Add version for Llamaindex with FastApi and Auth0 #10

AdamWozniewski · 2025-08-20T14:15:17Z

No description provided.

priley86 · 2025-08-21T15:07:27Z

py-llamaindex/frontend/public/images/arch-bg.png

can we update this diagram to reflect Python server and the new resources here?

cc: @deepu105

I can create one for this probably next week

priley86 · 2025-08-21T15:16:17Z

py-llamaindex/README.md

+
+## About the template
+
+This template scaffolds an Auth0 + LlamaIndex.js + Next.js starter app. It mainly uses the following libraries:


priley86 · 2025-08-21T15:28:43Z

py-llamaindex/backend/app/api/routes/chat.py

+                        "access_token": auth_session.get("token_sets")[0].get(
+                            "access_token"
+                        ),
+                        "refresh_token": auth_session.get("refresh_token"),


is refresh token really needed? It seems a security hazard to me. Posted a thread about this similar issue today in JS examples.

We should probably try and standardize on this, but my preference is to only rely on exchanged / short-lived access token here.

started a thread today:
https://auth0.slack.com/archives/C089AS3M9C5/p1755783912796319

for llamaIndex this should not be needed, this was a workaround for LangGraph since the AI SDK expects this instead of LangGraph custom auth

deepu105 · 2025-08-21T15:16:41Z

py-llamaindex/backend/app/agents/tools/shop_online_li.py

+
+    user_email = user.get("email", "unknown")
+
+    index = get_index()


This is suppose to show the CIBA stuff no?

deepu105 · 2025-08-21T15:18:59Z

py-llamaindex/backend/app/api/routes/chat.py

+):
+    try:
+        query_string = str(request.url.query)
+        target_url = f"{settings.LANGGRAPH_API_URL}/{full_path}"


why is this needed for llamaindex

deepu105 · 2025-08-21T15:19:44Z

py-llamaindex/backend/app/api/routes/chat.py

+@agent_router.api_route(
+    "/{full_path:path}", methods=["GET", "POST", "DELETE", "PATCH", "PUT", "OPTIONS"]
+)
+async def api_route(


this entire implementation should not be needed for Llamaindex since the LlamaIndex agent can live in route

deepu105 · 2025-08-21T15:20:08Z

py-llamaindex/README.md

+
+## About the template
+
+This template scaffolds an Auth0 + LlamaIndex.js + Next.js starter app. It mainly uses the following libraries:


Suggested change

This template scaffolds an Auth0 + LlamaIndex.js + Next.js starter app. It mainly uses the following libraries:

This template scaffolds an Auth0 + LlamaIndex + FastAPI starter app. It mainly uses the following libraries:

deepu105 · 2025-08-21T15:22:37Z

py-llamaindex/README.md

+
+```bash
+source .venv/bin/activate
+uv run uvicorn app.main:app --reload


isn't is fastapi dev app/main.py

deepu105 · 2025-08-21T15:26:47Z

py-llamaindex/backend/app/agents/tools/google_calendar_li.py

+from googleapiclient.discovery import build
+from llama_index.core.tools import FunctionTool
+
+# UWAGA: to jest wariant dla LlamaIndex


deepu105 · 2025-08-21T15:28:34Z

py-llamaindex/backend/app/agents/tools/google_calendar_li.py

+list_upcoming_events_li = FunctionTool.from_defaults(
+    name="list_upcoming_events",
+    description="List upcoming events from the user's Google Calendar (requires a federated Google connection).",
+    fn=list_upcoming_events_li_fn,


where are we wrapping the tool with federatedConnections?

deepu105 · 2025-08-21T15:29:13Z

py-llamaindex/backend/app/api/routes/agent_li.py

+        access_token = auth_session.get("token_sets", [{}])[0].get("access_token")
+        stream = agent.stream_chat(
+            query,
+            tool_kwargs={


why are this needed?

