Remove diagnostic startup print statements

Add startup print statements to diagnose Space hang

Fix startup hang: force CPU-only torch to avoid 2GB CUDA packages on CPU Space

Update recommendations.md with DeepInfra vs Novita comparison

Add recommendations.md and ignore .code-workspace files

Switch streaming call to provider-aware chat_completion

CHANGELOG: add commit hash for streaming change

Enable streaming responses for Sanyu chatbot

CHANGELOG: note HF token swap (Timothy Kintu) and credit incident

Revert to Qwen2.5-72B-Instruct; GLM-5.2 depleted free credits

Test GLM-5.2 (753B MoE) on Novita; fallback is Qwen2.5-72B

Revert to Qwen2.5-72B-Instruct; 8B failed prompt reasoning test

CHANGELOG: document model/prompt changes and pending 8B assessment

Try meta-llama/Llama-3.1-8B-Instruct on Novita; fallback is Qwen2.5-72B

Fix model not supported: switch to Qwen2.5-72B-Instruct on Novita

Switch to raw httpx call to expose full Novita 400 error body

Add debug logging to capture full 400 response body from Novita

Fix 400 Bad Request: replace non-ASCII chars unsupported by Novita API

Upgrade META_PROMPT to v2 (counsellor-reviewed)

Add CHANGELOG.md covering full project history

Add thumbs-up/down feedback system with comment form

Add conversation logging to atwine/CAPS-Sanyu-conversations dataset

Fix model not supported on Novita; fix Gradio 6 theme deprecation warning

Switch LLM to Llama-3.1-70B-Instruct

Fix double system message bug; add RAG_INSTRUCTIONS; explicit empty-context fallback

Fix TypeError: move theme from ChatInterface to wrapping gr.Blocks

Fix TypeError: remove deprecated Chatbot params bubble_full_width and show_copy_button

Improve UI: warm theme, Sanyu persona, starter prompts, layout tweaks

Fix 400 Bad Request: switch provider from hf-inference to novita

Switch LLM from Gemini to Llama-3.2-3B-Instruct via HF Inference API

Add retry logic for Gemini 503 UNAVAILABLE errors

Replace FAISS index with portable numpy embeddings in pkl

Fix retrieval: bypass FAISS search() using numpy dot product

Fix FAISS search() for both old and new API versions

Fix FAISS search() API compatibility for newer versions

Add RAG pipeline: build_index.py, retrieval.py, and FAISS knowledge base

Update README with full project documentation

Longer Prompt with improved formatting

Reverting to longer prompt

short meta-prompt version 1

Trying a really long system prompt

Fix content extraction from Gradio message format

Handle both dict and tuple history formats

Remove type argument, fix history as tuples

Fix history format for multi-turn conversation

Change to gemini 2.5

Update app.py

Trying a different gemini model

Update app.py

Switch to smaller model