commit

app.py

@@ -19,7 +19,7 @@ questions_to_run = [
     #"4fc2f1ae-8625-45b5-ab34-ad4433bc21f8",     # OK
     #"6f37996b-2ac7-44b0-8e68-6d28256631b4",     # ??? don't understand the question
     #"9d191bce-651d-4746-be2d-7ef8ecadb9c2",     # later
-    "cabe07ed-9eca-40ea-8ead-410ef5e83f91",      # <----
+    "cabe07ed-9eca-40ea-8ead-410ef5e83f91",      # <---
     #"3cef3a44-215e-4aed-8e3b-b1e3f08063b7",
     #"99c9cc74-fdc8-46c6-8f8d-3ce2d3bfeea3",
     #"305ac316-eef6-4446-960a-92d80d542f82",

my_agent.py

@@ -12,6 +12,9 @@ from my_prompt_config import MyPromptConfig
 from my_tool_wiki_page_section import MyWikiPageSectionTool
 from my_tool_wiki_filter_tables import MyWikiTableFilterTool
 from my_tool_wiki_featured_articles import MyWikiFeaturedArticles
+from my_tools_libretexts import MyLibreTextsBookshelvesTool, MyLibreTextsBooksTool
+from my_tools_libretexts import MyLibreTextsBookSectionsTool, MyLibreTextsBookSectionParagraphsTool
+from my_tools_libretexts import MyLibreTextsParagraphContentsTool
 from dotenv import load_dotenv
 
 # https://huggingface.co/docs/transformers/model_doc/mistral?usage=Pipeline
@@ -49,6 +52,13 @@ class MyAgent:
         self.model = InferenceClientModel(model_id=self.MODEL_REASONING)
         # --- HF Inference ------------------------------------------------------------------------------
 
+        # Instruction how to specify tools
+        #   1. Implement a tool (one or more tools per file)
+        #   2. Add imports for used tools in the header for this file
+        #   3. Add file(s) with tools to the 'additional_authorized_imports' below
+        #   4. Add tool(s) to 'tools' below
+        #   5. Add tool(s) name(s) to prompt configuration in my_prompt_config.py
+
         self.reasoning_agent = CodeAgent(
             name="CourseAssistant",
             description="General AI Assistant",
@@ -62,6 +72,11 @@ class MyAgent:
                 MyWikiTableFilterTool(),
                 WebSearchTool(),
                 MyWikiFeaturedArticles(),
+                MyLibreTextsBookshelvesTool(),
+                MyLibreTextsBooksTool(),
+                MyLibreTextsBookSectionsTool(),
+                MyLibreTextsBookSectionParagraphsTool(),
+                MyLibreTextsParagraphContentsTool(),
                 FinalAnswerTool(),
             ],
             model=self.model,
@@ -87,7 +102,9 @@ class MyAgent:
                 "my_tool_reverse_string",
                 "my_tool_wiki_page_section",
                 "my_tool_wiki_table_filter",
-                "my_tool_wiki_featured_articles"
+                "my_tool_wiki_featured_articles",
+                "my_tools_libretexts",
+                "my_base_libretexts_api"
             ],
         )
         #web_search_agent = CodeAgent(

my_base_libretexts_api.py

@@ -1,7 +1,6 @@
 import requests
-import json
-from lxml import etree
 from lxml import html
+from html_to_markdown import convert_to_markdown
 
 # https://gist.github.com/scionoftech/0f35d5e231be2cf46823d774023268b6
 # https://www.mediawiki.org/wiki/API:Main_page
@@ -26,10 +25,10 @@ class MyLibreTextsAPI:
         #    print("-----------------------------------")
         #    print(f"Tag {element.tag} : title: {element.attrib['title']} href: {element.attrib['href']}")
 
