feat(telegram): attribute forwarded messages #37

Merged

weiwen merged 1 commit from sandcastle/issue-25 into main

2026-07-05 14:11:36 +08:00

weiwen commented

2026-07-05 13:41:58 +08:00

Owner

Summary

When a Telegram message is a forward, prepend a compact attribution prefix to its text before it reaches the pi pipeline, so the model knows where the content came from.

Format: [Forwarded from <name>, YYYY-MM-DD] — or [Forwarded] when the origin has no displayable name (e.g. a nameless chat).

What changed

format_forward_prefix(origin: &MessageOrigin) — maps all four MessageOrigin variants (User, HiddenUser, Chat, Channel) to a human-readable attribution string with an ISO date.
apply_forward_prefix(extraction, prefix) — prepends the prefix to Extraction::Content text; leaves Unhandled and other variants untouched.
extract_message_content now wraps the renamed extract_message_content_raw, reading msg.forward_origin() and applying the prefix in one place before returning.
Five unit tests cover both functions: named user, hidden user, prefix applied, prefix absent (None), and Unhandled passthrough.

Key decisions

Attribution is injected as plain text rather than a separate metadata field — keeps the pi integration simple with no protocol changes.
Date format is YYYY-MM-DD for unambiguous, locale-neutral display.

Reviewer checklist

Prefix format reads naturally in actual pi conversations
All four MessageOrigin arms are handled (confirm no future variants are expected by teloxide)
Tests cover the edge cases you care about

Closes #25

## Summary When a Telegram message is a forward, prepend a compact attribution prefix to its text before it reaches the pi pipeline, so the model knows where the content came from. **Format:** `[Forwarded from <name>, YYYY-MM-DD]` — or `[Forwarded]` when the origin has no displayable name (e.g. a nameless chat). ## What changed - **`format_forward_prefix(origin: &MessageOrigin)`** — maps all four `MessageOrigin` variants (User, HiddenUser, Chat, Channel) to a human-readable attribution string with an ISO date. - **`apply_forward_prefix(extraction, prefix)`** — prepends the prefix to `Extraction::Content` text; leaves `Unhandled` and other variants untouched. - **`extract_message_content`** now wraps the renamed `extract_message_content_raw`, reading `msg.forward_origin()` and applying the prefix in one place before returning. - Five unit tests cover both functions: named user, hidden user, prefix applied, prefix absent (None), and Unhandled passthrough. ## Key decisions - Attribution is injected as plain text rather than a separate metadata field — keeps the pi integration simple with no protocol changes. - Date format is `YYYY-MM-DD` for unambiguous, locale-neutral display. ## Reviewer checklist - [ ] Prefix format reads naturally in actual pi conversations - [ ] All four `MessageOrigin` arms are handled (confirm no future variants are expected by teloxide) - [ ] Tests cover the edge cases you care about Closes #25

weiwen added the

ready-for-human

label

2026-07-05 13:41:58 +08:00

weiwen added 1 commit

2026-07-05 13:41:58 +08:00

RALPH: feat(telegram): attribute forwarded messages (closes #25 )

CI / check (pull_request) Failing after 56s

Details

87513d3163

Task: Detect Telegram forward_origin on incoming messages and prepend a
compact attribution line so pi knows the content originated elsewhere.

Key decisions:
- format_forward_prefix(origin: &MessageOrigin) -> String: maps all four
  MessageOrigin variants (User, HiddenUser, Chat, Channel) to
  "[Forwarded from <name>, YYYY-MM-DD]"; falls back to "[Forwarded]" when
  no displayable name is available (e.g. Chat/Channel with no title).
- apply_forward_prefix(extraction, prefix) -> Extraction: pure helper that
  prepends the prefix string to Extraction::Content.text, leaving
  FileTooLarge and Unhandled untouched. Extracted for unit-testability.
- extract_message_content now delegates raw extraction to
  extract_message_content_raw (renamed from the previous body) then
  applies the prefix in one place, so text, photo, and document branches
  all benefit automatically.
- MessageOrigin added to the teloxide::types import list.

Files changed:
- src/telegram/mod.rs: format_forward_prefix, apply_forward_prefix,
  extract_message_content refactor, extract_message_content_raw,
  5 new unit tests (User prefix, HiddenUser prefix, prefix composition
  with content, None-prefix no-op, Unhandled pass-through).

Blockers/notes:
- No Rust toolchain in sandbox; CI must validate compilation and tests.
  Logic follows the same async extraction pattern used for documents and
  photos merged in prior commits.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

weiwen merged commit 6087d159e9 into main

2026-07-05 14:11:36 +08:00