Test conditional retrieval

.gitignore

@@ -45,3 +45,6 @@ __pycache__/
 
 # ignore adalflow cache
 /adalflow
+
+# Ignore test files
+rag_logs.md

rag_logs.md

@@ -0,0 +1,206 @@
+
+### 2025-02-07 12:31:49
+🔍 No conversation history, not a clarification.
+
+### 2025-02-07 12:31:50
+
+#### Context Usage
+- 🔎 Retrieved new documents for query
+- 📚 Number of new documents: 20
+- 📄 Document paths:
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/evaluation.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/evaluation.rst
+  - tutorials/rag/rag.py
+  - adalflow/adalflow/core/retriever.py
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/db.rst
+
+
+### 2025-02-07 12:32:21
+
+#### Query Analysis
+- 🤔 Query: 'can you give me example usage?'
+- 📝 Result: This is a clarification
+
+
+### 2025-02-07 12:32:21
+
+#### Context Usage
+- ♻️ Reusing previous context for clarification query
+- 📚 Number of reused documents: 20
+- 📄 Document paths:
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/evaluation.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/evaluation.rst
+  - tutorials/rag/rag.py
+  - adalflow/adalflow/core/retriever.py
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/db.rst
+
+
+### 2025-02-07 12:33:21
+
+#### Query Analysis
+- 🤔 Query: 'can you use groq client instead of openai for the generation'
+- 📝 Result: This is a clarification
+
+
+### 2025-02-07 12:33:21
+
+#### Context Usage
+- ♻️ Reusing previous context for clarification query
+- 📚 Number of reused documents: 20
+- 📄 Document paths:
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/evaluation.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/evaluation.rst
+  - tutorials/rag/rag.py
+  - adalflow/adalflow/core/retriever.py
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/db.rst
+
+
+### 2025-02-07 12:33:55
+
+#### Query Analysis
+- 🤔 Query: 'provide me with the whole example'
+- 📝 Result: This is a clarification
+
+
+### 2025-02-07 12:33:55
+
+#### Context Usage
+- ♻️ Reusing previous context for clarification query
+- 📚 Number of reused documents: 20
+- 📄 Document paths:
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/evaluation.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/evaluation.rst
+  - tutorials/rag/rag.py
+  - adalflow/adalflow/core/retriever.py
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/db.rst
+
+
+### 2025-02-07 12:34:35
+
+#### Query Analysis
+- 🤔 Query: 'provide me with the whole example but  use the groq client'
+- 📝 Result: This is a clarification
+
+
+### 2025-02-07 12:34:35
+
+#### Context Usage
+- ♻️ Reusing previous context for clarification query
+- 📚 Number of reused documents: 20
+- 📄 Document paths:
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/evaluation.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/evaluation.rst
+  - tutorials/rag/rag.py
+  - adalflow/adalflow/core/retriever.py
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/retriever.rst
+  - docs/source/tutorials/rag_playbook.rst
+  - docs/source/tutorials/db.rst
+
+
+### 2025-02-07 12:35:16
+
+#### Query Analysis
+- 🤔 Query: 'can you explain me about agent compoenent.'
+- 📝 Result: This is a new question
+
+
+### 2025-02-07 12:35:17
+
+#### Context Usage
+- 🔎 Retrieved new documents for query
+- 📚 Number of new documents: 20
+- 📄 Document paths:
+  - adalflow/adalflow/components/agent/README.md
+  - docs/source/tutorials/agent.rst
+  - docs/source/tutorials/component.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/index.rst
+  - docs/source/tutorials/agent.rst
+  - docs/source/apis/components/index.rst
+  - docs/source/tutorials/index.rst
+  - adalflow/tests/test_react_agent.py
+  - docs/source/tutorials/component.rst
+  - use_cases/unsorted/simple_qa_groq.py
+  - docs/source/apis/core/index.rst
+  - docs/source/tutorials/auto_text_grad.rst
+  - adalflow/tests/test_component.py
+  - docs/source/tutorials/adalcomponent.rst
+  - use_cases/unsorted/simple_qa_memory.py
+  - docs/source/tutorials/auto_text_grad.rst
+  - docs/source/tutorials/index.rst
+  - use_cases/unsorted/simple_qa_trainable.py
+  - docs/source/tutorials/component.rst
+

src/rag.py

@@ -1,5 +1,7 @@
-from typing import Any, List
+from typing import Any, List, Tuple, Optional
 from uuid import uuid4
+import os
+from datetime import datetime
 
 import adalflow as adal
 from adalflow.core.types import (
@@ -16,6 +18,7 @@
 from config import configs
 from src.data_pipeline import DatabaseManager
 from adalflow.utils import printc
+from dataclasses import dataclass, field
 
 
 class Memory(DataComponent):
@@ -41,6 +44,14 @@ def add_dialog_turn(self, user_query: str, assistant_response: str):
         self.current_conversation.append_dialog_turn(dialog_turn)
 