-        bookshelf = [(element.attrib['title'], element.attrib['href']) for element in elements]
-        print(f"Bookshelf:\n{bookshelf}\n\n")
+        bookshelves = [{"title": element.attrib['title'], "url": element.attrib['href']} for element in elements]
+        #print(f"Bookshelf:\n{bookshelf}\n\n")
 
-        return bookshelf
+        return bookshelves
 
     def get_books(self, bookshelf_url):
         html_content = requests.get(
@@ -43,8 +42,8 @@ class MyLibreTextsAPI:
         #    print("-----------------------------------")
         #    print(f"Tag {element.tag} : title: {element.attrib['title']} href: {element.attrib['href']}")
 
-        books = [(element.attrib['title'], element.attrib['href']) for element in elements]
-        print(f"Books:\n{books}\n\n")
+        books = [{"title": element.attrib['title'], "url": element.attrib['href']} for element in elements]
+        #print(f"Books:\n{books}\n\n")
 
         return books
 
@@ -60,8 +59,8 @@ class MyLibreTextsAPI:
         #    print("-----------------------------------")
         #    print(f"Tag {element.tag} : title: {element.attrib['title']} href: {element.attrib['href']}")
 
-        sections = [(element.attrib['title'], element.attrib['href']) for element in elements]
-        print(f"Sections:\n{sections}\n\n")
+        sections = [{"title": element.attrib['title'], "url": element.attrib['href']} for element in elements]
+        #print(f"Sections:\n{sections}\n\n")
 
         return sections
 
@@ -77,21 +76,18 @@ class MyLibreTextsAPI:
         #    print("-----------------------------------")
         #    print(f"Tag {element.tag} : title: {element.attrib['title']} href: {element.attrib['href']}")
 
-        paragraphs = [(element.attrib['title'], element.attrib['href']) for element in elements]
-        print(f"Paragraphs:\n{paragraphs}\n\n")
+        paragraphs = [{"title": element.attrib['title'], "url": element.attrib['href']} for element in elements]
+        #print(f"Paragraphs:\n{paragraphs}\n\n")
 
         return paragraphs
 
-    def get_section_contents(self, paragraph_url):
+    def get_paragraph_contents(self, paragraph_url):
         html_content = requests.get(
             paragraph_url,
             headers=self.user_agent_headers,
         ).text
-        tree = html.fromstring(html_content)
-        link_class = "internal"
-        elements = tree.xpath(f"//a[@class='{link_class}']")
 
-        # TODO get contents and convert to MD format
+        markdown = convert_to_markdown(html_content)
 
-        return contents
+        return markdown
 

my_prompt_config.py

@@ -33,6 +33,11 @@ class MyPromptConfig:
             * _my_tool_wiki_table_filter : filter the subsection and tables on the Wikipedia page for provided years
             * _my_tool_wiki_featured_articles : get nominator and summary information for Wikipedia featured articles 
             for given month and year
+            * _my_tool_libretexts_bookshelves : Get a list of bookshelves in LibreTexts library of chemistry materials
+            * _my_tool_libretexts_books : Get a list of books in LibreTexts bookshelf
+            * _my_tool_libretexts_book_sections : Get a list of sections in a book in LibreTexts materials
+            * _my_tool_libretexts_book_section_paragraphs : Get a list of section paragraphs in a book in LibreTexts materials
+            * _my_tool_libretexts_paragraph_contents : Get contents of a paragraph in a book in LibreTexts materials
             
             If the question mentions image or other file, use one of provided tools to load it
             using task_id associated with the question.

