Commit History

fix: Robust Welcome Message (Sync Init)
524e7b6

neuralworm commited on

fix: Refine Oracle Output (Remove ELS Text, Ensure English)
baca426

neuralworm commited on

fix: Restore Legacy Parity for Oracle (Sura=Quran, Rev=Matthew)
787e371

neuralworm commited on

fix: Robust Deployment & Global Translation
932d1ec

neuralworm commited on

fix: Chat Output & Tool Hiding
5713c03

neuralworm commited on

feat: Agentic Positive Prompting & Scenario Tests
058ae1e

neuralworm commited on

UI: Simplified layout and fixed overflow bug
4718250

neuralworm commited on

Refactor: Modularized codebase and consolidated Oracle bridges
b7f3dcc

neuralworm commited on

Fix Role Alternation crash for assistant-led conversations and extend simulation tests
de6582c

neuralworm commited on

Sage 6.5: Transform into lean text-only agent, remove RAG/Voice/Mongo, optimize model loading, and extend UI/Logic verification tests
d1af7e9

neuralworm commited on

fix: agent history integration, OOM protection, and UI cleanup
e09f395

neuralworm commited on

feat: minimalist UI, auto-translation with phrase-based cache, placeholder update
32d508e

neuralworm commited on

feat: remove MongoDB and Sacred Knowledge tab, add agent welcome message, fix typos
80db0c8

neuralworm commited on

chore: save RAG WIP as patch
c42114b

neuralworm commited on

feat: Enforce global reproducibility with seed 42 (random, numpy, torch)
bbb5fd8

neuralworm commited on

refactor: Pass model_size explicitly (1b for HF, 4b for Local) instead of using ENV
3160a58

neuralworm commited on

fix: Use AutoModelForCausalLM for Gemma 3 1B (text-only) to fix loading issues
ac1099d

neuralworm commited on

feat: Explicit AutoProcessor/Tokenizer logic and verified 1B/4B/4B-Quant tool calling
61324b6

neuralworm commited on

feat: Add MODEL_SIZE env var (1b/4b) and fix loading logic (skip quantization for 1b)
af14da0

neuralworm commited on

feat: Switch back to gemma-3-4b-it but enable 4-bit quantization (BNB) to fix memory/tool issues
2e496bf

neuralworm commited on

fix: Implement fallback to AutoTokenizer for text-only Gemma models (1B)
8e936bf

neuralworm commited on

feat: Switch to lighter google/gemma-3-1b-it model to fix HF Spaces OOM
7737f23

neuralworm commited on

fix: Disable disk offloading by removing device_map='auto' and using explicit .to(device)
61c8e04

neuralworm commited on

chore: Pin ML dependencies (torch, transformers, etc.) to exact local versions
49d012d

neuralworm commited on

fix: Remove server_name arg from app.py for HF Spaces compatibility
490b060

neuralworm commited on

refactor: Modularize app structure (app_module.py + thin entry points)
195c86d

neuralworm commited on

chore: Prepare app.py for HF Spaces (remove SSL/share), keep local config in app_local.py
6c97eaf

neuralworm commited on

chore: Pin langchain dependencies to exact versions
20ffe70

neuralworm commited on

chore: Pin Gradio to 5.49.1 and unpin python-Levenshtein
2c63a44

neuralworm commited on

feat: Add support for 'name' argument in Oracle tool calls
f126517

neuralworm commited on

fix: Convert embedded repo to regular directory
e8510d3

neuralworm commited on

feat: Integrate daily-psalms-api and configure Git LFS for DBs
24f87c3

neuralworm commited on

Merge branch 'main' of https://huggingface.co/spaces/neuralworm/gemma-sage
cb2b9a6

neuralworm commited on

initial commit
28ad2a7
verified

neuralworm commited on

fix: Improve conversation fluidity with prompt injection and verified filtering
d5040cf

neuralworm commited on

feat: Agentic Oracle, Multi-Message Bubbles, Filtering, and UI Fixes
8ab43a3

neuralworm commited on

add openai data export compatibility
24766ca

neuralworm commited on

add openai data export compatibility
dac3615

neuralworm commited on

initial commit
7bcd142

neuralworm commited on

initial commit
af1dacb
verified

neuralworm commited on