improve selenium page retriever, move bshr running example to example…

…s, fix other examples, update readme, refactor, remove asserts

.gitignore

@@ -620,4 +620,4 @@ MigrationBackup/
 # dotenv
 .env
 
-*.db
+examples/output

README.md

@@ -57,7 +57,7 @@ At the heart of ChatFlock is the Conductor, a novel entity that determines the s
 - [Multi-Participant Chat with a Composition Generator with a deep (3+ level) Hierarchical Composition](examples/automatic_hierarchical_chat_composition.py)
 
 #### End-to-End Examples
-- [BSHR (Brainstorm-Search-Hypothesize-Refine) Loop](chatflock/use_cases/bshr.py) - Based on [David Shapiro's](https://github.com/daveshap/BSHR_Loop) idea.
+- [BSHR (Brainstorm-Search-Hypothesize-Refine) Loop](examples/bshr_loop.py) - Based on [David Shapiro's](https://github.com/daveshap/BSHR_Loop) idea.
 
 ## 🚀 Features
 

chatflock/ai_utils.py

@@ -23,11 +23,12 @@ def execute_chat_model_messages(
 ) -> str:
     chat_model_args = chat_model_args or {}
 
-    assert "functions" not in chat_model_args, (
-        "The `functions` argument is reserved for the "
-        "`execute_chat_model_messages` function. If you want to add more "
-        "functions use the `functions` argument to this method."
-    )
+    if "functions" in chat_model_args:
+        raise ValueError(
+            "The `functions` argument is reserved for the "
+            "`execute_chat_model_messages` function. If you want to add more "
+            "functions use the `functions` argument to this method."
+        )
 
     if tools is not None and len(tools) > 0:
         chat_model_args["functions"] = [format_tool_to_openai_function(tool) for tool in tools]

chatflock/base.py

@@ -250,9 +250,8 @@ def __init__(
         max_total_messages: Optional[int] = None,
         hide_messages: bool = False,
     ):
-        assert max_total_messages is None or max_total_messages > 0, (
-            "Max total messages must be None or greater than " "0."
-        )
+        if max_total_messages is not None and max_total_messages <= 0:
+            raise ValueError("Max total messages must be None or greater than 0.")
 
         self.backing_store = backing_store
         self.renderer = renderer

chatflock/structured_string.py

@@ -37,18 +37,22 @@ def to_text(self, level: int = 0) -> str:
 class StructuredString:
     sections: List[Section]
 
-    def __getitem__(self, item):
-        assert isinstance(item, str)
+    def __getitem__(self, item) -> Section:
+        if not isinstance(item, str):
+            raise TypeError(f"Item must be of type str, not {type(item)}.")
 
         relevant_sections = [section for section in self.sections if section.name == item]
         if len(relevant_sections) == 0:
             raise KeyError(f"No section with name {item} exists.")
 
         return relevant_sections[0]
 
-    def __setitem__(self, key, value):
-        assert isinstance(key, str)
-        assert isinstance(value, Section)
+    def __setitem__(self, key, value) -> None:
+        if not isinstance(key, str):
+            raise TypeError(f"Key must be of type str, not {type(key)}.")
+
+        if not isinstance(value, Section):
+            raise TypeError(f"Value must be of type Section, not {type(value)}.")
 
         try:
             section = self[key]
@@ -63,5 +67,5 @@ def __str__(self) -> str:
 
         return result
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         return self.__str__()

chatflock/use_cases/bshr.py

@@ -3,21 +3,14 @@
 
 import datetime
 import json
-import os
 from functools import partial
-from pathlib import Path
 
 import questionary
-from dotenv import load_dotenv
 from halo import Halo
-from langchain.cache import SQLiteCache
 from langchain.callbacks.manager import CallbackManagerForToolRun
-from langchain.chat_models import ChatOpenAI
 from langchain.chat_models.base import BaseChatModel
-from langchain.globals import set_llm_cache
 from langchain.llms.openai import OpenAI
 from langchain.memory import ConversationSummaryBufferMemory
-from langchain.text_splitter import TokenTextSplitter
 from langchain.tools import BaseTool
 from pydantic import BaseModel, Field
 
@@ -33,9 +26,6 @@
 from chatflock.structured_string import Section, StructuredString
 from chatflock.use_cases.request_response import get_response
 from chatflock.web_research import WebSearch
-from chatflock.web_research.page_analyzer import OpenAIChatPageQueryAnalyzer
-from chatflock.web_research.page_retrievers.selenium_retriever import SeleniumPageRetriever
-from chatflock.web_research.search import GoogleSerperSearchResultsProvider
 from chatflock.web_research.web_research import WebResearchTool
 