deepu105 · 2025-08-21T15:29:33Z

py-llamaindex/backend/app/agents/tools/user_info_li.py

+from app.core.config import settings
+
+async def get_user_info_li_fn(access_token: str | None) -> str:
+    if not access_token:


why not get access token from user session

deepu105 · 2025-08-21T15:30:40Z

py-llamaindex/backend/app/core/auth0_ai.py

+    scopes=["openid", "product:buy"],
+    binding_message=lambda product,
+    quantity: f"Do you want to buy {quantity} {product}",
+    user_id=lambda *_, **__: ensure_config()


this is for langchain, for llamaindex we should be able to get this from user session

priley86 · 2025-08-21T15:42:03Z

py-llamaindex/backend/app/agents/tools/google_calendar_li.py

+
+# UWAGA: to jest wariant dla LlamaIndex
+from auth0_ai_llamaindex.federated_connections import (
+    get_access_token_for_connection,


Believe this library is currently using "refresh_token" exchange. We should enhance this library in the core to support subject_token_type urn:ietf:params:oauth:token-type:access_token / access token exchange as well:

https://github.com/auth0-lab/auth0-ai-python/blob/main/packages/auth0-ai/auth0_ai/authorizers/federated_connection_authorizer.py#L192-L197

This will likely be useful in the case that LangGraph server is hosted separately.

^^ we have a separate ticket for this effort this sprint. If we keep the langgraph agent embedded, the RT flow here now is fine. If the langgraph server / agent is separate server, my current assumption is we should use AT flow and add that support to the core.

deepu105

The PR feels incomplete, can we make it WIP and mark as ready once internally reviewed/tested

priley86 · 2025-08-21T15:47:44Z

py-llamaindex/backend/app/agents/assistant0_li.py

+agent = ReActAgentWorker.from_tools(
+    tools=tools,
+    llm=llm,
+    verbose=True,
+    system_prompt=(
+        "You are a personal assistant named Assistant0. "
+        "Use tools when helpful; prefer get_context_docs for knowledge base queries. "
+        "Render email-like bodies as markdown (no code fences)."
+    ),
+).as_agent()


how are you finding the interrupt support w/ this ReActAgent worker, and compatibility w/ langgraph's useStream hook? Would it be better to use separate langgraph-cli server w/ access token exchange flow in this case?

It seems a simpler use case may also be like this one.

fwiw, i am finding a lot of gaps w/ createReactAgent on the typescript side (and not sure how this is behaving on the python side). Ideally we can also support interrupt flows for step-up auth 🤞 . If using useStream hook from langgraph-sdk on the client, the langgraph-cli approach may be easier b/c it comes w/ langgraph protocol fully implemented. If sticking w/ this approach and embedded agent w/ ReActAgent, you may find it easier to explore a custom SSE stream (and replace useStream hook on the client w/ custom handlers), example. Leave this to you though to determine as to how to reflect.

priley86 · 2025-08-21T17:32:58Z

py-llamaindex/README.md

+
+- [LlamaIndex's Python framework](https://docs.llamaindex.ai/en/stable/#introduction)
+- The [Auth0 AI SDK](https://github.com/auth0-lab/auth0-ai-python) and [Auth0 FastAPI SDK](https://github.com/auth0/auth0-fastapi) to secure the application and call third-party APIs.
+- [Auth0 FGA](https://auth0.com/fine-grained-authorization) to define fine-grained access control policies for your tools and RAG pipelines.


should we also highlight that we are using a lot of Langgraph libraries / sdks w/ this example now? Or is that still t.b.d.?

If we go w/ Langgraph, just wanting to call it out better in README. Also agree w/ @deepu105 's feedback, that if we already have a langgraph sample, capturing some of the additional things in the existing example:
https://github.com/auth0-samples/auth0-assistant0/tree/main/py-langchain

priley86 · 2025-08-21T19:11:57Z

py-llamaindex/frontend/README.md

@@ -0,0 +1,83 @@
+# Start the frontend server
+
+Rename `.env.example` file to `.env` in the `frontend` directory.


think we are missing the .env.example here

priley86 · 2025-08-21T19:22:20Z

py-llamaindex/README.md

+To start with the basic examples, you'll just need to add your OpenAI API key and Auth0 credentials.
+
+- To start with the examples, you'll just need to add your OpenAI API key and Auth0 credentials for the Web app.
+  - You can setup a new Auth0 tenant with an Auth0 Web App and Token Vault following the Prerequisites instructions [here](https://auth0.com/ai/docs/call-others-apis-on-users-behalf).


