fix(telegram): strip raw markdown in MarkdownV2 fallback paths

[Vitalymt]

Problem

When Telegram rejects MarkdownV2 formatting in edit_message(), the fallback sends raw content with literal **bold** markers and no parse_mode. The send() method already handles this correctly by calling _strip_mdv2().

Root Cause

Two code paths in edit_message() send text=content without stripping markdown on fallback:

1. MarkdownV2 parse failure: Falls back to raw content with no parse_mode
2. Flood control retry: Retries with raw content and no parse_mode

Additionally, _strip_mdv2() didn't handle raw **bold** patterns — only the MarkdownV2-converted *bold*.

Fix

Three changes in gateway/platforms/telegram.py:

1. _strip_mdv2(): Add regex to strip raw **bold** before the already-handled *bold*
2. edit_message() MarkdownV2 fallback: Use _strip_mdv2(content) + parse_mode=None (matching send() behaviour)
3. edit_message() flood-wait retry: Use already-formatted MarkdownV2 text, or fall back to _strip_mdv2()

Plus test update in tests/gateway/test_telegram_format.py.

Testing

All 101 tests in test_telegram_format.py pass.

[teknium1]

Thanks for tracing the fallback paths. Current main already contains the parse-error fallback and raw-**bold** stripping from this PR's scope (plugins/platforms/telegram/adapter.py:318-321,4012-4036, introduced by f4531feee).

Problems

• The short flood-control retry is still a live gap: after the initial formatted final edit raises RetryAfter, plugins/platforms/telegram/adapter.py:4085-4089 retries raw content without parse_mode.
• The PR's test covers parse-error fallback only (tests/gateway/test_telegram_format.py:891-907), not that RetryAfter branch.

Suggested changes

• Salvage the retry portion into plugins/platforms/telegram/adapter.py, preserving the finalized MarkdownV2 payload and parse mode on retry.
• Add a RetryAfter regression test asserting the retry arguments.

This is an automated hermes-sweeper review.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs address the same Telegram MarkdownV2 fallback defect with effectively identical diffs: both strip raw **bold**, send plain text after parse failure, and preserve the formatted payload during flood-control retry. Current main already contains the stripping and parse-error fallback portions, leaving the RetryAfter path as the remaining gap.

Related pull requests

• #33304 related — (+17/-4) — consolidate here, but revise before merge: despite the keep_open review on #33304, the diff is only partially applicable because main already contains the raw-**bold** stripping and parse-error fallback; retain the still-relevant formatted RetryAfter retry and add the requested regression test for its arguments.
• #34570 [closed] duplicate — (+17/-4) — duplicate of #33304: its diff is effectively identical and does not add the missing RetryAfter regression test. Although closed, it remains relevant as corroboration of the same failure mode and is superseded by the earlier PR and its contributor review context.

Duplicates

#34570 duplicates #33304; both apply the same production and test changes.

Suggested consolidation

Merge #33304 only after rebasing it onto the current plugins/platforms/telegram/adapter.py implementation, narrowing it to the remaining RetryAfter fix, and adding the contributor-requested RetryAfter regression test; keep #34570 closed as a duplicate of #33304.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup33304 ["PRs duplicating each other"]
        P33304["PR #33304 (open)"]
        P34570["PR #34570 (closed)"]
    end
    class P33304 open
    class P34570 closed
    class P33304 target
    click P33304 "https://github.com/NousResearch/hermes-agent/pull/33304"
    click P34570 "https://github.com/NousResearch/hermes-agent/pull/34570"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed or no verify verdict yet (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 7 kB of PR diffs, 2 kB of issue/PR text, 1 kB of discussion (2 comments), 1 verify verdict. verdicts reflect diff content, not PR titles. Part of an automated triage batch.