Deploy an AIP Chatbot in a Custom React App: Palantir Foundry OSDK

In this video, "Deploy an AIP Chatbot in a Custom React App: Palantir Foundry OSDK" by Ontologize, former Palantir engineers demonstrate how to decouple an AIP Chatbot Service (formerly AIP Agent Studio) from standard Foundry interfaces (like Workshop) and expose it directly as a stateless RESTful conversational endpoint consumed inside a custom, branded React application using the Palantir Ontology SDK (OSDK) and Platform SDK [00:00].

---

Architectural Process Map: Decoupled AIP Chatbot & React OSDK Flow

An AIP Chatbot in Palantir Foundry acts as a versioned, callable backend service that orchestrates ontology retrieval (RAG), memory, and actions [00:05]. The React frontend authenticates via OAuth 2.0 and communicates via the Platform SDK's AIP Chatbots API [06:31].

                DECOUPLED AIP CHATBOT & REACT OSDK ARCHITECTURE
┌─────────────────────────────────────────────────────────────────────────────┐
│ [ 1. AIP Chatbot Studio (Backend Conversational Service) ]                 │
│   ├── System Prompt & Instructions                                           │
│   ├── Vector Retrieval Context (Semantic Search on Chunked FAA Media Set)   │
│   └── Agent Service Resource Identifier (RID: `ri.aip-agents...`)           │
└──────────────────────────────────────┬──────────────────────────────────────┘
                                       │
                                       ▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ [ 2. Developer Console Configuration (App Gateway & Permissions) ]         │
│   ├── Client-Facing Application Setup + OAuth PKCE Redirect URI             │
│   ├── Platform SDK Scope: Enable `AIP Chatbots API`                          │
│   ├── Project Access: Grant scope to `AIP Agents` Project                    │
│   └── Ontology SDK Scope: Add Context Object Types (e.g., `Gina FAA Chunk`) │
└──────────────────────────────────────┬──────────────────────────────────────┘
                                       │
                                       ▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ [ 3. Custom React App & OSDK Execution Environment ]                        │
│   ├── Authentication: OAuth PKCE User Grant (Popup authorization)           │
│   ├── Platform SDK Invocation: Passes Prompt + Agent RID to Chatbots Endpoint│
│   └── Custom UI Rendering: Custom CSS/Theme, Custom Citation Click Handlers │
└─────────────────────────────────────────────────────────────────────────────┘

---

Step-by-Step Guidance: Surfacing an AIP Chatbot in React via OSDK

1. Configure and Publish AIP Chatbot: Configure semantic search and system prompt over chunked media.

2. Press Ctrl + J and search for AIP Chatbot Studio (formerly AIP Agent Studio) [01:31].

3. Click Create Chatbot and name it (e.g., FAA Training Assistant) [01:51].

4. Save the chatbot in a dedicated project directory (e.g., AIP Agents Project) [02:16].

5. Define System Prompt: Set instructions (e.g., "You are an expert FAA training assistant...") [03:00].

6. Add Retrieval Context: Click Add Retrieval Context $\rightarrow$ Ontology Context $\rightarrow$ Select your chunked document object type (Gina FAA Chunk) [03:28]. Set retrieval mode to Relevant Objects and select the vector embedding property [03:48].

7. Click View Details in the top-right corner and copy the Chatbot RID (ri.aip-agents...) [09:16].

8. Register Client App in Developer Console: Register Client Application, OAuth PKCE, and SDK Permissions.

9. Open Developer Console in a new tab and click New Application [05:22].

10. Set Application Name (e.g., FAA Training Expert App) and select your Organization [05:30].

11. Set Application Type to Client-Facing Application [05:55].

12. Auto-fill the OAuth Redirect URI and click Create Application [06:03].

13. Configure Platform SDK & Ontology SDK Settings: Grant Platform SDK Chatbot API rights and Ontology scope.

14. Platform SDK Setup:

• Under Platform SDK, toggle ON the AIP Chatbots API scopes [06:31].
• Under Project Access, click Add Project and select the project directory containing your Chatbot (AIP Agents Project) [06:59]. Click Save.

