feat(tutor): simplify documents dict to be added to llm

k8s/welearn-api/values.dev.yaml

@@ -4,7 +4,7 @@ ingress:
 config:
   nonSensitive:
     AZURE_API_BASE: "https://apim-welearn-openai.azure-api.net"
-    AZURE_GPT_40_API_BASE: "https://welearn-openai-sweden.openai.azure.com"
+    AZURE_GPT_4O_API_BASE: "https://welearn-openai-sweden.openai.azure.com"
     PG_HOST: dev-lab-projects-backend.postgres.database.azure.com
     AZURE_LLAMA_31_8B_API_BASE: https://meta-llama-3-1-8b-instruct-wl.swedencentral.models.ai.azure.com/
     AZURE_LLAMA_31_70B_API_BASE: https://meta-llama-3-1-70b-instruct-wl.swedencentral.models.ai.azure.com/

k8s/welearn-api/values.prod.yaml

@@ -4,7 +4,7 @@ ingress:
 config:
   nonSensitive:
     AZURE_API_BASE: "https://apim-welearn-openai.azure-api.net"
-    AZURE_GPT_40_API_BASE: "https://welearn-openai-sweden.openai.azure.com"
+    AZURE_GPT_4O_API_BASE: "https://welearn-openai-sweden.openai.azure.com"
     PG_HOST: prod-prod-projects-backend.postgres.database.azure.com
     AZURE_LLAMA_31_8B_API_BASE: https://meta-llama-3-1-8b-instruct-wl.swedencentral.models.ai.azure.com/
     AZURE_LLAMA_31_70B_API_BASE: https://meta-llama-3-1-70b-instruct-wl.swedencentral.models.ai.azure.com/

k8s/welearn-api/values.yaml

@@ -39,7 +39,7 @@ config:
     CORPUS: conversation
     MODELS_FOLDER_ROOTS: "/models"
     AZURE_API_BASE: "https://welearn-openai.openai.azure.com/openai/"
-    AZURE_GPT_40_API_BASE: "https://welearn-openai-sweden.openai.azure.com/"
+    AZURE_GPT_4O_API_BASE: "https://welearn-openai-sweden.openai.azure.com/"
     CLIENT_ORIGINS_REGEX: '^{{ join "|" (values .Values.allowedHostsRegexes | sortAlpha ) }}$'
     PG_USER: welearn_datastack
     PG_DATABASE: welearn_datastack

src/app/services/tutor/agents.py

@@ -70,8 +70,6 @@ def __init__(self, model_client: ChatCompletionClient) -> None:
                         "Assessment Methods: Propose methods to evaluate student progress."
                         "Course Schedule: Create a week-by-week breakdown with key topics and associated learning outcomes."
                         "References: Include all sources that you use to construct the syllabus."
-                        "3. Save a file to the current directory:"
-                        f"Syllabus File: Contains the drafted syllabus based on the provided template. You must follow this template :\n {TEMPLATES['template0']}. Name the file: uni-prof_syllabus.md"
                     ),
                     expected_output=f"You must follow this template :\n {TEMPLATES['template0']}",
                 )
@@ -81,9 +79,19 @@ def __init__(self, model_client: ChatCompletionClient) -> None:
 
     @message_handler
     async def handle_documents_and_themes(
-        self, message: TutorSearchResponse, ctx: MessageContext
+        self, message: MessageWithResources, ctx: MessageContext
     ) -> None:
-        prompt = f"Using the content in TEXT CONTENTS, you generate a syllabus that is engaging and coherent in relation to the THEMES extracted from these contents. \n\nTEXT CONTENTS:\n{message.extracts[0].original_document}\n\nTHEMES:\n{message.extracts[0].themes}"
+        contents = []
+        for curr_content in message.content:
+            if isinstance(curr_content, TutorSearchResponse):
+                contents.append(curr_content.extracts)
+
+        themes = []
+        for curr_content in contents:
+            if isinstance(curr_content, TutorSearchResponse):
+                themes.extend(curr_content.extracts)
+
+        prompt = f"Using the content in TEXT CONTENTS, you generate a syllabus that is engaging and coherent in relation to the THEMES extracted from these contents. \n\nTEXT CONTENTS:\n{contents}\n\nTHEMES:\n{themes}"
 
