feat(telegram): support copy text buttons

[kilhyeonjun]

Summary

• add Telegram COPY_BUTTON: label | text markers that render as Bot API CopyTextButton inline keyboard buttons
• support both normal sends and streaming final edits, while leaving markers visible when CopyTextButton is unavailable
• document the marker syntax and constraints for Telegram users

Test Plan

• uv run --extra messaging python -m pytest tests/gateway/test_telegram_copy_buttons.py tests/gateway/test_telegram_clarify_buttons.py tests/gateway/test_telegram_rich_messages.py tests/gateway/test_telegram_thread_fallback.py -q -o 'addopts='
• git diff --check
• uv run --extra messaging python -m py_compile plugins/platforms/telegram/adapter.py tests/gateway/test_telegram_copy_buttons.py

[teknium1]

Thanks for the Telegram copy-button implementation. The feature is still absent from current main, but this branch needs focused salvage before it can be reviewed as that feature.

Problems

• The diff again contains unrelated Codex usage/credential telemetry and network-circuit work (agent/credential_usage.py:1, hermes_cli/codex_usage.py:1, agent/network_circuit_breaker.py:1), despite the linked #59443 discussion stating that those commits had been accidentally included. Please keep the Telegram feature isolated.
• plugins/platforms/telegram/adapter.py:3944 strips markers and creates copy_markup, but oversized final edits return through _edit_overflow_split at lines 3981-3985. That helper has no markup parameter (:4149-4157), so the claimed streaming-final support drops the copy keyboard for overflowed responses. tests/gateway/test_telegram_copy_buttons.py:180-202 covers only the short path.

Suggested changes

• Split the unrelated work from this PR and reconcile the Telegram change with the later adapter refactors on current main.
• Pass markup through the overflow-final path and add a >4096 UTF-16 streamed-final regression test.

Automated hermes-sweeper review.

[teknium1]

For a final response over 4096 UTF-16 units, this stripped content returns through _edit_overflow_split at lines 3981-3985. That helper accepts no copy_markup, so the keyboard computed here is lost before the normal edit_message_text(... reply_markup=...) path. Please plumb markup through the overflow split and add a >4096 streamed-final regression test.

[GottZ]

This was generated by AI during triage.

Summary

Three PRs address the Telegram copy-text-button feature by parsing COPY_BUTTON markers and attaching CopyTextButton keyboards, but #59443 and #59445 mix that implementation with unrelated Codex telemetry, while #69153 isolates the Telegram change on current main. Unlike the earlier diffs, #69153 also carries the keyboard through oversized streaming-final responses and fallback paths.

Related pull requests

• #59443 [closed] duplicate — (+1147/-10) — superseded: The closed PR implements marker parsing, normal sends, and short final edits, but most of its diff is unrelated Codex quota and credential-usage work; it remains relevant as the original implementation and was explicitly replaced by the cleaner #59445.
• #59445 related — (+2332/-11) — superseded: It retains the same Telegram implementation but again includes unrelated Codex telemetry and network-circuit changes, and its overflow-final path drops reply_markup. Despite the keep_open review on #59445, that review explicitly requires focused salvage; #69153 provides that clean split and fixes the identified overflow defect.
• #69153 duplicate — (+417/-14) — preferred: This current-main, Telegram-only diff preserves malformed markers, supports normal sends and short final edits, and passes copy markup into _edit_overflow_split, attaching it to the final continuation and preserving it through plain-text, reply, flood-retry, and reactive-overflow fallbacks.

Duplicates

#59443, #59445, and #69153 implement substantially the same Telegram COPY_BUTTON feature; #69153 is the clean superseding iteration, while #59443 and #59445 contain unrelated changes and lack complete overflow handling.

Suggested consolidation

Merge #69153 after normal maintainer review — it is the focused implementation and directly addresses the blocking scope and overflow concerns documented on #59445. Close #59445 as superseded by #69153; #59443 is already closed and remains superseded by the later clean iterations.

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 Dup59443 ["PRs duplicating each other"]
        P59443["PR #59443 (closed)"]
        P59445["PR #59445 (open)"]
        P69153["PR #69153 (open)"]
    end
    class P59443 closed
    class P59445 open
    class P69153 open
    class P59445 target
    click P59443 "https://github.com/NousResearch/hermes-agent/pull/59443"
    click P59445 "https://github.com/NousResearch/hermes-agent/pull/59445"
    click P69153 "https://github.com/NousResearch/hermes-agent/pull/69153"

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 3 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 198 kB of PR diffs, 3 kB of issue/PR text, 3 kB of discussion (4 comments), 3 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.

[GottZ]

This was generated by AI during triage.

Delta since our previous triage comment

@teknium1 extended our analysis by identifying an uncovered standalone Telegram delivery path in #69153: cron and send_message delegate to _send_telegram, which neither extracts COPY_BUTTON markers nor attaches reply_markup. This changes #69153 from ready for normal maintainer review to requiring focused salvage and standalone-path coverage before merge.

Changed pull requests

• #69153 duplicate — (+417/-14) — keep open: The focused diff fixes adapter sends and overflow/fallback edits, but the new keep_open review shows that standalone Telegram delivery still emits valid markers as visible text without copy-button markup; reuse the parser/markup logic in _send_telegram and test normal, chunked, and fallback sends.

Suggested consolidation

Keep #69153 open until the standalone Telegram path and tests are added; the prior recommendation to prefer #69153 over #59445 otherwise remains unchanged.

Complex graph unchanged since our previous triage comment.

Cross-PR triage: Reviewed 3 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 198 kB of PR diffs, 3 kB of issue/PR text, 3 kB of discussion (5 comments), 0 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.