 
@@ -63,7 +53,7 @@ def load_state(state_file: Optional[str]) -> Optional[BHSRState]:
         return None
 
     try:
-        with open(state_file, "r") as f:
+        with open(state_file) as f:
             data = json.load(f)
             return BHSRState.model_validate(data)
     except FileNotFoundError:
@@ -445,7 +435,7 @@ def run_brainstorm_search_hypothesize_refine_loop(
                 break
 
             has_feedback = questionary.confirm(
-                "The information need seems to have have been satisficed. Do you have " "any feedback?"
+                "The information need seems to have have been satisficed. Do you have any feedback?"
             ).ask()
 
             if not has_feedback:
@@ -495,51 +485,3 @@ def _run(self, query: str, run_manager: Optional[CallbackManagerForToolRun] = No
         )
 
         return hypothesis
-
-
-if __name__ == "__main__":
-    load_dotenv()
-
-    output_dir = Path(os.getenv("OUTPUT_DIR", "../../output"))
-    output_dir.mkdir(exist_ok=True, parents=True)
-
-    n_search_results = 2
-
-    state_file = str(output_dir / "bshr_state.json")
-    llm_cache = SQLiteCache(database_path=str(output_dir / "llm_cache.db"))
-    set_llm_cache(llm_cache)
-
-    chat_model = ChatOpenAI(temperature=0.0, model="gpt-4-1106-preview")
-    chat_model_for_analysis = ChatOpenAI(
-        temperature=0.0,
-        model="gpt-3.5-turbo-1106",
-    )
-
-    try:
-        max_context_size = OpenAI.modelname_to_contextsize(chat_model_for_analysis.model_name)
-    except ValueError:
-        max_context_size = 12000
-
-    web_search = WebSearch(
-        chat_model=chat_model,
-        search_results_provider=GoogleSerperSearchResultsProvider(),
-        page_query_analyzer=OpenAIChatPageQueryAnalyzer(
-            chat_model=chat_model_for_analysis,
-            page_retriever=SeleniumPageRetriever(),
-            text_splitter=TokenTextSplitter(chunk_size=max_context_size, chunk_overlap=max_context_size // 5),
-            use_first_split_only=True,
-        ),
-    )
-
-    spinner = Halo(spinner="dots")
-
-    hypothesis = run_brainstorm_search_hypothesize_refine_loop(
-        confirm_satisficed=True,
-        web_search=web_search,
-        chat_model=chat_model,
-        n_search_results=n_search_results,
-        state_file=state_file,
-        spinner=spinner,
-    )
-
-    print(f"Final Answer:\n----------------\n{hypothesis}\n----------------")

chatflock/web_research/__init__.py

@@ -2,4 +2,10 @@
 from .page_analyzer import OpenAIChatPageQueryAnalyzer, PageQueryAnalyzer
 from .web_research import WebSearch
 
-__all__ = ["WebSearch", "PageQueryAnalyzer", "TransientHTTPError", "NonTransientHTTPError"]
+__all__ = [
+    "WebSearch",
+    "PageQueryAnalyzer",
+    "OpenAIChatPageQueryAnalyzer",
+    "TransientHTTPError",
+    "NonTransientHTTPError",
+]

chatflock/web_research/page_analyzer.py

@@ -8,15 +8,11 @@
 from langchain.text_splitter import TextSplitter
 from pydantic import BaseModel
 
-from chatflock.backing_stores import InMemoryChatDataBackingStore
-from chatflock.base import Chat
-from chatflock.conductors import RoundRobinChatConductor
 from chatflock.parsing_utils import string_output_to_pydantic
-from chatflock.renderers import NoChatRenderer
 from chatflock.structured_string import Section, StructuredString
 
 from ..participants.langchain import LangChainBasedAIChatParticipant
-from ..participants.user import UserChatParticipant
+from ..use_cases.request_response import get_response
 from .errors import NonTransientHTTPError, TransientHTTPError
 from .page_retrievers import PageRetriever
 
@@ -84,6 +80,8 @@ def analyze(self, url: str, title: str, query: str, spinner: Optional[Halo] = No
             return PageQueryAnalysisResult(
                 answer=f"The query could not be answered because an error occurred while retrieving the page: {e}"
             )
+        finally:
+            self.page_retriever.close()
 
         cleaned_html = clean_html(html)
 
