Spaces:
Sleeping
Sleeping
Update kadiApy_ragchain.py
Browse files- kadiApy_ragchain.py +99 -38
kadiApy_ragchain.py
CHANGED
@@ -27,8 +27,12 @@ class KadiApyRagchain:
|
|
27 |
print("Start retrieving:")
|
28 |
#doc_contexts = self.retrieve_contexts(query, k=2, filter={"dataset_category": "kadi_apy_docs"})
|
29 |
#code_contexts = self.retrieve_contexts(rewritten_query, k=3, filter={"usage": code_library_usage_prediction})
|
30 |
-
|
31 |
|
|
|
|
|
|
|
|
|
32 |
# Vanilla
|
33 |
#doc_contexts = self.retrieve_contexts(query, k=3, filter={"dataset_category": "kadi_apy_docs"})
|
34 |
#code_contexts = self.retrieve_contexts(query, k=5, filter={"dataset_category": "kadi_apy_source_code"})
|
@@ -39,14 +43,13 @@ class KadiApyRagchain:
|
|
39 |
|
40 |
# Format contexts
|
41 |
print("Formatting docs:")
|
42 |
-
|
43 |
-
|
44 |
-
formatted_contexts = self.format_documents(context)
|
45 |
|
46 |
# Generate response
|
47 |
print("Start generatin repsonsse:")
|
48 |
-
|
49 |
-
response = self.generate_response(query, chat_history, formatted_contexts)
|
50 |
|
51 |
# Add the response to the existing query in the conversation history
|
52 |
#self.add_to_conversation(llm_response=response)
|
@@ -109,6 +112,38 @@ class KadiApyRagchain:
|
|
109 |
)
|
110 |
return self.llm.invoke(rewrite_prompt).content
|
111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
|
113 |
def predict_library_usage(self, query):
|
114 |
"""
|
@@ -136,7 +171,44 @@ class KadiApyRagchain:
|
|
136 |
context = self.vector_store.similarity_search(query = query, k=k, filter=filter)
|
137 |
return context
|
138 |
|
139 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
140 |
# """
|
141 |
# Generate a response using the retrieved contexts and the LLM.
|
142 |
# """
|
@@ -146,8 +218,8 @@ class KadiApyRagchain:
|
|
146 |
# prompt = f"""
|
147 |
# You are a Python programming assistant specialized in the "Kadi-APY" library.
|
148 |
# The "Kadi-APY" library is a Python package designed to facilitate interaction with the REST-like API of a software platform called Kadi4Mat.
|
149 |
-
# Your task is to answer the user's query based on the guidelines, and if needed, combine understanding provided by
|
150 |
-
# "
|
151 |
|
152 |
# Guidelines if generating code:
|
153 |
# - Display the complete code first, followed by a concise explanation in no more than 5 sentences.
|
@@ -161,11 +233,8 @@ class KadiApyRagchain:
|
|
161 |
# Chat History:
|
162 |
# {formatted_history}
|
163 |
|
164 |
-
#
|
165 |
-
# {
|
166 |
-
|
167 |
-
# Code Snippets:
|
168 |
-
# {code_context}
|
169 |
|
170 |
# Query:
|
171 |
# {query}
|
@@ -173,41 +242,33 @@ class KadiApyRagchain:
|
|
173 |
# return self.llm.invoke(prompt).content
|
174 |
|
175 |
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
# Update the prompt with history included
|
183 |
prompt = f"""
|
184 |
You are a Python programming assistant specialized in the "Kadi-APY" library.
|
185 |
The "Kadi-APY" library is a Python package designed to facilitate interaction with the REST-like API of a software platform called Kadi4Mat.
|
186 |
-
Your task is to
|
187 |
-
"Context" contains snippets from the source code and/or code examples
|
188 |
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
-
|
194 |
-
-
|
195 |
-
- If the user's query cannot be fulfilled based on the provided snippets, reply with "The API does not support the requested functionality."
|
196 |
-
- If the user's query does not implicate any task, reply with a question asking the user to elaborate.
