PR #68144: fix(discord): set ctx.To to user: target for DM inbound so downstream builds direct session keys

• Repository: openclaw/openclaw
• Author: hclsys
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/68144

Description (problem / solution / changelog)

Summary

Fixes #68126 — Discord DM inbound messages were landing with `ctx.To = channel:<dmChannelId>` instead of `user:<userId>`. Downstream mirror and delivery-recovery then built `agent:main:discord:channel:<id>` session keys for what are semantically direct conversations, and Discord's API returned `Unknown Channel` errors on recovery reads.

Same fix family as the closed-without-merge attempt in #47626 (brokemac79, 2026-04-06); the bug was re-reported today with a fresh stack trace.

Root cause

In `extensions/discord/src/monitor/message-handler.process.ts`, `effectiveTo` was computed before `dmConversationTarget` was resolved:

```ts const effectiveTo = autoThreadContext?.To ?? replyTarget; // ← replyTarget is always channel:<dmChannelId> // ... later ... const dmConversationTarget = isDirectMessage ? resolveDiscordConversationIdentity({ isDirectMessage, userId: author.id }) : undefined; const lastRouteTo = dmConversationTarget ?? effectiveTo; // ← correctly user:... for DMs ```

`OriginatingTo` and `lastRouteTo` are already patched via `dmConversationTarget`, so the DM correction logic was already in the file — `ctx.To` itself was just missing it.

Fix

Reorder so `dmConversationTarget` is computed before `effectiveTo`, then thread it into the fallback chain:

```ts const effectiveTo = autoThreadContext?.To ?? dmConversationTarget ?? replyTarget; ```

The outgoing `deliverTarget` is unchanged because Discord's API sends DMs via channel id — only the semantic `ctx.To` that downstream consumers read is corrected. `lastRouteTo` simplifies naturally under the new order (`dmConversationTarget ?? effectiveTo` is equivalent when effectiveTo is already `dmConversationTarget` on DMs).

Test plan

• Added a regression case in `extensions/discord/src/monitor/message-handler.process.test.ts`:
  • `sets ctx.To to user target for DM inbound so downstream builds direct session keys`
• Widened the existing `getLastDispatchCtx` helper to expose `To`, `From`, and `ChatType` so downstream tests can assert ctx shape.
• The existing DM lastRoute test (`stores DM lastRoute with user target for direct-session continuity`) still passes — the `lastRouteTo` derivation is unchanged in outcome.
• `NODE_OPTIONS=--max-old-space-size=8192 npx tsc --noEmit` — 247 baseline on main, 247 on branch (no new errors).
• `pnpm exec oxlint` on both touched files — 0 warnings, 0 errors.

Local full vitest is blocked by pre-existing `test/non-isolated-runner.ts` drift (reproduces on `main`) — CI runs the new case normally.

Risk

• `deliverTarget` is untouched, so the outbound Discord API call is identical to today's behavior.
• `OriginatingTo` already uses `dmConversationTarget ?? replyTarget` (line 464), so every downstream ctx field now agrees on the DM-user semantic.
• Guild/channel paths (`isDirectMessage === false`) are unaffected because `dmConversationTarget` is `undefined` and the fallback chain collapses to the prior `replyTarget`.

Changed files

• extensions/discord/src/monitor/message-handler.process.test.ts (modified, +36/-2)
• extensions/discord/src/monitor/message-handler.process.ts (modified, +9/-5)