I just pushed v22 of my project : a local AI companion for Radarr, that goes beyond generic genre or TMDb lists.
This isn’t “yet another recommender”. It’s your personal taste explorer that actually gets the vibe you want in natural language and builds recommendations starting from your existing library.
Key highlights from a real recent run:
- Command:
--mood "dystopian films like Idiocracy, Gattaca or In Time" - Output: Metropolis (1927), V for Vendetta, Children of Men, Brazil (1985), Minority Report, Dark City, Equilibrium, Upgrade, The Road… → oppressive/surveillance/inequality/societal critique atmosphere, not just “dark sci-fi”.
How it works :
- Starts by sampling random movies from your Radarr collection (or uses your mood/like/saga input).
- Asks a local Ollama LLM (e.g. mistral-small:22b) for 25 thematic suggestions based on atmosphere/vibe.
- Validates each via OMDb (IMDb rating, genres, plot, director, cast…).
- Scores intelligently: IMDb rating + genre match + director/actor bonus + plot embedding similarity (cosine on Ollama embeddings).
- Adds the top ones directly to Radarr (with confirmation: all / one-by-one / no).
- Persistent blacklist to avoid repeats.
Different modes :
--mood "dark psychological thrillers with unreliable narrators", any vibe you describe--like "Parasite" --mood "mind-bending class warfare"(or just--like "Whiplash")--saga(auto-detects incomplete sagas in your library and suggests missing entries) or--saga "Star Wars"--director "Kubrick"/--actor "De Niro"/--cast "Pacino De Niro"(movies where they co-star)--analyze→ full library audit + gaps (e.g. “You’re missing Kurosawa classics and French New Wave”)--watchlist→ import from Letterboxd/IMDb--auto→ perfect for daily cron / Task Scheduler (wake up to 10 fresh additions)
Standout features:
- 100% local + privacy-first (Ollama + free OMDb API only)
- No cloud AI, no tracking
- colored console output, logs, stats, HTML/CSV exports
- Synopsis preview before adding
- Configurable quality profile, min IMDb, availability filters
- Works on Windows, Linux, Mac
GitHub (clean single-file Python script + great README):
https://github.com/nikodindon/radarr-movie-recommender
If you’re tired of generic Discover lists, Netflix-style randomness, or manual hunting give it a spin. The vibe/mood mode + auto saga completion really change how you expand your collection.
Let me know what you think, any weird mood examples you’d like to test, or features you’d want added!
I remember building something vaguely related in a university course on AI before ChatGPT was released and the whole LLM thing hadn’t taken off.
The user had the option to enter a couple movies (so long as they were present in the weird semantic database thing our professor told us to use) and we calculated a similarity matrix between them and all other movies in the database based on their tags and by putting the description through a natural language processing pipeline.
The result was the user getting a couple surprisingly accurate recommendations.
Considering we had to calculate this similarity score for every movie in the database it was obviously not very efficient but I wonder how it would scale up against current LLM models, both in terms of accuracy and energy efficiency.
One issue, if you want to call it that, is that our approach was deterministic. Enter the same movies, get the same results. I don’t think an LLM is as predictable for that
Maybe lowering the temperature will help with this?
Besides, a tinge of randomness could even be considered a fun feature.
I’m not an expert, but LLMs should still be deterministic. If you run the model with 0 creativity (or whatever the randomness setting is called) and provide exactly the same input, it should provide the same output. That’s not how it’s usually configured, but it should be possible. Now, if you change the input at all (change order of movies, misspell a title, etc) then the output can change in an unpredictable way
Yes. I think determinism a misunderstood concept. In computing, it means exact same input leads to always the same output. Could be a correct result or entirely wrong, though. As long as it stays the same, it’s deterministic. There’s some benefit in introducing randomness to AI. But it can be run in an entirely deterministic way as well. Just depends on the settings. (It’s called “temperature”.)