Repository navigation
Home
Kevin ortega edited this page Feb 6, 2026
·
2 revisions
Welcome to the official CineGram Wiki. Here you will find detailed documentation on internal mechanics.
- Cause: Python process might be hung.
- Solution: Check the terminal. Ensure internet connectivity.
- Cause: Invalid API Key or TMDB is blocked on your network.
-
Solution: Verify
TMDB_API_KEYin.env. Restart router if using dynamic IP.
- Cause: Ollama is not running.
-
Solution: Open terminal and run
ollama serve. The bot performs a "Health Check" at startup to warn you.
- Cause: Timeout downloading image from TMDB.
- Solution: The bot has a 10s timeout. If your net is slow, it might fail. Only a black placeholder is used to prevent crashing.
CineGram uses a 3-layer system to identify movies:
-
Regex Filtering (Fast):
- Uses
guessitto split "Movie.2024.mp4" into Name and Year. - If valid and found on TMDB, stops here.
- Uses
-
Spam Cleaning (Intermediate):
- If name contains "CUEVANA", "LATINO", "1080p", it aggressively strips these terms.
-
Artificial Inference (Slow but Precise):
- If all else fails, it sends the filename and caption to Ollama.
- System prompt: "Act as a movie expert. Extract the real title from this garbage text...".
- AI returns clean JSON:
{"title": "The Matrix", "year": "1999"}.
To prevent cropping heads in vertical posters when fitting to horizontal (Youtube/Telegram):
- Take original image.
- Create 1920x1080 background.
- Stretch original to fill background, apply Blur (30px) and Darkening.
- Paste original (aspect fit) in center.
- Overlay text and logos.
Ensures 100% readability and premium aesthetics.