my_tool_libretext_book_shelves.py

@@ -1,49 +0,0 @@
-from smolagents import Tool
-from my_base_wiki_api import MyWikiAPI
-
-
-#AUTHORIZED_TYPES = [
-#    "string",
-#    "boolean",
-#    "integer",
-#    "number",
-#    "image",
-#    "audio",
-#    "array",
-#    "object",
-#    "any",
-#    "null",
-#]
-
-class MyWikiFeaturedArticles(Tool):
-    name = "_my_tool_wiki_featured_articles"
-    description = """
-        Extracts a list of nominators and summaries of featured Wikipedia articles for the provided month and year
-        To invoke the tool use code as below
-        <code>
-        featured = _my_tool_wiki_featured_articles(month='January', year='2005')
-        </code>
-    """
-
-    inputs = {
-        "month": {
-            "type": "string",
-            "description": "name of the month for example January",
-        },
-        "year": {
-            "type": "integer",
-            "description": "Year expressed in 4 digit notation, for example 2009",
-        },
-    }
-
-    output_type = "string"
-
-    is_initialized = True
-
-    def __init__(self):
-        print(f"***KS*** Wiki featured articles tool initializing ...")
-        self.wiki = MyWikiAPI()
-
-    def forward(self, month, year):
-        result = self.wiki.get_featured_articles(month, year)
-        return result

my_tools_libretexts.py

@@ -0,0 +1,165 @@
+from smolagents import Tool
+from my_base_libretexts_api import MyLibreTextsAPI
+
+
+#AUTHORIZED_TYPES = [
+#    "string",
+#    "boolean",
+#    "integer",
+#    "number",
+#    "image",
+#    "audio",
+#    "array",
+#    "object",
+#    "any",
+#    "null",
+#]
+
+
+class MyLibreTextsBase(Tool):
+    is_initialized = True
+
+    def __init__(self):
+        print(f"***KS*** LibreTexts Base Tool initializing ...")
+        self.api = MyLibreTextsAPI()
+
+
+class MyLibreTextsBookshelvesTool(MyLibreTextsBase):
+    name = "_my_tool_libretexts_bookshelves"
+    description = """
+        Get a list of bookshelves in LibreTexts library of chemistry materials
+        Returns an array with dictionary items containing 'title' and 'url' of a bookshelf
+        <code>
+        bookshelves = _my_tool_libretexts_bookshelves()
+        </code>
+    """
+
+    inputs = {
+    }
+
+    output_type = "array"
+
+    is_initialized = True
+
+    def __init__(self):
+        MyLibreTextsBase.__init__(self)
+        print(f"***KS*** LibreTexts Bookshelves Tool initializing ...")
+
+    def forward(self):
+        return self.api.get_bookshelves()
+
+
+class MyLibreTextsBooksTool(MyLibreTextsBase):
+    name = "_my_tool_libretexts_books"
+    description = """
+        Get a list of books in LibreTexts bookshelf
+        Returns an array with dictionary items containing 'title' and 'url' of a book
+        <code>
+        books = _my_tool_libretexts_books(bookshelf_url='https://chem.libretexts.org/Bookshelves/Inorganic_Chemistry')
+        </code>
+    """
+
+    inputs = {
+        "bookshelf_url": {
+            "type": "string",
+            "description": "Bookshelf URL",
+        },
+    }
+
+    output_type = "array"
+
+    is_initialized = True
+
+    def __init__(self):
+        MyLibreTextsBase.__init__(self)
+        print(f"***KS*** LibreTexts Books Tool initializing ...")
+
+    def forward(self, bookshelf_url):
+        return self.api.get_books(bookshelf_url)
+
+
+class MyLibreTextsBookSectionsTool(MyLibreTextsBase):
+    name = "_my_tool_libretexts_book_sections"
+    description = """
+        Get a list of sections in a book in LibreTexts materials
+        Returns an array with dictionary items containing 'title' and 'url' of a section
+        <code>
+        sections = _my_tool_libretexts_book_sections(book_url='https://chem.libretexts.org/Bookshelves/Inorganic_Chemistry/Inorganic_Chemistry_(Saito)')
+        </code>
+    """
+
+    inputs = {
+        "book_url": {
+            "type": "string",
+            "description": "Book URL",
+        },
+    }
+
+    output_type = "array"
+
+    is_initialized = True
+
+    def __init__(self):
+        MyLibreTextsBase.__init__(self)
+        print(f"***KS*** LibreTexts Book Sections Tool initializing ...")
+
+    def forward(self, book_url):
+        return self.api.get_book_sections(book_url)
+
+
+class MyLibreTextsBookSectionParagraphsTool(MyLibreTextsBase):
+    name = "_my_tool_libretexts_book_section_paragraphs"
+    description = """
+        Get a list of section paragraphs in a book in LibreTexts materials
+        Returns an array with dictionary items containing 'title' and 'url' of a paragraph
+        <code>
+        paragraphs = _my_tool_libretexts_book_section_paragraphs(section_url='https://chem.libretexts.org/Bookshelves/Inorganic_Chemistry/Inorganic_Chemistry_(Saito)/03%3A_Reactions')
+        </code>
+    """
+
+    inputs = {
+        "section_url": {
+            "type": "string",
+            "description": "Book section URL",
+        },
+    }
+
+    output_type = "array"
+
+    is_initialized = True
+
+    def __init__(self):
+        MyLibreTextsBase.__init__(self)
+        print(f"***KS*** LibreTexts Book Section Paragraphs Tool initializing ...")
+
+    def forward(self, section_url):
+        return self.api.get_book_section_paragraphs(section_url)
+
+
+class MyLibreTextsParagraphContentsTool(MyLibreTextsBase):
+    name = "_my_tool_libretexts_paragraph_contents"
+    description = """
+        Get contents of a paragraph in a book in LibreTexts materials
+        Returns contents in Markdown format
+        <code>
+        markdown = _my_tool_libretexts_paragraph_contents(paragraph_url='https://chem.libretexts.org/Bookshelves/Inorganic_Chemistry/Inorganic_Chemistry_(Saito)/03%3A_Reactions/3.01%3A_Thermodynamics')
+        </code>
+    """
+
+    inputs = {
+        "paragraph_url": {
+            "type": "string",
+            "description": "Paragraph URL",
+        },
+    }
+
+    output_type = "string"
+
+    is_initialized = True
+
+    def __init__(self):
+        MyLibreTextsBase.__init__(self)
+        print(f"***KS*** LibreTexts Paragraph Contents Tool initializing ...")
+
+    def forward(self, paragraph_url):
+        return self.api.get_paragraph_contents(paragraph_url)

