fix(slack): reduce token bloat by skipping thread context on existing sessions by sourman · Pull Request #32133 · openclaw/openclaw

sourman · 2026-03-02T20:13:07Z

Summary

Thread history and thread starter were being fetched and included on every message in a Slack thread, causing unnecessary token bloat. The session transcript already contains the full conversation history, so re-fetching and re-injecting thread history on each turn is redundant.

What Changed

src/slack/monitor/message-handler/prepare.ts:551 - Only fetch thread history for NEW sessions (!threadSessionPreviousTimestamp)
src/slack/monitor/message-handler/prepare.ts:641 - Only include ThreadStarterBody for NEW sessions
Updated test to reflect new behavior

Before vs After

Before: Every message in a Slack thread included [Thread history - for context] with the entire thread conversation, growing larger each turn.

After: Thread history context wrapper is not needed since the agent has context form response. No further context is included

Session Evidence

Message 2 ("checking in"):

[Thread history - for context]
[Slack Ahmed Mansour] <@U0AF3PBGARJ> ping
[Slack comply] Pong. What do you need?
...

Message 3 ("what is the fifth letter in four"):

[Thread history - for context]
[... entire conversation repeated again ...]
[Slack Ahmed Mansour] checking in
[Slack comply] All clear here...

With this fix, messages do NOT include the redundant thread history wrapper.

Impact

Cost: Fewer API calls to Slack for thread history (only on first message)
Tokens: Significant reduction for multi-turn thread conversations
Latency: Reduced per-message processing time

Test Results

All 345 Slack tests pass
Updated test skips loading thread history when thread session already exists in store (bloat fix) to verify new behavior

Fixes #32121

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 012d84ec17

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-03-02T21:08:30Z

src/slack/monitor/message-handler/prepare.ts

    });
-    if (threadInitialHistoryLimit > 0) {
+    // Only fetch thread history for NEW sessions (existing sessions already have this context in their transcript)
+    if (threadInitialHistoryLimit > 0 && !threadSessionPreviousTimestamp) {


Check thread session freshness before skipping context preload

This condition treats any existing store entry as an active thread session, but session reuse is actually gated by freshness in initSessionState (src/auto-reply/reply/session.ts:411-424) and stale entries are reset (default mode is daily in src/config/sessions/reset.ts:20-23). In the common case where a thread has an old updatedAt (or idle-expired entry), prepareSlackMessage now skips loading thread history/starter context even though a brand new agent session will be created, so the first message after reset loses prior thread context and can produce incorrect replies.

Useful? React with 👍 / 👎.

greptile-apps · 2026-03-02T21:12:31Z

Greptile Summary

Prevents redundant thread context from being fetched and included on every message in a Slack thread, significantly reducing token bloat and API calls for multi-turn conversations.

Key Changes:

Only fetches thread history for NEW thread sessions (!threadSessionPreviousTimestamp) - existing sessions already have complete conversation history in their transcript
Only includes ThreadStarterBody for NEW sessions - subsequent turns don't need this re-injected
Reduces API calls from 2 per message (starter + history) to 1 per message (starter only for metadata/media) for existing sessions

Implementation:

prepare.ts:556 - Added && !threadSessionPreviousTimestamp condition to thread history fetch
prepare.ts:647 - Made ThreadStarterBody conditional based on session existence
Thread starter is still fetched for ALL messages (needed for thread labels and media fallback), but its body is only included in context for first turn
Test updated to validate that existing sessions skip both history fetch and body inclusion

Impact:

Significant token savings for multi-turn thread conversations
Reduced Slack API calls (only on first message per thread)
Faster per-message processing for existing thread sessions
All 345 Slack tests pass

Confidence Score: 5/5

This PR is safe to merge with minimal risk - changes are focused, well-tested, and backward compatible
Score reflects: (1) straightforward implementation with only two conditional checks added, (2) comprehensive test coverage for both new and existing session scenarios, (3) all 345 Slack tests passing, (4) well-motivated optimization addressing real token bloat issue, (5) backward compatible with no breaking changes, (6) clear code comments explaining rationale
No files require special attention - changes are minimal and focused

_{Last reviewed commit: 012d84e}

… sessions Thread history and thread starter were being fetched and included on every message in a Slack thread, causing unnecessary token bloat. The session transcript already contains the full conversation history, so re-fetching and re-injecting thread history on each turn is redundant. Now thread history is only fetched for new thread sessions (!threadSessionPreviousTimestamp). Existing sessions rely on their transcript for context. Fixes openclaw#32121

@sourman

…anks @sourman)

steipete · 2026-03-02T21:29:45Z

Landed via temp rebase onto main.

Gate: pnpm -s vitest run src/slack/monitor/message-handler/prepare.test.ts src/slack/monitor/allow-list.test.ts
Land commit: fc8f18e0233cdaf823c9523c12e50be4e34f34be
Merge commit: 2438fde

Thanks @sourman!

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fc8f18e02b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-02T21:32:08Z

src/slack/monitor/message-handler/prepare.ts

    });
-    if (threadInitialHistoryLimit > 0) {
+    // Only fetch thread history for NEW sessions (existing sessions already have this context in their transcript)
+    if (threadInitialHistoryLimit > 0 && !threadSessionPreviousTimestamp) {


Check session freshness before skipping thread context

The new gate threadInitialHistoryLimit > 0 && !threadSessionPreviousTimestamp treats any stored updatedAt as an active thread session, but session reuse is actually decided by freshness in initSessionState via evaluateSessionFreshness (src/auto-reply/reply/session.ts). Fresh evidence is that a stale entry (daily/idle reset) still has updatedAt, so this branch skips loading thread history (and the later starter-body assignment also skips starter text), even though the run will create a brand-new session; that makes the first post-reset thread message lose prior context and can produce incorrect replies.

