio to solve event loop issues # Apply nest_asyncio to solve event loop issues
@@ -48,6 +48,7 @@ print(f"EMBEDDING_MAX_TOKEN_SIZE: {EMBEDDING_MAX_TOKEN_SIZE}")
if not os.path.exists(WORKING_DIR): if not os.path.exists(WORKING_DIR):
os.mkdir(WORKING_DIR) os.mkdir(WORKING_DIR)
async def llm_model_func( async def llm_model_func(
prompt, system_prompt=None, history_messages=[], **kwargs prompt, system_prompt=None, history_messages=[], **kwargs
) -> str: ) -> str:
@@ -77,8 +78,8 @@ async def get_embedding_dim():
embedding_dim = embedding.shape[1] embedding_dim = embedding.shape[1]
return embedding_dim return embedding_dim
async def init():
async def init():
# Detect embedding dimension # Detect embedding dimension
embedding_dimension = 1024 # await get_embedding_dim() embedding_dimension = 1024 # await get_embedding_dim()
print(f"Detected embedding dimension: {embedding_dimension}") print(f"Detected embedding dimension: {embedding_dimension}")
@@ -88,15 +89,15 @@ async def init():
# We storage data in unified tables, so we need to set a `workspace` parameter to specify which docs we want to store and query # We storage data in unified tables, so we need to set a `workspace` parameter to specify which docs we want to store and query
# Below is an example of how to connect to Oracle Autonomous Database on Oracle Cloud # Below is an example of how to connect to Oracle Autonomous Database on Oracle Cloud
oracle_db = OracleDB(
oracle_db = OracleDB(config={ config={
"user": "", "user": "",
"password": "", "password": "",
"dsn": "", "dsn": "",
"config_dir": "path_to_config_dir", "config_dir": "path_to_config_dir",
"wallet_location": "path_to_wallet_location", "wallet_location": "path_to_wallet_location",
"wallet_password": "wallet_password", "wallet_password": "wallet_password",
"workspace":"company" "workspace": "company",
} # specify which docs you want to store and query } # specify which docs you want to store and query
) )
@@ -116,7 +117,7 @@ async def init():
), ),
graph_storage="OracleGraphStorage", graph_storage="OracleGraphStorage",
kv_storage="OracleKVStorage", kv_storage="OracleKVStorage",
vector_storage="OracleVectorDBStorage" vector_storage="OracleVectorDBStorage",
) )
# Setthe KV/vector/graph storage's `db` property, so all operation will use same connection pool # Setthe KV/vector/graph storage's `db` property, so all operation will use same connection pool
@@ -147,9 +148,11 @@ class QueryRequest(BaseModel):
only_need_context: bool = False only_need_context: bool = False
only_need_prompt: bool = False only_need_prompt: bool = False
class DataRequest(BaseModel): class DataRequest(BaseModel):
limit: int = 100 limit: int = 100
class InsertRequest(BaseModel): class InsertRequest(BaseModel):
text: str text: str
@@ -164,6 +167,7 @@ class Response(BaseModel):
rag = None rag = None
@asynccontextmanager @asynccontextmanager
async def lifespan(app: FastAPI): async def lifespan(app: FastAPI):
global rag global rag
@@ -172,7 +176,10 @@ async def lifespan(app: FastAPI):
yield yield
app = FastAPI(title="LightRAG API", description="API for RAG operations",lifespan=lifespan) app = FastAPI(
title="LightRAG API", description="API for RAG operations", lifespan=lifespan
)
@app.post("/query", response_model=Response) @app.post("/query", response_model=Response)
async def query_endpoint(request: QueryRequest): async def query_endpoint(request: QueryRequest):
@@ -188,7 +195,7 @@ async def query_endpoint(request: QueryRequest):
mode=request.mode, mode=request.mode,
only_need_context=request.only_need_context, only_need_context=request.only_need_context,
only_need_prompt=request.only_need_prompt, only_need_prompt=request.only_need_prompt,
top_k=top_k top_k=top_k,
), ),
) )
return Response(status="success", data=result) return Response(status="success", data=result)

1
examples/lightrag_oracle_demo.py
View File

@@ -97,7 +97,6 @@ async def main():
graph_storage="OracleGraphStorage", graph_storage="OracleGraphStorage",
kv_storage="OracleKVStorage", kv_storage="OracleKVStorage",
vector_storage="OracleVectorDBStorage", vector_storage="OracleVectorDBStorage",
addon_params={"example_number": 1, "language": "Simplfied Chinese"}, addon_params={"example_number": 1, "language": "Simplfied Chinese"},
) )