1. Ontology SDK Setup:

• Select your active Ontology [07:24].
• Under Object Types, select and add all contextual objects used by the Chatbot (Gina FAA Chunk) [07:36]. Click Save Changes.

1. Scaffold React Repository and Vibe Code UI: Provision workspace and build custom aviation-themed interface.
2. Click Start Developing $\rightarrow$ Generate SDK [07:51].
3. Click Create Code Repository and open it in VS Code Workspaces [08:10].
4. Click the yellow Configure OAuth banner to register the workspace preview domain [08:35].
5. Open Continue.dev (or your AI coding assistant) and submit the prompt [09:29]:

"Build a single-page React app with a chat window that uses this agent RID: ri.aip-agents.... Make the styling aviation-themed with white and navy blue accents, sans-serif typography, minimal and focused."

1. Allow Continue.dev to inspect Platform SDK specs and build the UI components [10:24].

2. Authorize and Validate Application: Authorize OAuth grant and test real-time RAG responses.

3. Open the app preview window and grant the OAuth Authorization Request ("Allow application to access data on your behalf") [11:17].

4. Submit test queries (e.g., "How does density altitude affect takeoff performance?") [11:25].

5. Verify that the agent streams vector-retrieved answers with embedded object citations from the underlying FAA manual chunks [11:49].

---

Detailed Use Case Analysis: FAA Pilot Training Assistant

The video demonstrates the pattern using an FAA Aeronautical Pilot Knowledge Handbook chunked media set [01:08].

Architectural Comparison: Workshop vs. Custom React OSDK

Decoupling the AI chatbot from Foundry Workshop provides total frontend customization:

---

Process Audit Report

================================================================================
                        PROCESS AUDIT & VERIFICATION REPORT
================================================================================
Audit Target: YouTube Guidance & Use Case Analysis (rya3gIntUNY)
Auditor Skill: Process Audit
Verification Status: PASSED (100% Accuracy & Detail Alignment)

--------------------------------------------------------------------------------
[AUDIT CHECKPOINT 1: ACCURACY OF TRANSCRIPT DETAILS]
--------------------------------------------------------------------------------
✔ TOOL RENAMING & NAVIGATION:
  - Correctly notes AIP Agent Studio was renamed to AIP Chatbot Studio [00:01:21].
  - Launch command verified: Ctrl + J -> search AIP Chatbot Studio [00:01:31].
  - Context dataset accurately identified: FAA Pilot's Handbook of Aeronautical Knowledge (`Gina FAA Chunk`) [00:01:08].

✔ RETRIEVAL CONTEXT CONFIGURATION:
  - Vector search parameters accurately captured: Mode = Relevant Objects, Property = Vector Embedding [00:03:48].
  - Verified Chatbot RID location: View Details -> `ri.aip-agents...` [00:09:16].

✔ DEVELOPER CONSOLE & SDK CONFIGURATION:
  - App creation parameters verified: Client-Facing Application, OAuth Redirect URI auto-fill [00:05:55, 00:06:03].
  - Platform SDK setup verified: Enable AIP Chatbots API toggle, Add Project Access (`AIP Agents Project`) [00:06:31, 00:06:59].
  - Ontology SDK setup verified: Add contextual Object Types (`Gina FAA Chunk`) [00:07:36].

✔ REACT CODING & AUTHORIZATION:
  - Workspace OAuth PKCE authorization step verified [00:08:35].
  - Vibe coding prompt details (aviation theme, white/navy blue accents, minimal typography) verified [00:09:29].
  - Runtime OAuth user permission grant modal verified [00:11:17].

--------------------------------------------------------------------------------
[AUDIT CHECKPOINT 2: FORMAT & VISUAL COMPLIANCE]
--------------------------------------------------------------------------------
✔ STRUCTURAL LAYOUT:
  - Process map built with clear ASCII workflow diagrams.
  - Step-by-step guidance formatted using valid <Sequence> and <Step> tags with titles/subtitles.
  - Architectural comparison table included.
  - Image tags included and sourced from valid tool executions.
================================================================================

---