@@ -92,71 +90,65 @@ def analyze(self, url: str, title: str, query: str, spinner: Optional[Halo] = No
         answer = "No answer yet."
         for i, doc in enumerate(docs):
             text = doc.page_content
-            chat = Chat(
-                backing_store=InMemoryChatDataBackingStore(),
-                renderer=NoChatRenderer(),
-                initial_participants=[
-                    UserChatParticipant(),
-                    LangChainBasedAIChatParticipant(
-                        name="Web Page Query Answerer",
-                        role="Web Page Query Answerer",
-                        personal_mission="Answer queries based on provided (partial) web page content from the web.",
-                        chat_model=self.chat_model,
-                        other_prompt_sections=[
+
+            query_answerer = LangChainBasedAIChatParticipant(
+                name="Web Page Query Answerer",
+                role="Web Page Query Answerer",
+                personal_mission="Answer queries based on provided (partial) web page content from the web.",
+                chat_model=self.chat_model,
+                other_prompt_sections=[
+                    Section(
+                        name="Crafting a Query Answer",
+                        sub_sections=[
+                            Section(
+                                name="Process",
+                                list=[
+                                    "Analyze the query and the given content",
+                                    "If context is provided, use it to answer the query.",
+                                    "Summarize the answer in a comprehensive, yet succinct way.",
+                                ],
+                                list_item_prefix=None,
+                            ),
                             Section(
-                                name="Crafting a Query Answer",
-                                sub_sections=[
-                                    Section(
-                                        name="Process",
-                                        list=[
-                                            "Analyze the query and the given content",
-                                            "If context is provided, use it to answer the query.",
-                                            "Summarize the answer in a comprehensive, yet succinct way.",
-                                        ],
-                                        list_item_prefix=None,
-                                    ),
-                                    Section(
-                                        name="Guidelines",
-                                        list=[
-                                            "If the answer is not found in the page content, it's insufficent, or not relevant "
-                                            "to the query at all, state it clearly.",
-                                            "Do not fabricate information. Stick to provided content.",
-                                            "Provide context for the next call (e.g., if a paragraph was cut short, include "
-                                            "relevant header information, section, etc. for continuity). Assume the content is "
-                                            "partial content from the page. Be very detailed in the context.",
-                                            "If unable to answer but found important information, include it in the context "
-                                            "for the next call.",
-                                            "Pay attention to the details of the query and make sure the answer is suitable "
-                                            "for the intent of the query.",
-                                            "A potential answer might have been provided. This means you thought you found "
-                                            "the answer in a previous partial text for the same page. You should double-check "
-                                            "that and provide an alternative revised answer if you think it's wrong, "
-                                            "or repeat it if you think it's right or cannot be validated using the current "
-                                            "text.",
-                                        ],
-                                    ),
+                                name="Guidelines",
+                                list=[
+                                    "If the answer is not found in the page content, it's insufficent, or not relevant "
+                                    "to the query at all, state it clearly.",
+                                    "Do not fabricate information. Stick to provided content.",
+                                    "Provide context for the next call (e.g., if a paragraph was cut short, include "
+                                    "relevant header information, section, etc. for continuity). Assume the content is "
+                                    "partial content from the page. Be very detailed in the context.",
+                                    "If unable to answer but found important information, include it in the context "
+                                    "for the next call.",
+                                    "Pay attention to the details of the query and make sure the answer is suitable "
+                                    "for the intent of the query.",
+                                    "A potential answer might have been provided. This means you thought you found "
+                                    "the answer in a previous partial text for the same page. You should double-check "
+                                    "that and provide an alternative revised answer if you think it's wrong, "
+                                    "or repeat it if you think it's right or cannot be validated using the current "
+                                    "text.",
                                 ],
-                            )
+                            ),
                         ],
-                    ),
+                    )
                 ],
-                max_total_messages=2,
             )
-            chat_conductor = RoundRobinChatConductor()
-            final_answer = chat_conductor.initiate_chat_with_result(
-                chat=chat,
-                initial_message=str(
+
+            final_answer, _ = get_response(
+                query=str(
                     StructuredString(
                         sections=[
                             Section(name="Query", text=query),
                             Section(name="Url", text=url),
                             Section(name="Title", text=title),
                             Section(name="Previous Answer", text=answer),
-                            Section(name="Page Content", text=text),
+                            Section(name="Page Content", text=f"```{text}```"),
                         ]
                     )
                 ),
+                answerer=query_answerer,
             )
+
             result = string_output_to_pydantic(
                 output=final_answer, chat_model=self.chat_model, output_schema=PageQueryAnalysisResult
             )

chatflock/web_research/page_retrievers/base.py

@@ -6,3 +6,6 @@
 class PageRetriever(abc.ABC):
     def retrieve_html(self, url: str, **kwargs: Any) -> str:
         raise NotImplementedError()
+
+    def close(self) -> None:
+        pass

chatflock/web_research/page_retrievers/fallback.py

@@ -5,7 +5,8 @@
 
 class RetrieverWithFallback(PageRetriever):
     def __init__(self, retrievers: Sequence[PageRetriever]):
-        assert len(retrievers) > 0, "Must provide at least one retriever."
+        if len(retrievers) == 0:
+            raise ValueError("Must provide at least one retriever.")
 
         self.retrievers = retrievers