8
lightrag/kg/oracle_impl.py
View File

@@ -114,7 +114,9 @@ class OracleDB:
logger.info("Finished check all tables in Oracle database") logger.info("Finished check all tables in Oracle database")
async def query(self, sql: str, multirows: bool = False) -> Union[dict, None]: async def query(
self, sql: str, params: dict = None, multirows: bool = False
) -> Union[dict, None]:
async with self.pool.acquire() as connection: async with self.pool.acquire() as connection:
connection.inputtypehandler = self.input_type_handler connection.inputtypehandler = self.input_type_handler
connection.outputtypehandler = self.output_type_handler connection.outputtypehandler = self.output_type_handler
@@ -256,7 +258,7 @@ class OracleKVStorage(BaseKVStorage):
item["__vector__"], item["__vector__"],
] ]
# print(merge_sql) # print(merge_sql)
await self.db.execute(merge_sql, data) await self.db.execute(merge_sql, values)
if self.namespace == "full_docs": if self.namespace == "full_docs":
for k, v in self._data.items(): for k, v in self._data.items():
@@ -266,7 +268,7 @@ class OracleKVStorage(BaseKVStorage):
) )
values = [k, self._data[k]["content"], self.db.workspace] values = [k, self._data[k]["content"], self.db.workspace]
# print(merge_sql) # print(merge_sql)
await self.db.execute(merge_sql, data) await self.db.execute(merge_sql, values)
return left_data return left_data
async def index_done_callback(self): async def index_done_callback(self):

6
lightrag/llm.py
View File

@@ -70,8 +70,8 @@ async def openai_complete_if_cache(
model=model, messages=messages, **kwargs model=model, messages=messages, **kwargs
) )
content = response.choices[0].message.content content = response.choices[0].message.content
if r'\u' in content: if r"\u" in content:
content = content.encode('utf-8').decode('unicode_escape') content = content.encode("utf-8").decode("unicode_escape")
print(content) print(content)
if hashing_kv is not None: if hashing_kv is not None:
await hashing_kv.upsert( await hashing_kv.upsert(
@@ -542,7 +542,7 @@ async def openai_embedding(
texts: list[str], texts: list[str],
model: str = "text-embedding-3-small", model: str = "text-embedding-3-small",
base_url: str = None, base_url: str = None,
api_key: str = None api_key: str = None,
) -> np.ndarray: ) -> np.ndarray:
if api_key: if api_key:
os.environ["OPENAI_API_KEY"] = api_key os.environ["OPENAI_API_KEY"] = api_key

