allow user to append own source content (#15)

* add accordion item for creating custom source items

* add shadcn-card component

* wip: add form for custom user source

* cleanup custom source componentn

* add shadcn input component

* add components to get custom source by website url and copy/paste

* add logic for extracting pdf and page content from url

* wip: add component for rendering pdf content

* define id field on URLContent

* fetch custom sources from specified links

- supports .pdf and any public webpage

* retain formatting in web and pdf content; improve page content rendering

* cleanup

* wip: add logic to store and fetch custom sources

* add endpoint to fetch custom sources for a given audiocast session

* create a component for rendering audio_souces - ai-generated custom

* setup firebase and rxjs

* add a safe_to_dict method to db_manager

* use generate-custom-source endpoint when adding custom source

- this allows saving the custom source to firestore

* fetch custom sources on mount page/drawer

* read custom_sources from firebase using the js_sdk

* add firebase.json and optimize_deps

* parse firebase_config env variable in vite.config

* define .firebaserc

* update custom_source model; add toast notification

* consolidate and cleanup

* wip: create workflow to store copy/paste source

* add useful helpers to CustomSourceManager

* add endpoint and request logic to add copy/paste source

* improve the layouts of copypast_source and website_url_source forms

* refactor and cleanup

* use improved module names

* finalize endpoint to store uploaded content

* improve class and module names

* add ui logic for uploading dropped files

* upload dragged&dropped file as form_data; allow up to 10mb

* add a title field on source_content model

* ensure the user cannot create more than 5 custom sources

* allow uploading a file using the file dialog

* cleanup

* write source_content to db after it's generated

* add user provided custom sources to the tts_prompt

* specify the correct interface for _get_custom_sources

* improve the prompt to exclude extraneous information in user-provided sources

* add FIREBASE_CONFIG to env

* remove firebase_config from env_vars

.firebaserc

@@ -0,0 +1,3 @@
+{
+    "projects": {}
+}

.github/workflows/deploy_app.yml

@@ -24,6 +24,7 @@ env:
   CONFIG_FILE: "app/app.yaml"
   SERVICE: audiora-app
   API_BASE_URL: ${{ secrets.API_BASE_URL }}
+  FIREBASE_CONFIG: ${{ secrets.FIREBASE_CONFIG }}
 
 jobs:
   prepare:

api/requirements.txt

@@ -30,4 +30,8 @@ uvicorn
 uvloop
 redis[hiredis]
 
-async-web-search
+async-web-search
+
+lxml 
+beautifulsoup4 
+pypdf[crypto]

api/src/main.py

@@ -2,7 +2,7 @@
 from time import time
 from typing import Any, Callable, Generator
 
-from fastapi import BackgroundTasks, FastAPI, HTTPException, Request
+from fastapi import BackgroundTasks, FastAPI, Form, HTTPException, Request, UploadFile
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, StreamingResponse
 from fastapi_utilities import add_timer_middleware
@@ -13,6 +13,18 @@
     SessionChatItem,
     SessionChatRequest,
 )
+from src.utils.custom_sources.base_utils import SourceContent
+from src.utils.custom_sources.extract_url_content import ExtractURLContent, ExtractURLContentRequest
+from src.utils.custom_sources.generate_url_source import (
+    CustomSourceManager,
+    CustomSourceModel,
+    DeleteCustomSourcesRequest,
+    GenerateCustomSourceRequest,
+    GetCustomSourcesRequest,
+    generate_custom_source,
+)
+from src.utils.custom_sources.save_copied_source import CopiedPasteSourceRequest, save_copied_source
+from src.utils.custom_sources.save_uploaded_sources import UploadedFiles
 from src.utils.generate_audiocast import (
     GenerateAudioCastRequest,
     GenerateAudioCastResponse,
@@ -54,12 +66,16 @@ async def log_request_headers(request: Request, call_next: Callable):
 
 
 @app.get("/")
-async def root():
+def root():
     return {"message": "Hello World"}
 
 
 @app.post("/chat/{session_id}", response_model=Generator[str, Any, None])
-async def chat_endpoint(session_id: str, request: SessionChatRequest, background_tasks: BackgroundTasks):
+def chat_endpoint(
+    session_id: str,
+    request: SessionChatRequest,
+    background_tasks: BackgroundTasks,
+):
     """Chat endpoint"""
     category = request.contentCategory
     db = SessionManager(session_id, category)
@@ -85,18 +101,20 @@ async def generate_audiocast_endpoint(
     request: GenerateAudioCastRequest,
     background_tasks: BackgroundTasks,
 ):
-    result = await generate_audiocast(request, background_tasks)
-    return result
+    return await generate_audiocast(request, background_tasks)
 
 
 @app.get("/audiocast/{session_id}", response_model=GenerateAudioCastResponse)
-async def get_audiocast_endpoint(session_id: str):
+def get_audiocast_endpoint(session_id: str):
     result = get_audiocast(session_id)
     return result
 