Can we describe some place how you are configuring this app with Auth? E.g. "Create an API..", "Create a SPA application"... etc.

I'm having trouble following this at the moment, and it's quite important as far as what auth flow we want to use.

If going for an Auth0 API/resource server for the Fast API + embedded llamaindex agent, and a SPA application for the client, we'll like want to make use of the new access token / token vault flow. (more like this example w/ a SPA: auth0/docs-v2#39)

If going for a FastAPI (app) (Regular Web App for the Auth0 client) like the existing example here, and an embedded llamaindex agent, we'll likely be fine w/ the existing refresh token / token vault exchange flow.

If going for a FastAPI (app) (Regular Web App for the Auth0 client) like the existing example here, and an external llamaindex agent/external LangGraph server, we'll likely want access token / token vault exchange flow (and the AT provided to Langgraph server).

priley86 · 2025-08-21T19:35:40Z

nice work getting this started @AdamWozniewski 👍 - please let us know once you've settled on approach here and when this is rdy for testing. ty 🙇

priley86 · 2025-08-25T15:56:00Z

py-llamaindex/README.md

+  - Create a **Regular Web App** for FastAPI.
+  - (Optional) Create an API if you also expose protected endpoints to SPAs.
+- FastAPI handles cookie-based session and federated connections via Token Vault.
+- Tools do **not** receive tokens as arguments; they read them from the session or use the federated-connection wrapper.


super helpful 👍 👍

Seems this will work ok and RT -> AT / token exchange flow is desired if Regular Web App is used, and agent is embedded as we are here. I'm good w/ this if others are 👍

priley86 · 2025-08-25T16:37:49Z

...ontend/src/components/auth0-ai/FederatedConnections/FederatedConnectionsInterruptHandler.tsx

+  };
+}
+
+export function FederatedConnectionInterruptHandler({


is interrupt flow working w/ useStream? I'm not sure about useStream /langgraph integration and llamaindex ReActAgent worker. Seems there is a possibility of using the abort signals for this type of interrupt server side:
https://ts.llamaindex.ai/docs/workflows#with-workflows

is it worth investigating llamama index's client sdk for this:
https://ts.llamaindex.ai/docs/chat-ui

i see a Fast API example:
https://github.com/run-llama/chat-ui/tree/main/examples/fastapi

Maybe some React components can still be shared.

It looks like on Next.js side of things, AI SDK was used w/ llamaindex server. Another possibility to consider.

https://github.com/auth0-lab/auth0-ai-js/blob/main/examples/calling-apis/chatbot/app/(llamaindex)/components/chat.tsx#L6

cc: @jcenturion @pmalouin @deepu105 - any thoughts?

priley86 reviewed Aug 21, 2025

View reviewed changes

deepu105 requested changes Aug 21, 2025

View reviewed changes

priley86 reviewed Aug 21, 2025

View reviewed changes

deepu105 requested changes Aug 21, 2025

View reviewed changes

priley86 reviewed Aug 21, 2025

View reviewed changes

Add version for Llamaindex with FastApi and Auth0

07b5424

AdamWozniewski force-pushed the main branch from a5f3687 to 07b5424 Compare August 22, 2025 21:04

priley86 reviewed Aug 25, 2025

View reviewed changes


		## About the template

		This template scaffolds an Auth0 + LlamaIndex.js + Next.js starter app. It mainly uses the following libraries:


		user_email = user.get("email", "unknown")

		index = get_index()

		@@ -0,0 +1,83 @@
		# Start the frontend server

		Rename `.env.example` file to `.env` in the `frontend` directory.

Add version for Llamaindex with FastApi and Auth0 #10

Are you sure you want to change the base?

Add version for Llamaindex with FastApi and Auth0 #10

Uh oh!

Conversation

AdamWozniewski commented Aug 20, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

priley86 Aug 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

deepu105 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

priley86 commented Aug 21, 2025

Uh oh!

priley86 Aug 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

priley86 Aug 21, 2025 •

edited

Loading

priley86 Aug 25, 2025 •

edited

Loading