76
lightrag/operate.py
View File

@@ -249,10 +249,14 @@ async def extract_entities(
ordered_chunks = list(chunks.items()) ordered_chunks = list(chunks.items())
# add language and example number params to prompt # add language and example number params to prompt
language = global_config["addon_params"].get("language",PROMPTS["DEFAULT_LANGUAGE"]) language = global_config["addon_params"].get(
"language", PROMPTS["DEFAULT_LANGUAGE"]
)
example_number = global_config["addon_params"].get("example_number", None) example_number = global_config["addon_params"].get("example_number", None)
if example_number and example_number < len(PROMPTS["entity_extraction_examples"]): if example_number and example_number < len(PROMPTS["entity_extraction_examples"]):
examples="\n".join(PROMPTS["entity_extraction_examples"][:int(example_number)]) examples = "\n".join(
PROMPTS["entity_extraction_examples"][: int(example_number)]
)
else: else:
examples = "\n".join(PROMPTS["entity_extraction_examples"]) examples = "\n".join(PROMPTS["entity_extraction_examples"])
@@ -263,7 +267,8 @@ async def extract_entities(
completion_delimiter=PROMPTS["DEFAULT_COMPLETION_DELIMITER"], completion_delimiter=PROMPTS["DEFAULT_COMPLETION_DELIMITER"],
entity_types=",".join(PROMPTS["DEFAULT_ENTITY_TYPES"]), entity_types=",".join(PROMPTS["DEFAULT_ENTITY_TYPES"]),
examples=examples, examples=examples,
language=language) language=language,
)
continue_prompt = PROMPTS["entiti_continue_extraction"] continue_prompt = PROMPTS["entiti_continue_extraction"]
if_loop_prompt = PROMPTS["entiti_if_loop_extraction"] if_loop_prompt = PROMPTS["entiti_if_loop_extraction"]
@@ -396,6 +401,7 @@ async def extract_entities(
return knowledge_graph_inst return knowledge_graph_inst
async def kg_query( async def kg_query(
query, query,
knowledge_graph_inst: BaseGraphStorage, knowledge_graph_inst: BaseGraphStorage,
@@ -408,7 +414,9 @@ async def kg_query(
context = None context = None
example_number = global_config["addon_params"].get("example_number", None) example_number = global_config["addon_params"].get("example_number", None)
if example_number and example_number < len(PROMPTS["keywords_extraction_examples"]): if example_number and example_number < len(PROMPTS["keywords_extraction_examples"]):
examples = "\n".join(PROMPTS["keywords_extraction_examples"][:int(example_number)]) examples = "\n".join(
PROMPTS["keywords_extraction_examples"][: int(example_number)]
)
else: else:
examples = "\n".join(PROMPTS["keywords_extraction_examples"]) examples = "\n".join(PROMPTS["keywords_extraction_examples"])
@@ -422,7 +430,7 @@ async def kg_query(
kw_prompt_temp = PROMPTS["keywords_extraction"] kw_prompt_temp = PROMPTS["keywords_extraction"]
kw_prompt = kw_prompt_temp.format(query=query, examples=examples) kw_prompt = kw_prompt_temp.format(query=query, examples=examples)
result = await use_model_func(kw_prompt) result = await use_model_func(kw_prompt)
logger.info(f"kw_prompt result:") logger.info("kw_prompt result:")
print(result) print(result)
try: try:
json_text = locate_json_string_body_from_string(result) json_text = locate_json_string_body_from_string(result)
@@ -500,40 +508,68 @@ async def _build_query_context(
ll_kewwords, hl_keywrds = query[0], query[1] ll_kewwords, hl_keywrds = query[0], query[1]
if query_param.mode in ["local", "hybrid"]: if query_param.mode in ["local", "hybrid"]:
if ll_kewwords == "": if ll_kewwords == "":
ll_entities_context,ll_relations_context,ll_text_units_context = "","","" ll_entities_context, ll_relations_context, ll_text_units_context = (
warnings.warn("Low Level context is None. Return empty Low entity/relationship/source") "",
"",
"",
)
warnings.warn(
"Low Level context is None. Return empty Low entity/relationship/source"
)
query_param.mode = "global" query_param.mode = "global"
else: else:
ll_entities_context,ll_relations_context,ll_text_units_context = await _get_node_data( (
ll_entities_context,
ll_relations_context,
ll_text_units_context,
) = await _get_node_data(
ll_kewwords, ll_kewwords,
knowledge_graph_inst, knowledge_graph_inst,
entities_vdb, entities_vdb,
text_chunks_db, text_chunks_db,
query_param query_param,
) )
if query_param.mode in ["global", "hybrid"]: if query_param.mode in ["global", "hybrid"]:
if hl_keywrds == "": if hl_keywrds == "":
hl_entities_context,hl_relations_context,hl_text_units_context = "","","" hl_entities_context, hl_relations_context, hl_text_units_context = (
warnings.warn("High Level context is None. Return empty High entity/relationship/source") "",
"",
"",
)
warnings.warn(
"High Level context is None. Return empty High entity/relationship/source"
)
query_param.mode = "local" query_param.mode = "local"
else: else:
hl_entities_context,hl_relations_context,hl_text_units_context = await _get_edge_data( (
hl_entities_context,
hl_relations_context,
hl_text_units_context,
) = await _get_edge_data(
hl_keywrds, hl_keywrds,
knowledge_graph_inst, knowledge_graph_inst,
relationships_vdb, relationships_vdb,
text_chunks_db, text_chunks_db,
query_param query_param,
) )
if query_param.mode == 'hybrid': if query_param.mode == "hybrid":
entities_context, relations_context, text_units_context = combine_contexts( entities_context, relations_context, text_units_context = combine_contexts(
[hl_entities_context, ll_entities_context], [hl_entities_context, ll_entities_context],
[hl_relations_context, ll_relations_context], [hl_relations_context, ll_relations_context],
[hl_text_units_context,ll_text_units_context] [hl_text_units_context, ll_text_units_context],
)
elif query_param.mode == "local":
entities_context, relations_context, text_units_context = (
ll_entities_context,
ll_relations_context,
ll_text_units_context,
)
elif query_param.mode == "global":
entities_context, relations_context, text_units_context = (
hl_entities_context,
hl_relations_context,
hl_text_units_context,
) )
elif query_param.mode == 'local':
entities_context,relations_context,text_units_context = ll_entities_context,ll_relations_context,ll_text_units_context
elif query_param.mode == 'global':
entities_context,relations_context,text_units_context = hl_entities_context,hl_relations_context,hl_text_units_context
return f""" return f"""
# -----Entities----- # -----Entities-----
# ```csv # ```csv
@@ -550,7 +586,6 @@ async def _build_query_context(
# """ # """
async def _get_node_data( async def _get_node_data(
query, query,
knowledge_graph_inst: BaseGraphStorage, knowledge_graph_inst: BaseGraphStorage,
@@ -824,7 +859,6 @@ async def _get_edge_data(
return entities_context, relations_context, text_units_context return entities_context, relations_context, text_units_context
async def _find_most_related_entities_from_relationships( async def _find_most_related_entities_from_relationships(
edge_datas: list[dict], edge_datas: list[dict],
query_param: QueryParam, query_param: QueryParam,

