docs: remove remaining plain-english mention

feat: upgrade agent to llama-3.3-70b

docs: remove 'plain english' phrases

fix: chart visibility in print mode

fix: track event_logs.json in git

track csv with lfs

feat: migrate to Chronos-2 zero-shot with temperature covariates

ui/backend: pivot from chronos-2 multivariate to chronos-bolt-base fine-tuning architecture

ui: update chronos loading screens and progress labels to reflect chronos-2 multivariate model

fix: change hardcoded base model from chronos v1 to v2 to avoid pipeline loading warnings

fix: add robust regex json extraction in LangGraph agents to prevent JSONDecodeError from trailing LLM output

fix: correct quantile array shape handling for Chronos 2 predict_quantiles output

chore: explicitly track pjm datasets in gitignore

fix: bump scikit-learn to >=1.6.0 to resolve chronos-forecasting dependency conflict during docker build

fix: replace HereDoc Dockerfile syntax with explicit entrypoint file for HF Spaces compatibility

fix: correct static type checker error for pyrefly and narrow properties explicitly

feat: upgrade Chronos forecasting engine to Chronos-2 with 8192 token multivariate context window

feat: upgrade SARIMA to multivariate using temperature exogenous variables

perf: increase chronos samples to 500 and remove deterministic constraints

feat: remove replay topology button, enforce strict DB event fetching to stop hallucination, bump chronos num_samples to 100

perf: increase chronos sample size to 40 for smoother quantiles without breaking time budget

feat: tune chronos hyperparameters for deterministic accuracy

style: display available dataset range in target date help text

fix: add min and max bounds to targetDate input

style: remove unnecessary float() casting

fix: initialize data_stats to prevent NoneType error

feat: add backtest target date selector

style: update node 5a and 5b descriptions and explain conditional routing

style: update node 3 and 4 descriptions for grid operators

style: remove pipeline vertical line, relocate parallel execution text, remove auto refresh note

style: rewrite node 2a and 2b descriptions and explain parallel execution

style: update data validator description for grid operators

style: ui wording tweaks and default config updates

style: force page break before seasonality analysis tab on pdf print

style: fix pdf navbar, fix gauge background, update mock mandate text

style: fix pdf export overlapping charts, invisible gauge text, and hide top navbar

style: overhaul @media print rules for perfect pdf exports

style: remove context box from divergence report

style: make divergence and seasonality grids responsive for mobile screens

fix: restore stakeholder summary layout to fix reference error

style: layout refinements for tabs, heatmap, and removing trace

fix: use true semantic search vector score instead of pseudo random

style: center anomaly formula and add formula codeblock to model divergence metric

feat: add interactive replay modal, dynamic mandate tone, outcome badges, and vector scores

feat: completely overhaul RAG Historical Memory section to be an actionable threat dossier

fix: switch to llama-3.1-8b-instant and build frontend/dist to serve new ui changes

feat: add exact mathematical formulas for anomaly severity and divergence directly into UI boxes with dynamic placeholders

fix: switch LLM from llama-3.3-70b to llama3-8b-8192 to bypass daily Groq token limits

feat: hardcode 3-part semantic risk narrative in agent pipeline using dynamic dashboard values

feat: force LLM to output highly detailed, semantic 3-part operational briefing for summary rationale