Start addressing reveiw comments

Author: GovernmentPlates

ai_summary/indico_ai_summary/blueprint.py

@@ -7,11 +7,11 @@
 
 from indico.core.plugins import IndicoPluginBlueprint
 
-from indico_ai_summary.controllers import RHManageCategoryPrompts, SummarizeEvent
+from indico_ai_summary.controllers import RHManageCategoryPrompts, RHSummarizeEvent
 
 
 blueprint = IndicoPluginBlueprint('ai_summary', __name__, url_prefix='/plugin/ai-summary')
 
-blueprint.add_url_rule('/manage-category-prompts/<int:category_id>', 'manage_category_prompts',
+blueprint.add_url_rule('!/category/<int:category_id>/manage/prompts', 'manage_category_prompts',
                        RHManageCategoryPrompts, methods=('GET', 'POST'))
-blueprint.add_url_rule('/summarize-event/<int:event_id>', 'summarize_event', SummarizeEvent, methods=('POST',))
+blueprint.add_url_rule('/summarize-event/<int:event_id>', 'summarize_event', RHSummarizeEvent, methods=('POST',))

ai_summary/indico_ai_summary/client/components/ind_summarize_button.jsx

@@ -24,7 +24,7 @@ import {PromptControls, PromptEditor} from './PromptSelector';
 import SummaryPreview from './SummaryPreview';
 import './ind_summarize_button.module.scss';
 
-function SummarizeButton({categoryId, eventId, storedPrompts, streamResponse, llmInfo}) {
+function SummarizeButton({eventId, storedPrompts, streamResponse, llmInfo}) {
   const [selectedPromptIndex, setSelectedPromptIndex] = useState(0);
   const [prompts, setPrompts] = useState(storedPrompts);
   const selectedPrompt = prompts[selectedPromptIndex];
@@ -51,7 +51,7 @@ function SummarizeButton({categoryId, eventId, storedPrompts, streamResponse, ll
 
     if (streamResponse) {
       setStreamStopped(false);
-      // Streaming via indicoAxios
+      // Streaming via fetch
       const ctl = streamSummary(eventId, selectedPrompt.text, {
         onChunk: html => {
           // Replace each time with server snapshot

ai_summary/indico_ai_summary/client/services/summarize.js

@@ -130,6 +130,7 @@ export function streamSummary(eventId, prompt, {onChunk, onDone, onError} = {})
         method: 'POST',
         headers: {
           'Content-Type': 'application/json',
+          'X-CSRF-Token': document.getElementById('csrf-token').getAttribute('content'),
         },
         body: JSON.stringify({prompt: prompt || ''}),
         signal: abortController.signal,

ai_summary/indico_ai_summary/controllers.py

@@ -15,19 +15,19 @@
 from indico.modules.categories.controllers.base import RHManageCategoryBase
 from indico.modules.events.management.controllers.base import RHManageEventBase
 from indico.modules.events.notes.util import get_scheduled_notes
+from indico.util.string import sanitize_html
 
 from indico_ai_summary.llm_interface import LLMInterface
 from indico_ai_summary.models.prompt import Prompt
 from indico_ai_summary.schemas import PromptSchema
-from indico_ai_summary.utils import MarkupMode, chunk_text, convert_markup, generate_chunk_stream
+from indico_ai_summary.utils import chunk_text, markdown_to_html, html_to_markdown, generate_chunk_stream
 from indico_ai_summary.views import WPCategoryManagePrompts
 
 
 CATEGORY_SIDEMENU_ITEM = 'plugin_ai_summary_prompts'
 
 
 class RHManageCategoryPrompts(RHManageCategoryBase):
-
     def _process_args(self):
         RHManageCategoryBase._process_args(self)
 
