Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
@@ -14,7 +14,8 @@ from langchain.text_splitter import RecursiveCharacterTextSplitter
|
|
14 |
from langchain.vectorstores import Chroma
|
15 |
from langchain.vectorstores import MongoDBAtlasVectorSearch
|
16 |
from pymongo import MongoClient
|
17 |
-
from
|
|
|
18 |
|
19 |
_ = load_dotenv(find_dotenv())
|
20 |
|
@@ -35,7 +36,7 @@ MONGODB_INDEX_NAME = "default"
|
|
35 |
LLM_CHAIN_PROMPT = PromptTemplate(input_variables = ["question"], template = os.environ["LLM_TEMPLATE"])
|
36 |
RAG_CHAIN_PROMPT = PromptTemplate(input_variables = ["context", "question"], template = os.environ["RAG_TEMPLATE"])
|
37 |
|
38 |
-
WANDB_API_KEY = os.environ["WANDB_API_KEY"]
|
39 |
|
40 |
RAG_OFF = "Off"
|
41 |
RAG_CHROMA = "Chroma"
|
@@ -115,42 +116,42 @@ def rag_chain(llm, prompt, db):
|
|
115 |
completion = rag_chain({"query": prompt})
|
116 |
return completion, rag_chain
|
117 |
|
118 |
-
def wandb_trace(rag_option, prompt, completion, result, generation_info, llm_output, chain, err_msg, start_time_ms, end_time_ms):
|
119 |
-
wandb.init(project = "openai-llm-rag")
|
120 |
-
|
121 |
-
trace = Trace(
|
122 |
-
kind = "chain",
|
123 |
-
name = "" if (chain == None) else type(chain).__name__,
|
124 |
-
status_code = "success" if (str(err_msg) == "") else "error",
|
125 |
-
status_message = str(err_msg),
|
126 |
-
metadata = {"chunk_overlap": "" if (rag_option == RAG_OFF) else config["chunk_overlap"],
|
127 |
-
"chunk_size": "" if (rag_option == RAG_OFF) else config["chunk_size"],
|
128 |
-
} if (str(err_msg) == "") else {},
|
129 |
-
inputs = {"rag_option": rag_option,
|
130 |
-
"prompt": prompt,
|
131 |
-
"chain_prompt": (str(chain.prompt) if (rag_option == RAG_OFF) else
|
132 |
-
str(chain.combine_documents_chain.llm_chain.prompt)),
|
133 |
-
"source_documents": "" if (rag_option == RAG_OFF) else str([doc.metadata["source"] for doc in completion["source_documents"]]),
|
134 |
-
} if (str(err_msg) == "") else {},
|
135 |
-
outputs = {"result": result,
|
136 |
-
"generation_info": str(generation_info),
|
137 |
-
"llm_output": str(llm_output),
|
138 |
-
"completion": str(completion),
|
139 |
-
} if (str(err_msg) == "") else {},
|
140 |
-
model_dict = {"client": (str(chain.llm.client) if (rag_option == RAG_OFF) else
|
141 |
-
str(chain.combine_documents_chain.llm_chain.llm.client)),
|
142 |
-
"model_name": (str(chain.llm.model_name) if (rag_option == RAG_OFF) else
|
143 |
-
str(chain.combine_documents_chain.llm_chain.llm.model_name)),
|
144 |
-
"temperature": (str(chain.llm.temperature) if (rag_option == RAG_OFF) else
|
145 |
-
str(chain.combine_documents_chain.llm_chain.llm.temperature)),
|
146 |
-
"retriever": ("" if (rag_option == RAG_OFF) else str(chain.retriever)),
|
147 |
-
} if (str(err_msg) == "") else {},
|
148 |
-
start_time_ms = start_time_ms,
|
149 |
-
end_time_ms = end_time_ms
|
150 |
-
)
|
151 |
-
|
152 |
-
trace.log("evaluation")
|
153 |
-
wandb.finish()
|
154 |
|
155 |
def invoke(openai_api_key, rag_option, prompt):
