Skip to content

Preserve session context and durable Telegram progress#84828

Closed
ooiuuii wants to merge 1 commit into
openclaw:release/2026.5.19from
ooiuuii:agent/xiaozhua/yoyou-519-telegram-context
Closed

Preserve session context and durable Telegram progress#84828
ooiuuii wants to merge 1 commit into
openclaw:release/2026.5.19from
ooiuuii:agent/xiaozhua/yoyou-519-telegram-context

Conversation

@ooiuuii

@ooiuuii ooiuuii commented May 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Summary

  • preserve larger configured/session context windows over smaller runtime metadata when provider/model match
  • keep Telegram progress mode durable by sending tool progress as standalone messages instead of draft/edit previews
  • update Telegram dispatch tests for durable progress behavior on the 2026.5.19 release base

Test

  • ./node_modules/.bin/vitest run src/agents/session-context-tokens.test.ts src/agents/command/session-store.test.ts extensions/telegram/src/bot-message-dispatch.test.ts --maxWorkers=1

Real behavior proof

  • Behavior or issue addressed: Windows/Telegram OpenClaw 2026.5.19 was shrinking an explicitly configured 1050k session back to runtime-reported 272k after a successful agent run, and Telegram progress mode hid tool/progress updates in a repeatedly edited draft message instead of leaving durable 5.3-style progress messages.

  • Real environment tested: Real Windows/ROG OpenClaw gateway running OpenClaw 2026.5.19 (a185ca2) on port 18889, with the real Telegram channel enabled and reporting ON/OK. The Mac controller verified Windows gateway/control surfaces over the LAN/Tailscale dashboard path.

  • Exact steps or command run after this patch: Applied the equivalent Windows dist hotfix, restarted the Windows OpenClaw Gateway, checked /health, ran a real openclaw agent --agent main ... command against the Windows runtime, inspected C:\Users\Administrator\.openclaw\agents\main\sessions\sessions.json, checked recent runtime logs for schema/API errors, and exercised the real Windows Telegram bot path with a tool-progress turn.

  • Evidence after fix: Copied live output / redacted runtime evidence from the real setup:

    Windows Gateway http://127.0.0.1:18889/health: live
OpenClaw version: 2026.5.19 (a185ca2)
Telegram channel: ON/OK
Agent smoke command: returned OK
agent:main:main.sessionId == agent:main:telegram:default:direct:8589344021.sessionId
shared sessionId: 6f44523c-fc48-49da-925d-bc8bc1e9f224
agent:main:main.contextTokens: 1050000
agent:main:telegram:default:direct:8589344021.contextTokens: 1050000
verbose: full
toolProgressDetail: raw
recent logs: no new provider schema/API errors
Mac cross-access checks after patch: 18898=200, 18913=200, 18912=200
Telegram behavior after patch: tool/progress messages remained visible as standalone durable messages instead of being hidden in a repeatedly edited draft.

  • Observed result after fix: After the agent run, both main and Telegram direct sessions stayed aligned on the same session id with contextTokens=1050000, verbose=full, and toolProgressDetail=raw; the UI/status context no longer reflected the runtime-reported 272k value; Telegram progress/tool output remained durable and visible.

  • What was not tested: Full upstream CI on every package was not run locally; the PR relies on GitHub CI for full-matrix coverage. Production deployment was not tested.

@openclaw-barnacle openclaw-barnacle Bot added channel: telegram Channel integration: telegram agents Agent runtime and tooling size: M triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 21, 2026
@clawsweeper

clawsweeper Bot commented May 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Codex review: found issues before merge.

Workflow note: Future ClawSweeper reviews update this same comment in place.

How this review workflow works
  • ClawSweeper keeps one durable marker-backed review comment per issue or PR.
  • Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
  • A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
  • PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
  • Maintainers can also comment @clawsweeper review to request a fresh review only.
  • Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
  • Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
  • Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Summary
The PR preserves larger session context-token values across agent runs and changes Telegram progress mode from an edited progress draft to standalone durable progress messages, with related tests.

Reproducibility: yes. by source inspection: the PR path spreads deliveryBaseOptions into standalone progress delivery, current deliverReplies mirrors when transcriptMirror is present, and the new context helper takes the max of runtime and fallback context tokens. I did not run tests or live Telegram because this review is read-only.

PR rating
Overall: 🧂 unranked krab
Proof: 🐚 platinum hermit
Patch quality: 🧂 unranked krab
Summary: The proof signal is usable, but the patch is not quality-ready because the current diff has blocking message-delivery and session-state defects.

Rank-up moves:

  • Fix the transcript mirroring, label-only progress, and stale fallback context-token issues with focused regression tests.
  • After code changes, refresh the real Telegram proof with a redacted transcript, recording, live output, or logs that show durable progress behavior and session context preservation.
What the crustacean ranks mean
  • 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
  • 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
  • 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
  • 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
  • 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
  • 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
  • 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

Real behavior proof
Sufficient (live_output): The PR body includes copied live output from a real Windows gateway with Telegram enabled and session context checks; a Telegram Desktop recording would still help review the visible message-order semantics after fixes.