@@ -47,25 +47,18 @@ def _process_POST(self, prompts):
         return '', 204
 
 
-class SummarizeEvent(RHManageEventBase):
-
-    CSRF_ENABLED = False
-
-    def _process_args(self):
-        RHManageEventBase._process_args(self)
-
-    @use_kwargs({'prompt': fields.Str(load_default='')}, location='json')
+class RHSummarizeEvent(RHManageEventBase):
+    @use_kwargs({'prompt': fields.Str(required=True)})
     def _process(self, prompt):
-        event = self.event
-        prompt_template = prompt.strip()
-        notes = get_scheduled_notes(event)
+        prompt_template = prompt
+        notes = get_scheduled_notes(self.event)
         meeting_notes = '\n\n'.join(note.html for note in notes if hasattr(note, 'html'))
 
         if not meeting_notes.strip():
             current_plugin.logger.error("No meeting notes found from this event's contributions.")
             return jsonify({'error': "No meeting notes found from this event's contributions."}), 400
 
-        cleaned_text = convert_markup(meeting_notes, MarkupMode.HTML_TO_MARKDOWN)
+        cleaned_text = html_to_markdown(meeting_notes)
         chunks = chunk_text(cleaned_text)
         summaries = []
 
@@ -74,14 +67,14 @@ def _process(self, prompt):
             return jsonify({'error': 'LLM Auth Token is not set in plugin settings.'}), 400
 
         llm_model = LLMInterface(
-                model_name=current_plugin.settings.get('llm_model_name'),
-                host=current_plugin.settings.get('llm_host_name'),
-                url=current_plugin.settings.get('llm_provider_url'),
-                auth_token=current_plugin.settings.get('llm_auth_token'),
-                max_tokens=current_plugin.settings.get('llm_max_tokens'),
-                temperature=current_plugin.settings.get('llm_temperature'),
-                system_prompt=current_plugin.settings.get('llm_system_prompt')
-            )
+            model_name=current_plugin.settings.get('llm_model_name'),
+            host=current_plugin.settings.get('llm_host_header'),
+            url=current_plugin.settings.get('llm_provider_url'),
+            auth_token=current_plugin.settings.get('llm_auth_token'),
+            max_tokens=current_plugin.settings.get('llm_max_tokens'),
+            temperature=current_plugin.settings.get('llm_temperature'),
+            system_prompt=current_plugin.settings.get('llm_system_prompt'),
+        )
 
         if current_plugin.settings.get('llm_stream_response'):
             return Response(
@@ -103,8 +96,8 @@ def _process(self, prompt):
             summaries.append(response)
 
         combined_summary = '\n'.join(summaries)
-        html_output = convert_markup(combined_summary, MarkupMode.MARKDOWN_TO_HTML)
+        html_output = markdown_to_html(combined_summary)
 
         return {
-            'summary_html': html_output
+            'summary_html': sanitize_html(html_output)
         }

ai_summary/indico_ai_summary/llm_interface.py

@@ -31,7 +31,7 @@ class LLMInterface(llm.Model):
 
     can_stream = False
 
-    def __init__(self, model_name: str, host: str, url: str, auth_token: str, max_tokens: int = 1024,
+    def __init__(self, *, model_name: str, host: str, url: str, auth_token: str, max_tokens: int = 1024,
                  temperature: float = 0.5, system_prompt: str = '') -> None:
         super().__init__()
         self.model_name = model_name
@@ -86,7 +86,7 @@ def _stream_deltas(self, headers, payload) -> Generator[str, None, None]:
                     line = raw_line.strip()
                     # OpenAI-compatible providers prefix with 'data: '
                     if line.startswith('data:'):
-                        line = line[len('data:'):].strip()
+                        line = line.removeprefix('data:').strip()
                     if line == '[DONE]':
                         break
                     # Some providers may send keep-alives like ': ping' -> ignore non-JSON
@@ -110,12 +110,12 @@ def _extract_content_and_done(obj) -> tuple[str | None, bool]:
         """Extract content delta and completion status from a provider chunk."""
         choices = obj.get('choices') or []
         if choices:
-            ch0 = choices[0]
-            delta = ch0.get('delta') or {}
+            choice = choices[0]
+            delta = choice.get('delta') or {}
             content = delta.get('content')
             if content:
                 return content, False
-            if ch0.get('finish_reason'):
+            if choice.get('finish_reason'):
                 return None, True
         # Fallback: some providers may send raw 'content'
         content = obj.get('content')