requirements.txt

@@ -12,4 +12,5 @@ bs4
 wikitextparser
 mwparserfromhell
 lxml
-cssselect
+cssselect
+html-to-markdown

test_tools.py

@@ -19,30 +19,28 @@ def test_tool_libretextx_bookshelves(_exp):
     bookshelves = api.get_bookshelves()
 
     shelf_1 = bookshelves[0]
-    shelf_1_url = shelf_1[1]
+    shelf_1_url = shelf_1["url"]
 
     books = api.get_books(shelf_1_url)
 
     book_2 = books[1]
-    book_2_url = book_2[1]
+    book_2_url = book_2["url"]
 
     sections = api.get_book_sections(book_2_url)
 
     section_2 = sections[1]
-    section_2_url = section_2[1]
+    section_2_url = section_2["url"]
 
     paragraphs = api.get_book_section_paragraphs(section_2_url)
 
     paragraph_8 = paragraphs[7]
-    paragraph_8_url = paragraph_8[1]
+    paragraph_8_url = paragraph_8["url"]
 
-    #content = api.get_featured_articles(_month,_year)
-    #tool = MyWikiFeaturedArticles()
-    #results = tool(month=_month, year=_year)
-    #results_str = "\n".join(results)
-    #print(f"Results: {len(results)}\n{results_str}")
+    contents = api.get_paragraph_contents(paragraph_8_url)
 
-    assert 0 == 0
+    #print(f"Contents:\n{contents}")
+
+    assert "Louvrier" in contents
 
 
 @pytest.mark.skip(reason="disabled")