feat(ms-teams): two-way emoji reaction sync

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.

[KrisBraun]

What

Implements outbound emoji reaction write-back for the Microsoft Teams connector, completing two-way reaction sync.

Reactions a user adds/removes in Plot are now pushed to Teams via Microsoft Graph setReaction/unsetReaction. The runtime dispatches onNoteReactionChanged on the reacting user's own connector instance (routed via twist_instance_for_actor), so getApi resolves to that user's token and the reaction is attributed correctly in Teams — no actAs step (same model as Slack).

Details

• Channel messages and replies. A reply is addressed through its parent's replies/{id} collection, not at the top messages level. A new private GraphApi.channelReactionPath helper switches setChannelReaction/unsetChannelReaction to the reply-scoped endpoint when the note is a reply (note.key differs from meta.messageId, the thread's parent message); parents keep the top-level messages/{id} endpoint.
• DMs (chats).setChatReaction/unsetChatReaction on chats/{chatId}/messages/{id} (chats are flat, no replies).
• Emoji format. Passed through as Unicode — Graph accepts the grapheme directly. reactionCapabilities stays open-unicode / customEmoji: none.
• Stale "deferred to a follow-up" comments removed/updated; the per-actor dispatch they were waiting on now exists.

Inbound was already wired

extractTeamsMessageReactions + the transformChannelThread/transformDmThread transforms already attach reactions to notes during full and incremental sync — this PR only adds the missing outbound direction.

Scopes

No new OAuth scopes. Per the Graph docs, setReaction/unsetReaction require ChannelMessage.Send (channels) and Chat.ReadWrite / ChatMessage.Send (chats) — all already declared. Existing users do not need to re-authorize.

Testing

plot lint, tsc --noEmit, and pnpm build all clean. ms-teams has no existing test harness (no vitest devDep), so no unit test was added to avoid introducing test tooling for a single routing method; the reply-vs-parent path construction follows the Graph endpoint shapes documented above.

Follow-ups (core repo)

• Bump the public/ submodule pointer after merge.
• Add a docs/updates.md bullet (user-facing: Teams reactions now sync back).

🤖 Generated with Claude Code

[KrisBraun]

Superseded by #201. This PR was auto-closed when its head branch feat/ms-teams-reactions was renamed to feat/connector-reactions (now bundling the MS Teams + Google Chat reaction work). Same MS Teams commit (de1a09c) lives on in #201.