Musicpizzaman/Gemma4-E2B-it-CustomUI-llamafile

Gemma4-E2B-it-Custom.llamafile

Custom local llamafile package for Gemma 4 E2B IT. This is not an official Google, Unsloth, Mozilla, llamafile, or llama.cpp release.

Repository: https://huggingface.co/Musicpizzaman/Gemma4-E2B-it-CustomUI-llamafile

The single executable includes:

• Gemma 4 E2B IT Q3_K_M GGUF
• multimodal projector for image understanding
• custom browser UI with beginner-friendly setting tooltips
• image, PDF, text, code, and large CSV upload handling
• 50,000-token attachment budget
• multiple conversations and Markdown session save/load
• active context feedback and restart arguments
• response timing and meaning-preserving context compaction

Run:

.\Gemma4-E2B-it-Custom.llamafile

Then open:

http://127.0.0.1:8080/?v=context-sync-v10

On Windows, if double-click launch does not work, copy or rename the file with .exe at the end.

Context length

The embedded default is 128K tokens. Context is allocated at startup:

.\Gemma4-E2B-it-Custom.llamafile
.\Gemma4-E2B-it-Custom.llamafile --ctx-size 65536
.\Gemma4-E2B-it-Custom.llamafile --ctx-size 32768

The UI distinguishes the active server context from the saved restart target. Restart the process after changing the target. Shorthand such as - 31k is not a valid llamafile argument.

Files

Gemma4-E2B-it-Custom.llamafile
Size: 3570099395 bytes
SHA-256: 81B517C594EB73518A83AE5072EDB6834250B0FA6AE9A0EB443B355A6EAFD57C
UI build: context-sync-v10

Gemma4-E2B-Custom-Llamafile-Software-Design.pdf

The package runs locally by default and binds to 127.0.0.1. Building, downloading, or uploading the package requires internet access.