454d6cbe52
The existing-message overflow split path in stream_consumer.run() sealed the first chunk via _send_or_edit(chunk) (finalize=False) then reset _message_id to None — so that chunk was never edited again and never received the adapter's final rich-text pass. On Telegram, MarkdownV2 formatting is applied on the finalize edit, so early split messages of a long multi-part streamed reply rendered raw markdown (##, **bold**, code fences) while only the last chunk rendered correctly. Fix: seal the overflow chunk with finalize=True so it gets its final formatting pass before _message_id is cleared. Salvaged from #32609 (the streaming-format portion only; the PR's send_draft parse_mode change is already superseded on main, and its media-roots change conflicts with the current denylist + recency-window delivery model). |
||
|---|---|---|
| .. | ||
| lib | ||
| tests | ||
| whatsapp-bridge | ||
| analyze_livetest.py | ||
| benchmark_browser_eval.py | ||
| build_model_catalog.py | ||
| build_skills_index.py | ||
| check-windows-footguns.py | ||
| contributor_audit.py | ||
| discord-voice-doctor.py | ||
| docker_config_migrate.py | ||
| hermes-gateway | ||
| install.cmd | ||
| install.ps1 | ||
| install.sh | ||
| install_psutil_android.py | ||
| keystroke_diagnostic.py | ||
| kill_modal.sh | ||
| lint_diff.py | ||
| LIVETEST_README.md | ||
| profile-tui.py | ||
| release.py | ||
| run_tests.sh | ||
| run_tests_parallel.py | ||
| sample_and_compress.py | ||
| setup_open_webui.sh | ||
| tool_search_livetest.py | ||