ai_summary/indico_ai_summary/plugin.py

@@ -23,9 +23,9 @@
 
 from indico_ai_summary.blueprint import blueprint
 from indico_ai_summary.controllers import CATEGORY_SIDEMENU_ITEM
-from indico_ai_summary.models.prompt import Prompt
 from indico_ai_summary.schemas import PromptSchema
 from indico_ai_summary.views import WPCategoryManagePrompts
+from indico_ai_summary.utils import get_all_prompts
 
 
 class PromptManagerField(JSONField):
@@ -47,18 +47,16 @@ class PluginSettingsForm(IndicoForm):
                                          [DataRequired()],
                                          toggle=True,
                                          description=_('The authentication token for accessing the LLM provider.'))
-    llm_host_name = StringField(_('Host Name'),
+    llm_host_header = StringField(_('Host Name'),
                                  description=_('An optional host header to be added to the request (if '
                                                'required by your provider).'))
     llm_max_tokens = IntegerField(_('Max Tokens'),
                                   [NumberRange(min=1)],
-                                  default=1024,
                                   description=_('The maximum number of tokens to generate in the response. Defaults '
                                                 'to a maximum of 1024 tokens.'))
     llm_temperature = FloatField(_('Temperature'),
                                    [NumberRange(min=0, max=1)],
                                    widget=NumberInput(step='0.1'),
-                                   default=0.5,
                                    description=_('The sampling temperature to use, between 0 and 1. Higher values '
                                                  'like 0.8 will make the response more random, while lower values like '
                                                  '0.2 will make it more focused and deterministic.'))
@@ -76,7 +74,8 @@ class PluginSettingsForm(IndicoForm):
 
 class IndicoAISummaryPlugin(IndicoPlugin):
     """AI-assisted minutes summarization tool
-    Configure your LLM provider as well as predefined prompts to assist in summarizing event minutes.
+
+    Configure your LLM provider as well as predefined prompts to assist in summarizing meeting minutes.
     """
 
     configurable = True
@@ -86,7 +85,7 @@ class IndicoAISummaryPlugin(IndicoPlugin):
         'llm_provider_name': None,
         'llm_provider_url': None,
         'llm_auth_token': None,
-        'llm_host_name': None,
+        'llm_host_header': None,
         'llm_max_tokens': 1024,
         'llm_temperature': 0.5,
         'llm_system_prompt': '',
@@ -111,11 +110,12 @@ def get_blueprints(self):
 
     def _render_summarize_button(self, event):
         global_prompts = self.settings.get('prompts')
-        category_prompts = Prompt.query.with_parent(event.category).all()
+        category_prompts = get_all_prompts(event.category)
         all_prompts = global_prompts + PromptSchema(many=True).dump(category_prompts)
         stream_response = self.settings.get('llm_stream_response')
         show_info = self.settings.get('display_info')
 
+        llm_info = None
         if show_info:
             llm_info = {
                 'provider_name': self.settings.get('llm_provider_name'),
@@ -127,7 +127,7 @@ def _render_summarize_button(self, event):
                                       event=event,
                                       stored_prompts=all_prompts,
                                       stream_response=stream_response,
-                                      llm_info=(llm_info if show_info else None))
+                                      llm_info=llm_info)
 
     def _extend_category_menu(self, sender, category, **kwargs):
         return SideMenuItem(CATEGORY_SIDEMENU_ITEM, _('Prompts'),

ai_summary/indico_ai_summary/schemas.py

@@ -11,5 +11,5 @@
 
 
 class PromptSchema(mm.Schema):
-    name = fields.Str(required=True)
-    text = fields.Str(required=True)
+    name = fields.String(required=True)
+    text = fields.String(required=True)