Useful? React with 👍 / 👎.

…eload Previously, the fix only checked if a session timestamp existed, but didn't account for session freshness. A stale session (e.g., from yesterday before the daily reset hour) has a timestamp, so we skipped loading thread history, but initSessionState creates a brand new session because it's stale - causing the first message after reset to lose context. Now we check session freshness using the same logic as initSessionState: - Truly new session (no timestamp) → load thread history - Stale session (exists but expired) → load thread history (will be reset) - Fresh session (exists and active) → skip thread history (use transcript) Fixes comment on PR openclaw#32133

@sourman

…anks @sourman)

@sourman

…anks @sourman)

…eload Previously, the fix (openclaw#32133) only checked if a session timestamp existed, but didn't account for session freshness. A stale session (e.g., from yesterday before the daily reset hour) has a timestamp, so we skipped loading thread history, but initSessionState creates a brand new session because it's stale - causing the first message after reset to lose context. Now we check session freshness using the same logic as initSessionState: - Truly new session (no timestamp) → load thread history - Stale session (exists but expired) → load thread history (will be reset) - Fresh session (exists and active) → skip thread history (use transcript) Also fixes the wildcard bug in resolveChannelResetConfig so resetByChannel["*"] works as a catch-all for channel-specific session reset policies. Fixes comment on PR openclaw#32133

@sourman

…anks @sourman)

…eload Previously, the fix (openclaw#32133) only checked if a session timestamp existed, but didn't account for session freshness. A stale session (e.g., from yesterday before the daily reset hour) has a timestamp, so we skipped loading thread history, but initSessionState creates a brand new session because it's stale - causing the first message after reset to lose context. Now we check session freshness using the same logic as initSessionState: - Truly new session (no timestamp) → load thread history - Stale session (exists but expired) → load thread history (will be reset) - Fresh session (exists and active) → skip thread history (use transcript) Also fixes the wildcard bug in resolveChannelResetConfig so resetByChannel["*"] works as a catch-all for channel-specific session reset policies. Fixes comment on PR openclaw#32133

Previously, the fix (openclaw#32133) only checked if a session timestamp existed, but didn't account for session freshness. A stale session (e.g., from yesterday before the daily reset hour) has a timestamp, so we skipped loading thread history, but initSessionState creates a brand new session because it's stale - causing the first message after reset to lose context. Now we check session freshness using the same logic as initSessionState: - Truly new session (no timestamp) → load thread history - Stale session (exists but expired) → load thread history (will be reset) - Fresh session (exists and active) → skip thread history (use transcript) Also fixes the wildcard bug in resolveChannelResetConfig so resetByChannel["*"] works as a catch-all for channel-specific session reset policies. Fixes comment on PR openclaw#32133

@sourman

…anks @sourman)

openclaw-barnacle bot added channel: slack Channel integration: slack size: XS labels Mar 2, 2026

sourman force-pushed the fix/slack-thread-context-bloat branch from 48420ad to 012d84e Compare March 2, 2026 21:02

sourman marked this pull request as ready for review March 2, 2026 21:05

chatgpt-codex-connector bot reviewed Mar 2, 2026

View reviewed changes

sourman and others added 2 commits March 2, 2026 21:28

fix: trim repeated slack thread context payloads (openclaw#32133) (th…

fc8f18e

…anks @sourman)

steipete force-pushed the fix/slack-thread-context-bloat branch from 012d84e to fc8f18e Compare March 2, 2026 21:29

steipete merged commit 2438fde into openclaw:main Mar 2, 2026

chatgpt-codex-connector bot reviewed Mar 2, 2026

View reviewed changes

This was referenced Mar 2, 2026

fix(slack): skip thread history fetch for existing sessions (#32121) #32132

Closed

fix(slack): exclude thread starter from thread history to avoid token duplication #32140

Closed

planfit-alan pushed a commit to planfit/openclaw that referenced this pull request Mar 3, 2026

fix: trim repeated slack thread context payloads (openclaw#32133) (th…

f6bbe85

…anks @sourman)

sourman mentioned this pull request Mar 3, 2026

[Slack]: Drop unnecessary per-message metadata blocks in Slack threads #33042

Open

dawi369 pushed a commit to dawi369/davis that referenced this pull request Mar 3, 2026

fix: trim repeated slack thread context payloads (openclaw#32133) (th…

4fdec28

…anks @sourman)

This was referenced Mar 3, 2026

[slack] Check session freshness before skipping thread context** #33507

Open

fix(slack): thread sessions respect timeout (history + implicit mention) #33519

Draft

OWALabuy pushed a commit to kcinzgg/openclaw that referenced this pull request Mar 4, 2026

fix: trim repeated slack thread context payloads (openclaw#32133) (th…

7408ee6

…anks @sourman)

zooqueen pushed a commit to hanzoai/bot that referenced this pull request Mar 6, 2026

fix: trim repeated slack thread context payloads (openclaw#32133) (th…

885a63e

…anks @sourman)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(slack): reduce token bloat by skipping thread context on existing sessions#32133

fix(slack): reduce token bloat by skipping thread context on existing sessions#32133
steipete merged 2 commits intoopenclaw:mainfrom
sourman:fix/slack-thread-context-bloat

sourman commented Mar 2, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector bot left a comment

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Uh oh!

greptile-apps bot commented Mar 2, 2026

Uh oh!

steipete commented Mar 2, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

sourman commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What Changed

Before vs After

Session Evidence

Impact

Test Results

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot commented Mar 2, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

steipete commented Mar 2, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 2, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

sourman commented Mar 2, 2026 •

edited

Loading