fix: adjust temperature setting for generation to improve output quality

fix: виправлення кодування Промт 8.txt (CP1251 → UTF-8) та форматування prompts.py

feat: introduce few-shot prompting with semantic similarity for EU legal assistant

feat: setup progress bar for analysis action and swap batch/settings tabs

extend provider-specific prompt caching to analysis path

refactor prompt assembly for provider-specific caching

docs: Update HELP.md with analysis thinking info and latest models

feat: Add thinking parameters to Analysis mode and update models list in README

handoff: align help and trim debug leakage

handoff: harden config and align MVP surface

docs: mark legacy/summary files as pointers to canonical docs (dedup)

docs: reorganize markdown files into docs/; add docs index

docs: clarify scope (generation-first) and index snapshot freshness

feat: Refine legal position prompts for clarity and specificity in analysis workflow

feat: Add validation to skip duplicate header rows in batch processing

feat: Allow default analysis model in PrecedentAnalysisWorkflow if not specified

feat: Update OpenAI generation models to include display names and remove deprecated models

feat: Update OpenAI generation model configuration to include display name and default setting for GPT-5.4

feat: Update generation models in configuration to include GPT-5.4 as default

feat: Revise legal position prompt structure and rules for clarity and application

refactor: Update legal position prompts for clarity and specificity

feat: Add support for GPT-5.3 Chat Latest model in configuration and reasoning parameters

feat: Enhance OpenAI reasoning capabilities with configurable parameters

I cannot generate a commit message without knowing the specific changes made to `interface.py`.

test: Add script to verify Anthropic adaptive thinking and output effort parameters.

Support output_config effort parameter for Claude Adaptive Thinking

Re-enable extended thinking for Claude Haiku 4.5 models

Fix Anthropic thinking mode configuration logic

chore: Update dependencies and adjust interface logic.

chore: Update ignored files and refine prompt content.

feat: Introduce `document` input variable to the legal analysis prompt and update `PromptTemplate` import path.

feat: enhance text cleaning utility to remove HTML tags and entities, update documentation, and requirements.

feat: Enhance interactivity of Gradio interface and improve handling of empty response in PrecedentAnalysisWorkflow

feat: Update system prompt structure to include cache control in LLMAnalyzer and generate_legal_position functions

feat: Optimize caching for Anthropic and OpenAI prompts, restructure prompt variables for efficiency

feat: Initialize search components at module level for HF Spaces

feat: Improve user feedback during legal position generation with loading message and button interactivity

feat: Suppress asyncio event loop warnings on HF Spaces shutdown

feat: Enhance save button interactivity in Gradio interface for prompt changes

feat: Update system prompt and legal position template for clarity and specificity

feat: Add debug prompt logging functionality with environment variable support

feat: Оновлення UI для GPT-5.2 та додавання контролю MaxTokens

feat: Інтеграція GPT-5.2 з підтримкою reasoning

Add diagnostic logging and raw httpx connectivity test for OpenAI API

Disable HTTP/2 in httpx client for OpenAI and DeepSeek to avoid 421 Misdirected Request errors and ensure consistent connections in HF Spaces

Refactor httpx client usage for OpenAI and DeepSeek: implement context management to prevent socket leaks and improve error logging

Fix OpenAI/DeepSeek HF Spaces connectivity: use custom httpx.Client to handle asyncio conflicts

Enhance API key debugging: add length and prefix display for OPENAI_API_KEY; update provider availability checks to use os.getenv for real-time state

Enhance provider management: normalize response keys, add availability checks, and improve Gradio interface for AI providers

Revert httpx.Client customization: use default OpenAI SDK settings for async compatibility in HF Spaces