feat(ai): update code commenter, title generator, and add class explainer utils

b

perf(ai): implement high-speed snippet strategy for tags, attributes, CSS and JS

feat(ai): strengthen comment density prompt and fix UI refresh on AI update

fix(backend): update endpoints to handle extra frontend fields, fix health check, and improve AI HTML extraction

fix: set nemotron-nano-12b-v2-vl as primary model

debug: add full trace logging to diagnose model failures

refactor: replace cloudflare with nemotron-nano as primary, pure openrouter chain

fix: validate cloudflare output has comments/valid title before accepting

fix(cloudflare): merge system prompt into user prompt to force instruction following

fix(cloudflare): enforce explicit max_tokens to prevent truncation bugs

feat: replace primary model with Cloudflare Workers AI

fix: remove failing mistral model from fallback chain

fix: replace huggingface fallback with openrouter mistral

fix: switch to flash-lite and phi-3-mini-4k to resolve API limits

feat: complete 5-tier Gemini-first fallback architecture

Swap model order and enforce inline comment format

Update fallback AI models for stable benchmarking

feat: benchmark/production mode split, 3-retry backoff, /health_check endpoint

feat: add llama-3.3-8b to openrouter fallback chain

chore: fix huggingface failover timeouts and model

v2

google/gemma-4-27b-it:free

AI_Commentor-V0.1

RAG V1

v2.5

v2.4

v2.4

v2.4

v2.3

v2.3

v2.2

v2.2

v2

Fix: Preserve user HTML content when adding AI comments

Heroku deployment setup