|
156 |
if (openai_api_key == ""):
|
|
|
14 |
from langchain.vectorstores import Chroma
|
15 |
from langchain.vectorstores import MongoDBAtlasVectorSearch
|
16 |
from pymongo import MongoClient
|
17 |
+
from trace import wandb_trace
|
18 |
+
#from wandb.sdk.data_types.trace_tree import Trace
|
19 |
|
20 |
_ = load_dotenv(find_dotenv())
|
21 |
|
|
|
36 |
LLM_CHAIN_PROMPT = PromptTemplate(input_variables = ["question"], template = os.environ["LLM_TEMPLATE"])
|
37 |
RAG_CHAIN_PROMPT = PromptTemplate(input_variables = ["context", "question"], template = os.environ["RAG_TEMPLATE"])
|
38 |
|
39 |
+
#WANDB_API_KEY = os.environ["WANDB_API_KEY"]
|
40 |
|
41 |
RAG_OFF = "Off"
|
42 |
RAG_CHROMA = "Chroma"
|
|
|
116 |
completion = rag_chain({"query": prompt})
|
117 |
return completion, rag_chain
|
118 |
|
119 |
+
#def wandb_trace(rag_option, prompt, completion, result, generation_info, llm_output, chain, err_msg, start_time_ms, end_time_ms):
|
120 |
+
# wandb.init(project = "openai-llm-rag")
|
121 |
+
#
|
122 |
+
# trace = Trace(
|
123 |
+
# kind = "chain",
|
124 |
+
# name = "" if (chain == None) else type(chain).__name__,
|
125 |
+
# status_code = "success" if (str(err_msg) == "") else "error",
|
126 |
+
# status_message = str(err_msg),
|
127 |
+
# metadata = {"chunk_overlap": "" if (rag_option == RAG_OFF) else config["chunk_overlap"],
|
128 |
+
# "chunk_size": "" if (rag_option == RAG_OFF) else config["chunk_size"],
|
129 |
+
# } if (str(err_msg) == "") else {},
|
130 |
+
# inputs = {"rag_option": rag_option,
|
131 |
+
# "prompt": prompt,
|
132 |
+
# "chain_prompt": (str(chain.prompt) if (rag_option == RAG_OFF) else
|
133 |
+
# str(chain.combine_documents_chain.llm_chain.prompt)),
|
134 |
+
# "source_documents": "" if (rag_option == RAG_OFF) else str([doc.metadata["source"] for doc in completion["source_documents"]]),
|
135 |
+
# } if (str(err_msg) == "") else {},
|
136 |
+
# outputs = {"result": result,
|
137 |
+
# "generation_info": str(generation_info),
|
138 |
+
# "llm_output": str(llm_output),
|
139 |
+
# "completion": str(completion),
|
140 |
+
# } if (str(err_msg) == "") else {},
|
141 |
+
# model_dict = {"client": (str(chain.llm.client) if (rag_option == RAG_OFF) else
|
142 |
+
# str(chain.combine_documents_chain.llm_chain.llm.client)),
|
143 |
+
# "model_name": (str(chain.llm.model_name) if (rag_option == RAG_OFF) else
|
144 |
+
# str(chain.combine_documents_chain.llm_chain.llm.model_name)),
|
145 |
+
# "temperature": (str(chain.llm.temperature) if (rag_option == RAG_OFF) else
|
146 |
+
# str(chain.combine_documents_chain.llm_chain.llm.temperature)),
|
147 |
+
# "retriever": ("" if (rag_option == RAG_OFF) else str(chain.retriever)),
|
148 |
+
# } if (str(err_msg) == "") else {},
|
149 |
+
# start_time_ms = start_time_ms,
|
150 |
+
# end_time_ms = end_time_ms
|
151 |
+
# )
|
152 |
+
#
|
153 |
+
# trace.log("evaluation")
|
154 |
+
# wandb.finish()
|
155 |
|
156 |
def invoke(openai_api_key, rag_option, prompt):
|
157 |
if (openai_api_key == ""):
|