16
lightrag/prompt.py
View File

@@ -121,10 +121,12 @@ Output:
("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"Humanity's Response"{tuple_delimiter}"Alex and his team are the key figures in Humanity's Response to the unknown intelligence."{tuple_delimiter}"collective action, cosmic significance"{tuple_delimiter}8){record_delimiter} ("relationship"{tuple_delimiter}"Alex"{tuple_delimiter}"Humanity's Response"{tuple_delimiter}"Alex and his team are the key figures in Humanity's Response to the unknown intelligence."{tuple_delimiter}"collective action, cosmic significance"{tuple_delimiter}8){record_delimiter}
("relationship"{tuple_delimiter}"Control"{tuple_delimiter}"Intelligence"{tuple_delimiter}"The concept of Control is challenged by the Intelligence that writes its own rules."{tuple_delimiter}"power dynamics, autonomy"{tuple_delimiter}7){record_delimiter} ("relationship"{tuple_delimiter}"Control"{tuple_delimiter}"Intelligence"{tuple_delimiter}"The concept of Control is challenged by the Intelligence that writes its own rules."{tuple_delimiter}"power dynamics, autonomy"{tuple_delimiter}7){record_delimiter}
("content_keywords"{tuple_delimiter}"first contact, control, communication, cosmic significance"){completion_delimiter} ("content_keywords"{tuple_delimiter}"first contact, control, communication, cosmic significance"){completion_delimiter}
#############################""" #############################""",
] ]
PROMPTS["summarize_entity_descriptions"] = """You are a helpful assistant responsible for generating a comprehensive summary of the data provided below. PROMPTS[
"summarize_entity_descriptions"
] = """You are a helpful assistant responsible for generating a comprehensive summary of the data provided below.
Given one or two entities, and a list of descriptions, all related to the same entity or group of entities. Given one or two entities, and a list of descriptions, all related to the same entity or group of entities.
Please concatenate all of these into a single, comprehensive description. Make sure to include information collected from all the descriptions. Please concatenate all of these into a single, comprehensive description. Make sure to include information collected from all the descriptions.
If the provided descriptions are contradictory, please resolve the contradictions and provide a single, coherent summary. If the provided descriptions are contradictory, please resolve the contradictions and provide a single, coherent summary.
@@ -139,10 +141,14 @@ Description List: {description_list}
Output: Output:
""" """
PROMPTS["entiti_continue_extraction"] = """MANY entities were missed in the last extraction. Add them below using the same format: PROMPTS[
"entiti_continue_extraction"
] = """MANY entities were missed in the last extraction. Add them below using the same format:
""" """
PROMPTS["entiti_if_loop_extraction"] = """It appears some entities may have still been missed. Answer YES | NO if there are still entities that need to be added. PROMPTS[
"entiti_if_loop_extraction"
] = """It appears some entities may have still been missed. Answer YES | NO if there are still entities that need to be added.
""" """
PROMPTS["fail_response"] = "Sorry, I'm not able to provide an answer to that question." PROMPTS["fail_response"] = "Sorry, I'm not able to provide an answer to that question."
@@ -230,7 +236,7 @@ Output:
"high_level_keywords": ["Education", "Poverty reduction", "Socioeconomic development"], "high_level_keywords": ["Education", "Poverty reduction", "Socioeconomic development"],
"low_level_keywords": ["School access", "Literacy rates", "Job training", "Income inequality"] "low_level_keywords": ["School access", "Literacy rates", "Job training", "Income inequality"]
}} }}
#############################""" #############################""",
] ]

3
lightrag/utils.py
View File

@@ -56,7 +56,8 @@ def locate_json_string_body_from_string(content: str) -> Union[str, None]:
maybe_json_str = maybe_json_str.replace("'", '"') maybe_json_str = maybe_json_str.replace("'", '"')
json.loads(maybe_json_str) json.loads(maybe_json_str)
return maybe_json_str return maybe_json_str
except: except Exception:
pass
# try: # try:
# content = ( # content = (
# content.replace(kw_prompt[:-1], "") # content.replace(kw_prompt[:-1], "")