ai_summary/indico_ai_summary/templates/summarize_button.html

@@ -2,7 +2,7 @@
     category-id="{{ category.id }}"
     event-id="{{ event.id }}"
     stored-prompts="{{ stored_prompts | tojson | forceescape }}"
-    stream-response="{{ stream_response | tojson }}"
+    stream-response="{{ stream_response | tojson | forceescape }}"
     llm-info="{{ llm_info | tojson | forceescape }}"
 >
 </ind-summarize-button>

ai_summary/indico_ai_summary/utils.py

@@ -8,44 +8,35 @@
 import itertools
 import json
 from collections.abc import Generator
-from enum import Enum
 
-import html2text
-import markdown
+from html2text import HTML2Text
+from indico.util.string import render_markdown, sanitize_html
+from indico_ai_summary.models.prompt import Prompt
+from indico.modules.categories.models.categories import Category
 
 from indico_ai_summary.llm_interface import LLMInterface
 
 
-class MarkupMode(Enum):
-    """Supported markup conversion modes.
+def html_to_markdown(html_string: str) -> str:
+    """Convert a HTML string to Markdown.
 
-    Members
-    - ``HTML_TO_MARKDOWN``: convert HTML to Markdown
-    - ``MARKDOWN_TO_HTML``: convert Markdown to HTML
+    :param html_string: The input HTML string.
+    :return: The converted Markdown string.
     """
+    h = HTML2Text()
+    h.ignore_links = False
+    h.ignore_images = True
+    h.body_width = 0
+    return h.handle(html_string)
 
-    HTML_TO_MARKDOWN = 1
-    MARKDOWN_TO_HTML = 2
 
+def markdown_to_html(markdown_string: str) -> str:
+    """Convert a Markdown string to HTML using :func:`render_markdown` from the core.
 
-def convert_markup(markup: str, mode: MarkupMode) -> str:
-    """Convert markup between HTML and Markdown formats.
-
-    :param markup: The input markup string to be converted.
-    :param mode: The conversion mode. Use members of :class:`MarkupMode`,
-        for example ``MarkupMode.HTML_TO_MARKDOWN`` or
-        ``MarkupMode.MARKDOWN_TO_HTML``.
-    :return: The converted markup string.
+    :param markdown_string: The input Markdown string.
+    :return: The converted HTML string.
     """
-    if mode == MarkupMode.HTML_TO_MARKDOWN:
-        h = html2text.HTML2Text()
-        h.ignore_links = False
-        h.ignore_images = True
-        h.body_width = 0
-        return h.handle(markup)
-    if mode == MarkupMode.MARKDOWN_TO_HTML:
-        return markdown.markdown(markup)
-    raise ValueError(f'Unknown markup mode: {mode}')
+    return render_markdown(markdown_string)
 
 
 def chunk_text(text: str, max_tokens: int = 1500) -> list[str]:
@@ -71,9 +62,15 @@ def generate_chunk_stream(chunks: list[str], prompt: str, llm_model: LLMInterfac
         try:
             for delta in response_stream:
                 md_total += delta
-                html_snapshot = convert_markup(md_total, MarkupMode.MARKDOWN_TO_HTML)
+                html_snapshot = sanitize_html(markdown_to_html(md_total))
                 yield f"data: {json.dumps({'summary_html': html_snapshot})}\n\n"
         except Exception as e:
             yield f"data: {json.dumps({'error': f'Streaming error: {e}'})}\n\n"
             return
     yield 'data: [DONE]\n\n'
+
+
+def get_all_prompts(category: Category) -> set[Prompt]:
+    """Get all prompts defined for the given event/category."""
+    current_category = category or Category.get_root()
+    return set(Prompt.query.filter(Prompt.category_id.in_(categ['id'] for categ in current_category.chain)).all())