 
 @app.post("/generate-audiocast-source", response_model=str)
-async def generate_audiocast_source_endpoint(request: GenerateAudiocastSource, background_tasks: BackgroundTasks):
+async def generate_audiocast_source_endpoint(
+    request: GenerateAudiocastSource,
+    background_tasks: BackgroundTasks,
+):
     source_content = await generate_audiocast_source(request, background_tasks)
     if not source_content:
         raise HTTPException(status_code=500, detail="Failed to generate source content")
@@ -132,6 +150,50 @@ async def get_signed_url_endpoint(blobname: str):
 
 
 @app.post("/get-session-title", response_model=str)
-async def get_session_title_endpoint(request: GetSessionTitleModel, background_tasks: BackgroundTasks):
-    source_content = await get_session_title(request, background_tasks)
-    return source_content
+async def get_session_title_endpoint(
+    request: GetSessionTitleModel,
+    background_tasks: BackgroundTasks,
+):
+    return await get_session_title(request, background_tasks)
+
+
+@app.post("/extract-url-content", response_model=SourceContent)
+def extract_url_content_endpoint(request: ExtractURLContentRequest):
+    extractor = ExtractURLContent()
+    page_content = extractor._extract(request.url)
+    return page_content.model_dump()
+
+
+@app.post("/generate-url-source", response_model=SourceContent)
+def generate_url_source_endpoint(
+    request: GenerateCustomSourceRequest,
+    background_tasks: BackgroundTasks,
+):
+    return generate_custom_source(request, background_tasks)
+
+
+@app.post("/get-custom-sources", response_model=list[CustomSourceModel])
+async def get_custom_sources_endpoint(request: GetCustomSourcesRequest):
+    return CustomSourceManager(request.sessionId)._get_custom_sources()
+
+
+@app.post("/delete-custom-source", response_model=list[CustomSourceModel])
+def delete_custom_source_endpoint(request: DeleteCustomSourcesRequest):
+    manager = CustomSourceManager(request.sessionId)
+    manager._delete_custom_source(request.sourceId)
+    return "Deleted"
+
+
+@app.post("/save-copied-source", response_model=str)
+def save_copied_source_endpoint(request: CopiedPasteSourceRequest):
+    result = save_copied_source(request)
+    return result
+
+
+@app.post("/save-uploaded-sources", response_model=str)
+async def save_uploaded_files_endpoint(files: list[UploadFile], sessionId: str = Form(...)):
+    """
+    Save sources uploaded from the frontend
+    """
+    result = await UploadedFiles(session_id=sessionId)._save_sources(files)
+    return result

api/src/services/firestore_sdk.py

@@ -1,3 +1,4 @@
+from datetime import datetime
 from typing import Dict, Literal
 
 from firebase_admin.firestore import client, firestore
@@ -48,3 +49,16 @@ def _get_document(self, collection: Collection, doc_id: str):
 
     def _get_documents(self, collection: Collection):
         return self._get_collection(collection).stream()
+
+    @classmethod
+    def _safe_to_dict(cls, data: dict):
+        """
+        safely parse firestore data by converting convert all timestamp to string
+        """
+
+        def _safe_to_str(value: dict | str | datetime):
+            if isinstance(value, datetime):
+                return value.strftime("%Y-%m-%d %H:%M:%S")
+            return value
+
+        return {k: _safe_to_str(v) for k, v in data.items()}

api/src/utils/audiocast_request.py

@@ -21,22 +21,20 @@ async def _run(self):
         Returns:
             str: The audiocast source content
         """
-        source_content = await self.__use_openai(self.category, self.preference_summary)
+        additional_ctx = await self.get_context(self.preference_summary)
+        source_content = await self.__use_openai(self.category, self.preference_summary, additional_ctx)
         if not source_content:
             raise ValueError("Failed to generate audiocast source content")
 
         return self._refine(source_content)
 
-    async def __use_openai(self, category: ContentCategory, preference_summary: str):
+    async def __use_openai(self, category: ContentCategory, preference_summary: str, additional_ctx: str):
         """
         Generate audiocast source content using OpenAI.
         """
         refined_summary = re.sub("You want", "A user who wants", preference_summary, flags=re.IGNORECASE)
         refined_summary = re.sub("You", "A user", refined_summary, flags=re.IGNORECASE)
 