         start_time = time.time()
         llm_result = await self._model_client.create(
@@ -102,7 +110,7 @@ async def handle_documents_and_themes(
 
         await self.publish_message(
             MessageWithResources(
-                content=response, resources=message.documents, source=self.id.type
+                content=response, resources=message.resources, source=self.id.type
             ),
             topic_id=TopicId(sdg_expert_topic_type, source=self.id.key),
         )
@@ -139,8 +147,7 @@ def __init__(self, model_client: ChatCompletionClient, memory: ListMemory) -> No
                     "References: NEVER delete, summarize, or modify any references already present in the syllabus. Only append this section with any additional sources that you use to construct the syllabus. "
                 ),
                 expected_output=(
-                    f"1. Final Syllabus File: The revised syllabus, ready for the pedagogical engineer's review. You must follow this template :\n {TEMPLATES['template0']}. Name the file: SDG-exp_syllabus.md"
-                    "When you are done with generating the final output, reply with SDG EXPERT DONE."
+                    f"1. The revised syllabus, ready for the pedagogical engineer's review. You must follow this template :\n {TEMPLATES['template0']}."
                 ),
             ),
             memory=[memory],
@@ -217,9 +224,7 @@ def __init__(self, model_client: ChatCompletionClient, memory: ListMemory) -> No
                     "3. Check for consistency, clarity, and overall syllabus coherence to ensure usability for instructors."
                 ),
                 expected_output=(
-                    "Save a to the current directory:"
-                    f"1. Final Syllabus File: The polished syllabus, ready for user review. You must follow this template :\n {TEMPLATES['template0']}. Name the file: peda-eng_syllabus.md"
-                    "When you are done with generating the final output, reply with TERMINATE."
+                    f"1. Final Syllabus: The polished syllabus, ready for user review. You must follow this template :\n {TEMPLATES['template0']}."
                 ),
             ),
             memory=[memory],

src/app/services/tutor/models.py

@@ -38,8 +38,8 @@ class MessageWithAnalysis(BaseModel):
 
 
 class MessageWithResources(BaseModel):
-    content: Dict | str
-    resources: List[Document]
+    content: list[ExtractorOutput] | str
+    resources: List[Dict]
     source: str = "default"
 
 

src/app/services/tutor/tutor.py

@@ -22,7 +22,8 @@
     sdg_expert_topic_type,
     university_teacher_topic_type,
 )
-from src.app.services.tutor.models import Message, TaskResponse, TutorSearchResponse
+from src.app.services.tutor.models import Message, TaskResponse, TutorSearchResponse, MessageWithResources
+from src.app.services.tutor.utils import extract_doc_info
 
 settings = get_settings()
 
@@ -39,6 +40,8 @@
 async def tutor_manager(content: TutorSearchResponse) -> Message:
     queue = asyncio.Queue[TaskResponse]()
 
+    formatted_content = MessageWithResources(content=content.extracts, resources=extract_doc_info(content.documents))
+
     async def collect_result(
         _agent: ClosureContext, message: TaskResponse, ctx: MessageContext
     ) -> None:
@@ -102,7 +105,7 @@ async def collect_result(
     )
 
     await runtime.send_message(
-        content,
+        formatted_content,
         recipient=AgentId(university_teacher_topic_type, "default"),
     )
 

src/app/services/tutor/utils.py

@@ -1,3 +1,6 @@
+from src.app.models.documents import Document
+
+
 def build_system_message(
     role: str,
     backstory: str,
@@ -13,3 +16,21 @@ def build_system_message(
     if expected_output:
         message += f"\nThis is the expected criteria for your final answer: {expected_output}\nYou MUST return the actual complete content as the final answer, not a summary."
     return message
+
+
+def extract_doc_info(documents: list[Document]) -> list[dict]:
+    """
+    Extracts the document information from a list of documents.
+    Args:
+        documents (list[Document]): List of Document objects.
+    Returns:
+        list[dict]: List of dictionaries containing document information.
+    """
+    return [
+        {
+            "title": doc.payload.document_title,
+            "url": doc.payload.document_url,
+            "content": doc.payload.slice_content,
+        }
+        for doc in documents
+    ]