|
197 |
-
|
198 |
-
Chat History:
|
199 |
-
{formatted_history}
|
200 |
|
201 |
Context:
|
202 |
{context}
|
203 |
-
|
204 |
-
Query:
|
205 |
-
{query}
|
206 |
"""
|
207 |
-
|
208 |
|
209 |
|
210 |
|
|
|
|
|
211 |
def format_documents(self, documents):
|
212 |
formatted_docs = []
|
213 |
for i, doc in enumerate(documents, start=1):
|
|
|
27 |
print("Start retrieving:")
|
28 |
#doc_contexts = self.retrieve_contexts(query, k=2, filter={"dataset_category": "kadi_apy_docs"})
|
29 |
#code_contexts = self.retrieve_contexts(rewritten_query, k=3, filter={"usage": code_library_usage_prediction})
|
30 |
+
code_contexts = self.retrieve_contexts(query, k=2)
|
31 |
|
32 |
+
query_formulated_question= formulate_question(code_contexts)
|
33 |
+
print("question": query_formulated_question)
|
34 |
+
doc_contexts = self.retrieve_contexts(query_formulated_question, k=2, filter={"dataset_category": "kadi_apy_docs"})
|
35 |
+
|
36 |
# Vanilla
|
37 |
#doc_contexts = self.retrieve_contexts(query, k=3, filter={"dataset_category": "kadi_apy_docs"})
|
38 |
#code_contexts = self.retrieve_contexts(query, k=5, filter={"dataset_category": "kadi_apy_source_code"})
|
|
|
43 |
|
44 |
# Format contexts
|
45 |
print("Formatting docs:")
|
46 |
+
formatted_doc_contexts = self.format_documents(doc_contexts)
|
47 |
+
formatted_code_contexts = self.format_documents(code_contexts)
|
|
|
48 |
|
49 |
# Generate response
|
50 |
print("Start generatin repsonsse:")
|
51 |
+
response = self.generate_response(query, chat_history, formatted_doc_contexts, formatted_code_contexts)
|
52 |
+
#response = self.generate_response(query, chat_history, formatted_contexts)
|
53 |
|
54 |
# Add the response to the existing query in the conversation history
|
55 |
#self.add_to_conversation(llm_response=response)
|
|
|
112 |
)
|
113 |
return self.llm.invoke(rewrite_prompt).content
|
114 |
|
115 |
+
def ask_rag(self, context):
|
116 |
+
rewrite_prompt = (
|
117 |
+
f""" "Code snippets" consists of methods and/or class declaration from the source code of an API Wrapper Python library.
|
118 |
+
Your task is to write a question how to use
|
119 |
+
|
120 |
+
Guidelines for rewriting the query:
|
121 |
+
1. Identify the main action the user wants to perform (e.g., "Upload a file to a record," "Get users of a group").
|
122 |
+
2. Remove conversational elements like greetings or pleasantries (e.g., "Hello Chatbot", "I need you to help me with").
|
123 |
+
3. Exclude specific variable values (e.g., "ID of my record is '31'") unless essential to the intent.
|
124 |
+
4. Rephrase the query to match the format and keywords used in the docstrings, focusing on verbs and objects relevant to the action (e.g., "Add a record to a collection").
|
125 |
+
5. Given the query the user might need more than one action to achieve his goal. In this case the rewritten query has more than one action.
|
126 |
+
|
127 |
+
Examples:
|
128 |
+
- User query: "Create a Python script with a method that facilitates the creation of records. This method should accept an array of identifiers as a parameter and allow metadata to be added to each record."
|
129 |
+
- Rewritten query: "create records, add metadata to record"
|
130 |
+
- User query: "Hi, can you help me write Python code to add a record to a collection? The record ID is '45', and the collection ID is '12'."
|
131 |
+
Rewritten query: "add a record to a collection"
|
132 |
+
- User query: I need a python script with which i create a new record with the title: "Hello World" and then link the record to a given collection.
|
133 |
+
Rewritten query: "create a new record with title" , "link a record to a collection"
|
134 |
+
|
135 |
+
Based on these examples and guidelines, rewrite the following user query to align more effectively with the keywords used in the docstrings.
|
136 |
+
Do not include any addition comments, explanations, or text.
|
137 |
+
|
138 |
+
Original query:
|
139 |
+
{query}
|
140 |
+
"""
|
141 |
+
)
|
142 |
+
return self.llm.invoke(rewrite_prompt).content
|
143 |
+
|
144 |
+
|
145 |
+
|
146 |
+
|
147 |
|
148 |
def predict_library_usage(self, query):
|
149 |
"""
|
|
|
171 |
context = self.vector_store.similarity_search(query = query, k=k, filter=filter)
|
172 |
return context
|
173 |
|
174 |
+
def generate_response(self, query, chat_history, doc_context, code_context):
|
175 |
+
"""
|
176 |
+
Generate a response using the retrieved contexts and the LLM.
