Единая логика сводки и авто-целей оптимизатора (buildSummaryData)

feat(optimizer): stage and goal selection controls

fix(bm25): strict n-gram remove guidance

fix(ui): restore optimizer start

feat(optimizer): add diff highlighting

chore(hf): redeploy Space with latest LLM optimizer

chore: trigger Hugging Face Space rebuild

optimizer: per-goal deficit-based iteration and candidate budget

refactor(optimizer): apply unified per-goal iteration budget across all stages

fix(optimizer): keep BERT stage from plateau auto-skip

fix(optimizer): require bert target for each phrase

fix(ui): apply optimized Title from analysis snapshot + resolved_title in metrics

fix(optimizer): reject title goal in _llm_edit_chunk; doc deploy check for title_rewrite

fix(optimizer): title stage rewrites Title field only, return optimized_title

fix(optimizer): define _emit before first call (avoid UnboundLocalError)

Optimizer: SSE log panel + static app.js (SES-safe nv), no global loader; HF_SES doc

Revert "Optimizer: SSE progress, cancel job with partial result, UI progress card"

Revert "fix(ui): CSP-safe buttons — data-app-action delegation (HF Spaces)"

fix(ui): CSP-safe buttons — data-app-action delegation (HF Spaces)

Optimizer: SSE progress, cancel job with partial result, UI progress card

Revert to pre-progress-bar (ad42da6): remove SSE endpoint, optimizer callbacks, progress UI

Revert UI to pre-progress onclick buttons; keep optimizer SSE progress only

UI: replace inline onclick with addEventListener (SES/HF-safe)

Fix UI: var globals for inline handlers, safer loadProject/forceHideLoader

Optimizer: SSE progress stream + UI progress bar during LLM run

N-gram: rank by Freq/Avg, extra loop budget, overlapping window chunking

Enforce per-term ngram loop and underrepresented-only targeting.

Split n-gram summary into underrepresented and overrepresented blocks.

Align n-gram selection rules across summary and optimizer.

Implement n-gram tolerance stage rules and optimizer principles doc.

Add configurable BERT stage target and stage visibility in optimizer.

Implement staged multi-objective optimization flow after BERT.

Add ensemble phrase strategy mode for optimizer.

Add configurable phrase strategy mode for LLM optimizer.

Add multi-objective candidate utility for optimizer decisions.

Revert "Improve universal chunk selection with structural and diversity signals."

Improve universal chunk selection with structural and diversity signals.

Refine LLM prompt for semantic-vector chunk optimization.

Apply local-first accumulation for chunk-level optimizer progress.

Accept incremental BERT phrase gains during optimization.

Add semantic gap tolerance and detailed gap attribution.

Improve optimizer attribution and multi-edit batch search.

Add deep LLM-input diagnostics for chunk optimization debugging.

Implement two-stage optimizer acceptance with multi-chunk batching.

Improve optimizer span selection to avoid repeated chunk edits.

Trigger rebuild on Hugging Face without app code changes.

Expand full functional documentation with algorithmic details.

Improve optimizer candidate validation and BERT goal tracking.

Implement cascade state-machine for optimizer edit escalation

Add detailed optimizer debug trace for candidate-level visibility