Major upgrade for v5.0: Context limit handling, drop Playwright, single v4.1flash model & more

This release marks a comprehensive upgrade to the work done in V4. It adds explicit context limit handling (configurable limits, summaries, and automatic handover to fresh chats), drops Playwright and Chromium entirely in favor of Android client headers (kept only as a deprecated fallback), and consolidates onto a single DeepSeek v4.1flash model (removing file forking and all other models). It also resolves server hangs after multi-turn conversations, cookie loss on restart, and long-context processing failures (via session recovery and streaming preflight checks), hardens DSML/tool-call parsing, adds standard en_US locale headers, introduces a LICENSE, and delivers a complete README refactor (contributors, star history, usage warnings), among other improvements.

What's Changed

• fix html/xml tag stripping in tool argument values by @flame-0 in #10
• Fix Anthropic tool_result handling: emit role=tool messages and assistant tool_calls by @obruniningo-glitch in #11
• Harden parse_tools: ignore tool markup in code fences, keep repeated calls and companion text by @obruniningo-glitch in #12
• Centralize parent_message_id bookkeeping in next_parent() helper by @obruniningo-glitch in #13
• Fix long-context failure: session recovery, streaming preflight, token-capped prompt rebuild by @AmanCode22 in #14
• Fix server freeze after many chats + cookie loss across restarts by @AmanCode22 in #15
• Deprecate Playwright & Chromium by using Android headers and keep them as backup by @alan7383 in #16
• Support single DeepSeek v4.1flash model by @AmanCode22 in #19
• Fix DSML tool streaming edge cases by @AmanCode22 in #20
• Explicit context policy: configurable limits+summerize and rollover to new chat by @AmanCode22 in #23

New Contributors

• @flame-0 made their first contribution in #10
• @obruniningo-glitch made their first contribution in #11
• @alan7383 made their first contribution in #16

Full Changelog: v4.0...v5.0