|
177 |
+
"""
|
178 |
+
formatted_history = self.format_history(chat_history)
|
179 |
+
|
180 |
+
# Update the prompt with history included
|
181 |
+
prompt = f"""
|
182 |
+
You are a Python programming assistant specialized in the "Kadi-APY" library.
|
183 |
+
The "Kadi-APY" library is a Python package designed to facilitate interaction with the REST-like API of a software platform called Kadi4Mat.
|
184 |
+
Your task is to answer the user's query based on the guidelines, and if needed, combine understanding provided by
|
185 |
+
"Document Snippets" with the implementation details provided by "Code Snippets."
|
186 |
+
|
187 |
+
Guidelines if generating code:
|
188 |
+
- Display the complete code first, followed by a concise explanation in no more than 5 sentences.
|
189 |
+
|
190 |
+
General Guidelines:
|
191 |
+
- Refer to the "Chat History" if it provides context that could enhance your understanding of the user's query.
|
192 |
+
- Always include the "Chat History" if relevant to the user's query for continuity and clarity in responses.
|
193 |
+
- If the user's query cannot be fulfilled based on the provided snippets, reply with "The API does not support the requested functionality."
|
194 |
+
- If the user's query does not implicate any task, reply with a question asking the user to elaborate.
|
195 |
+
|
196 |
+
Chat History:
|
197 |
+
{formatted_history}
|
198 |
+
|
199 |
+
Document Snippets:
|
200 |
+
{doc_context}
|
201 |
+
|
202 |
+
Code Snippets:
|
203 |
+
{code_context}
|
204 |
+
|
205 |
+
Query:
|
206 |
+
{query}
|
207 |
+
"""
|
208 |
+
return self.llm.invoke(prompt).content
|
209 |
+
|
210 |
+
|
211 |
+
# def generate_response(self, query, chat_history, context):
|
212 |
# """
|
213 |
# Generate a response using the retrieved contexts and the LLM.
|
214 |
# """
|
|
|
218 |
# prompt = f"""
|
219 |
# You are a Python programming assistant specialized in the "Kadi-APY" library.
|
220 |
# The "Kadi-APY" library is a Python package designed to facilitate interaction with the REST-like API of a software platform called Kadi4Mat.
|
221 |
+
# Your task is to answer the user's query based on the guidelines, and if needed, combine understanding provided by "Context"
|
222 |
+
# "Context" contains snippets from the source code and/or code examples
|
223 |
|
224 |
# Guidelines if generating code:
|
225 |
# - Display the complete code first, followed by a concise explanation in no more than 5 sentences.
|
|
|
233 |
# Chat History:
|
234 |
# {formatted_history}
|
235 |
|
236 |
+
# Context:
|
237 |
+
# {context}
|
|
|
|
|
|
|
238 |
|
239 |
# Query:
|
240 |
# {query}
|
|
|
242 |
# return self.llm.invoke(prompt).content
|
243 |
|
244 |
|
245 |
+
|
246 |
+
|
247 |
+
def formulate_question(self, source_code):
|
248 |
+
"""
|
249 |
+
Generate a response using the retrieved contexts and the LLM.
|
250 |
+
"""
|
|
|
251 |
prompt = f"""
|
252 |
You are a Python programming assistant specialized in the "Kadi-APY" library.
|
253 |
The "Kadi-APY" library is a Python package designed to facilitate interaction with the REST-like API of a software platform called Kadi4Mat.
|
254 |
+
Your task is to formulate the next logical question a programmer would ask themselves to implement and run the method provided in the "context".
|
|
|
255 |
|
256 |
+
"Context" contains snippets from the source code and metadata that provide details about the method.
|
257 |
+
|
258 |
+
Guidelines for generating questions:
|
259 |
+
- The question should be specific to the programmer's intent of using the method within a Python script.
|
260 |
+
- Focus on determining the entry point of the class to which the method belongs.
|
261 |
+
- Avoid vague or general questions; be precise about the next actionable steps.
|
|
|
|
|
|
|
|
|
|
|
262 |
|
263 |
Context:
|
264 |
{context}
|
|
|
|
|
|
|
265 |
"""
|
266 |
+
return self.llm.invoke(prompt).content
|
267 |
|
268 |
|
269 |
|
270 |
+
|
271 |
+
|
272 |
def format_documents(self, documents):
|
273 |
formatted_docs = []
|
274 |
for i, doc in enumerate(documents, start=1):
|