 
+@dataclass
+class RAGAnswer(adal.DataClass):
+    rationale: str = field(default="", metadata={"desc": "Rationale for the answer."})
+    answer: str = field(default="", metadata={"desc": "Answer to the user query."})
+
+    __output_fields__ = ["rationale", "answer"]
+
+
 system_prompt = r"""
 You are a code assistant which answer's user question on a Github Repo. 
 You will receive user query, relevant context, and past conversation history.
@@ -75,16 +86,6 @@ def add_dialog_turn(self, user_query: str, assistant_response: str):
 <END_OF_USER_PROMPT>
 """
 
-from dataclasses import dataclass, field
-
-
-@dataclass
-class RAGAnswer(adal.DataClass):
-    rationale: str = field(default="", metadata={"desc": "Rationale for the answer."})
-    answer: str = field(default="", metadata={"desc": "Answer to the user query."})
-
-    __output_fields__ = ["rationale", "answer"]
-
 
 class RAG(adal.Component):
     __doc__ = """RAG with one repo.
@@ -119,6 +120,13 @@ def __init__(self):
             model_kwargs=configs["generator"]["model_kwargs"],
             output_processors=data_parser,
         )
+        self.previous_retrieved_documents = None
+
+        # Initialize log file
+        self.log_file = "rag_logs.md"
+        if os.path.exists(self.log_file):
+            # Clear previous logs when starting new session
+            open(self.log_file, 'w').close()
 
     def initialize_db_manager(self):
         self.db_manager = DatabaseManager()
@@ -136,15 +144,89 @@ def prepare_retriever(self, repo_url_or_path: str):
             document_map_func=lambda doc: doc.vector,
         )
 
-    def call(self, query: str) -> Any:
+    def log_to_file(self, message: str):
+        """Write log messages to file with timestamp"""
+        timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+        with open(self.log_file, 'a') as f:
+            f.write(f"\n### {timestamp}\n{message}\n")
+
+    def is_clarification_query(self, query: str) -> bool:
+        """
+        Determines if the current query is a clarification of a previous query.
+        """
+        if not self.memory():
+            self.log_to_file("🔍 No conversation history, not a clarification.")
+            return False
+
+        clarification_prompt = f"""
+        You are a clarification detector. Analyze if the query is a follow-up or clarification of the previous conversation.
+        Your response should include:
+        - A rationale explaining your reasoning
+        - A clear True/False answer
+
+        Output your response in this format:
+            {{
+            "rationale": "Your step-by-step reasoning here",
+            "answer": "True or False"
+            }}
+
+        Conversation History:
+        {self.memory()}
+
+        Query:
+        {query}
+        """
+        response = self.generator(
+            prompt_kwargs={
+                "conversation_history": self.memory(),
+                "system_prompt": clarification_prompt,
+            },
+        )
+
+        is_clarification = "true" in response.data.answer.lower()
+        log_message = f"""
+#### Query Analysis
+- 🤔 Query: '{query}'
+- 📝 Result: {'This is a clarification' if is_clarification else 'This is a new question'}
+"""
+        self.log_to_file(log_message)
+        return is_clarification
+
+    def call(self, query: str) -> Tuple[Any, Any]:
+        previous_context = (
+            self.previous_retrieved_documents[0].documents
+            if self.previous_retrieved_documents
+            else None
+        )
 
-        retrieved_documents = self.retriever(query)
+        is_clarification = self.is_clarification_query(query)
 
-        # fill in the document
-        retrieved_documents[0].documents = [
-            self.transformed_docs[doc_index]
-            for doc_index in retrieved_documents[0].doc_indices
-        ]
+        if is_clarification and self.previous_retrieved_documents:
+            retrieved_documents = self.previous_retrieved_documents
+            log_message = f"""
+#### Context Usage
+- ♻️ Reusing previous context for clarification query
+- 📚 Number of reused documents: {len(retrieved_documents[0].documents)}
+- 📄 Document paths:
+  {self._format_doc_paths(retrieved_documents[0].documents)}
+"""
+            self.log_to_file(log_message)
+        else:
+            retrieved_documents = self.retriever(query)
+            retrieved_documents[0].documents = [
+                self.transformed_docs[doc_index]
+                for doc_index in retrieved_documents[0].doc_indices
+            ]
+            self.previous_retrieved_documents = retrieved_documents
+
+            log_message = f"""
+#### Context Usage
+- 🔎 Retrieved new documents for query
+- 📚 Number of new documents: {len(retrieved_documents[0].documents)}
+- 📄 Document paths:
+  {self._format_doc_paths(retrieved_documents[0].documents)}
+"""
+            self.log_to_file(log_message)
 
         printc(f"retrieved_documents: {retrieved_documents[0].documents}")
         printc(f"memory: {self.memory()}")
@@ -168,6 +250,13 @@ def call(self, query: str) -> Any:
 
         return final_response, retrieved_documents
 
+    def _format_doc_paths(self, documents: List[Any]) -> str:
+        """Helper to format document paths for logging"""
+        return "\n  ".join(
+            f"- {doc.meta_data.get('file_path', 'unknown')}"
+            for doc in documents
+        )
+
 
 if __name__ == "__main__":
     from adalflow.utils import get_logger