Plex Concierge - A Guide to Overseerr ⤫ Custom GPT

Table of Contents

• Overview
• Alternatives
• Functionality
• Prerequisites
• Guide
• Notes
• Explanation
• Ideas
• Screenshots

---

Overview

This is a guide for how to create a media discovery and management chat interface by connecting Overseerr to ChatGPT's Custom GPT tools. You can use this yourself, share it with friends, or connect it to a chat application!

---

Alternatives

If you're looking for other chat-to-Plex tools, they exist! This is a new approach to an already-solved problem, with the benefit of a nice conversational flow.

• Requestrr - A Discord chatbot (GitHub)
• Doplarr - A Discord request bot (GitHub)
• Presumably more

---

Functionality

• Adding movies and shows to Plex via natural language
  • "Hey, can you add Big Buck Bunny?"
• Display trailer links for movies/shows discussed
• Disambiguates content with similar names
  • → "Did you mean Point Break 1991, or Point Break 2015?"
• Understanding whether you're asking to add 4K or non-4K content
  • "Can you add The Boys in 4k?"
• Offering progress updates and ETAs on downloads/requests
  • "How long until The Simpsons is ready?"
  • → "There are a few episodes still downloading, and they should all finish in about 2 hours and 25 minutes"
• Providing direct links to watch content on Plex
• Support for bulk requests
  • "Add the dark knight, the Simpsons, and the Simpsons movie"
• Handling recommendations or general questions about content
  • "What are Clint Eastwood's best movies?"
  • → (Results)
  • "Thanks, can you add the top 3?"

---

Prerequisites

• A ChatGPT Plus subscription (In order to create Custom GPTs)
• Overseerr (Configured with Sonarr and Radarr)
  • Your Overseerr instance must be available from the internet and have some sort of static URL/address.
  • Supports additional Sonarr and Radarr 4K instances, but not required.

---

Guide

1. Go to the My GPTs section of the ChatGPT website and click "Create a GPT". (GPT Plus Subscription required)
2. Click on the 'Configure'. Provide a name and description. Add an avatar if desired.
3. For 'Instructions', use the provided gpt-instructions.md file (use the raw version with markdown formatting).
4. (Optional) Add conversation starters, such as:
   • "Request a new show or movie"
   • "Find recommendations for popular movies and shows"
5. Configure 'Capabilities':
   • Web Search: On
   • Canvas: Off
   • DALL•E Image Generation: Off
   • Code Interpreter & Data Analysis: Off-
6. By this point, it should look something like this:

7. Create a New Action:
   • Authentication: Choose "API Key". Retrieve the API key from Overseer's Settings → General tab.
   • Schema: Use the provided overseerr-actions-schema.yml.
     • ❗️ Replace the server URL with your Overseerr URL.
   • Privacy Policy: Add any link (required if sharing the GPT).
   • By this point, it should look something like this:

8. Hit the 'Update' button in the top right.
9. Adjust privacy settings as desired. Use a private link for sharing; avoid making this public.
10. Done! Test and explore its features.

---

Notes

• Parameter Workarounds: Some unusual instructions were added in the schema, e.g., making 'seasons' required and defaulting it to 'all' for TV shows fixes issues with missing parameters.
• Search Behavior: Initial search queries may fail before retrying correctly due to URL encoding errors. This does not impact functionality.
• Debugging: Use the GPT editor's testing conversations to inspect request/response payloads sent to the Overseerr API.
• Custom GPT Limitations: GPTs can have up to 30 actions. Simplify schemas for complex tasks. This guide uses only 2 actions (search and request).

---

Explanation

As a bit of background for the steps I took to do this, in case you were curious to try it out yourself from scratch for this or another use case:

1. I read about Custom GPTs using OpenAI's Guide
2. I looked at Overseerr's API Documentation and checked that it had the sort of functionality I wanted to use (in this case, I wanted to search for movie and tv content, and make requests to add it to Plex)
3. I read about GPT Actions
4. I read the GPT Actions Getting Started Guide
5. I chatted with OpenAI's Actions GPT, and asked to read the OpenAPI Schema for Overseerr, and generate a new spec for just the /search and /request endpoints
6. I created the new Custom GPT, and created a new Action.
   1. I configured additional settings for API authorization
   2. I provided the schema I received from the Actions GPT chat
   3. I updated the schema to point to my Overseerr instance, since it defaulted to localhost
7. I tuned the operation of the Custom GPT through a mixture of prompted-edits and manual edits to it's instructions.
   1. After a period of tuning, it was getting a bit confused with conflicting instructions and edge-cases. I used a separate ChatGPT chat to ask for it to consolidate the instructions and provide the result as raw markdown, which I copied back into the instructions section of the custom GPT
8. I used the debugging functionality available in the Custom GPT editor's chat window to test out queries and inspect request/response payloads when things went wrong

---

Ideas

• Connect to Discord or WhatsApp for easier sharing and accessibility.
• Extend Functionality by adding new actions:
  • Deleting Content:
    • DELETE /media/{mediaId}
  • Fetching Ratings Data:
    • GET /movie/{movieId}
    • GET /tv/{tvId}
  • Fetching Similar Content:
    • GET /movie/{movieId}/similar
    • GET /movie/{movieId}/recommendations
    • GET /tv/{tvId}/similar
    • GET /tv/{tvId}/recommendations
• Integrate Sonarr/Radarr APIs for enhanced library management:
  • Force re-downloads
  • Specify preferred releases
• Deep Links for iOS Apps: Some bugs exist with deep links in ChatGPT. Reported issues pending resolution.

---

Screenshots

Adding a show and seeing download progress

Adding a show that already exists

Adding 4K Content

Checking Availability on Plex

Querying Download Progress

---