-        additional_context = await self.get_context(self.preference_summary)
-        print(f">>> Additional context: {additional_context}")
-
         response = get_openai().chat.completions.create(
             model="gpt-4o",
             messages=[
@@ -45,7 +43,7 @@ async def __use_openai(self, category: ContentCategory, preference_summary: str)
                     "content": generate_source_content_prompt(
                         category,
                         refined_summary,
-                        additional_context,
+                        additional_ctx,
                     ),
                 },
                 {

api/src/utils/audiocast_script_maker.py

@@ -13,9 +13,10 @@
 class AudioScriptMaker:
     category: ContentCategory
 
-    def __init__(self, category: ContentCategory, source_content: str):
+    def __init__(self, category: ContentCategory, source_content: str, compiled_custom_sources: str | None = None):
         self.category = category
         self.source_content = source_content
+        self.compiled_custom_sources = compiled_custom_sources
 
     def create(self, provider: AudioScriptProvider = "openai"):
         """
@@ -28,9 +29,11 @@ def create(self, provider: AudioScriptProvider = "openai"):
         """
         print("Generating audio script...")
         print(f"Category: {self.category}; Source content: {self.source_content}")
+        if self.compiled_custom_sources:
+            print(f"Custom sources: {self.compiled_custom_sources}")
 
         prompt_maker = TTSPromptMaker(self.category, Metadata())
-        system_prompt = prompt_maker.get_system_prompt(self.source_content)
+        system_prompt = prompt_maker.get_system_prompt(self.source_content, self.compiled_custom_sources)
 
         if provider == "anthropic":
             audio_script = self.__use_anthropic(system_prompt)

api/src/utils/audiocast_source_refiner.py

@@ -10,7 +10,7 @@ def __init__(self, category: ContentCategory, preference_summary: str):
         self.category = category
         self.preference_summary = preference_summary
 
-    def _refine(self, content):
+    def _refine(self, content: str):
         """
         Moderate and augment the source content to ensure it aligns with the user's preferences.
         """

api/src/utils/custom_sources/base_utils.py

@@ -0,0 +1,105 @@
+from typing import Literal, Optional, TypedDict, cast
+
+from google.cloud.firestore_v1 import DocumentReference
+from pydantic import BaseModel
+
+from src.services.firestore_sdk import (
+    Collection,
+    DBManager,
+    collections,
+)
+
+
+class SourceContent(BaseModel):
+    id: str
+    content: str
+    content_type: str
+    metadata: dict = {}
+    title: Optional[str] = None
+
+    def __str__(self):
+        result = f"Content: {self.content}"
+        if self.title:
+            return f"Title: {self.title}\n{result}"
+        return result
+
+
+class CustomSourceModel(SourceContent):
+    source_type: Literal["link", "copy/paste", "file_upload"]
+    url: Optional[str] = None
+    created_at: Optional[str] = None
+    updated_at: Optional[str] = None
+
+
+class SourceContentDict(TypedDict):
+    id: str
+    content: str
+    content_type: str
+    metadata: dict
+    title: Optional[str]
+
+
+class CustomSourceModelDict(SourceContentDict):
+    source_type: Literal["link", "copy/paste", "file_upload"]
+    url: Optional[str]
+    created_at: Optional[str]
+    updated_at: Optional[str]
+
+
+class CustomSourceManager(DBManager):
+    collection: Collection = collections["audiora_sessions"]
+    sub_collection = "custom_sources"
+
+    def __init__(self, session_id: str):
+        super().__init__()
+        self.doc_id = session_id
+
+    def _check_document(self):
+        """if the collection does not exist, create it"""
+        doc = self._get_document(self.collection, self.doc_id)
+        if not doc.exists:
+            raise Exception("Session not found")
+        return doc
+
+    def _get_doc_ref(self, source_id: str) -> DocumentReference:
+        self._check_document()
+        return (
+            self._get_collection(self.collection)
+            .document(self.doc_id)
+            .collection(self.sub_collection)
+            .document(source_id)
+        )
+
+    def _set_custom_source(self, data: CustomSourceModel):
+        return self._get_doc_ref(data.id).set(
+            {
+                **(data.model_dump()),
+                "created_at": self._timestamp,
+                "updated_at": self._timestamp,
+            }
+        )
+
+    def _get_custom_source(self, source_id: str) -> CustomSourceModel | None:
+        doc = self._get_doc_ref(source_id).get()
+        data = doc.to_dict()
+        if doc.exists and data:
+            return cast(CustomSourceModel, self._safe_to_dict(data))
+
+    def _get_custom_sources(self) -> list[CustomSourceModelDict]:
+        self._check_document()
+
+        try:
+            session_ref = self._get_collection(self.collection).document(self.doc_id)
+            docs = session_ref.collection(self.sub_collection).get()
+            return [
+                cast(CustomSourceModelDict, self._safe_to_dict(doc.to_dict()))
+                for doc in docs
+                if doc.exists and doc.to_dict()
+            ]
+
+        except Exception as e:
+            print(f"Error getting custom sources for Session: {self.doc_id}", e)
+            return []
+
+    def _delete_custom_source(self, source_id: str):
+        return self._get_doc_ref(source_id).delete()