Mantis proof suggestion
A Telegram Desktop proof would show whether progress messages are durable, ordered correctly, absent from transcript context, and compatible with toolProgress=false. A maintainer can ask Mantis to capture proof by posting a new PR comment that starts with the OpenClaw Mantis account mention, followed by:

telegram desktop proof: verify progress-mode Telegram tool updates stay visible, do not enter transcript context, and preserve toolProgress=false behavior.

Risk before merge

  • Merging as-is can mirror standalone Telegram progress/status messages into the durable assistant transcript, changing future model context.
  • Existing Telegram users with streaming.mode="progress" would move from one editable status draft to multiple standalone chat messages without a compatibility path or docs update.
  • Existing streaming.progress.toolProgress=false setups would lose their label-only progress indicator instead of keeping a quiet status draft until final delivery.
  • The context-token resolver can keep an older larger session context window after a provider/model switch because fallback context tokens are included in the max calculation regardless of model match.

Maintainer options:

  1. Fix semantics before merge (recommended)
    Patch standalone progress so non-final status messages never mirror into transcripts, preserve label-only progress mode, constrain context-token preservation to matching runtime model/provider or explicit config, and refresh real Telegram proof after the fixes.
  2. Accept a documented progress policy change
    Maintainers can intentionally make progress messages durable by default, but should update Telegram docs/config hints and provide upgrade proof for existing progress-mode settings before landing.
  3. Split session context from progress transport
    If durable Telegram progress needs more design, land a narrower session-context PR and leave Telegram progress transport unchanged for now.

Next step before merge
Human review is needed because the PR intentionally changes Telegram progress UX and needs author/maintainer revisions rather than an automated cleanup-only repair.

Security
Cleared: No dependency, workflow, permission, credential, or supply-chain changes were introduced; the blockers are functional delivery and session-state correctness.

Review findings

  • [P1] Stop mirroring standalone progress into transcripts — extensions/telegram/src/bot-message-dispatch.ts:1196
  • [P1] Preserve label-only progress mode — extensions/telegram/src/bot-message-dispatch.ts:681
  • [P1] Do not carry stale context windows across model switches — src/agents/session-context-tokens.ts:72
Review details

Best possible solution:

Keep the session-context fix narrow, make durable Telegram progress an explicit documented policy or opt-in mode, and preserve transcript hygiene plus existing progress configuration semantics before merge.

Do we have a high-confidence way to reproduce the issue?

Yes, by source inspection: the PR path spreads deliveryBaseOptions into standalone progress delivery, current deliverReplies mirrors when transcriptMirror is present, and the new context helper takes the max of runtime and fallback context tokens. I did not run tests or live Telegram because this review is read-only.

Is this the best way to solve the issue?

No. The session-context goal is valid, but the current patch is not the safest solution because it changes Telegram progress policy, transcript behavior, and context-window preservation too broadly.

Label changes:

  • add proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes copied live output from a real Windows gateway with Telegram enabled and session context checks; a Telegram Desktop recording would still help review the visible message-order semantics after fixes.
  • add status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Sufficient (live_output): The PR body includes copied live output from a real Windows gateway with Telegram enabled and session context checks; a Telegram Desktop recording would still help review the visible message-order semantics after fixes.
  • remove status: 📣 needs proof: Current PR status label is status: ⏳ waiting on author.

Label justifications:

  • P2: This is a focused bug/improvement PR with real user impact, but the remaining blockers are bounded to Telegram progress delivery and session context metadata.
  • merge-risk: 🚨 compatibility: The PR changes documented progress-mode behavior for existing Telegram configs from an edited draft to standalone visible messages.
  • merge-risk: 🚨 message-delivery: The new standalone progress path can send extra Telegram messages and does not yet preserve the existing label-only progress delivery semantics.
  • merge-risk: 🚨 session-state: The new context-token resolver can preserve stale larger context windows across model/provider changes, affecting compaction and session status.
  • rating: 🧂 unranked krab: Current PR rating is 🧂 unranked krab because proof is 🐚 platinum hermit, patch quality is 🧂 unranked krab, and The proof signal is usable, but the patch is not quality-ready because the current diff has blocking message-delivery and session-state defects.
  • status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Sufficient (live_output): The PR body includes copied live output from a real Windows gateway with Telegram enabled and session context checks; a Telegram Desktop recording would still help review the visible message-order semantics after fixes.
  • proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes copied live output from a real Windows gateway with Telegram enabled and session context checks; a Telegram Desktop recording would still help review the visible message-order semantics after fixes.
  • mantis: telegram-visible-proof: Mantis should capture Telegram visible proof. The PR changes visible Telegram progress-message behavior, which is well suited to a short Telegram Desktop proof recording.

Full review comments:

  • [P1] Stop mirroring standalone progress into transcripts — extensions/telegram/src/bot-message-dispatch.ts:1196
    deliveryBaseOptions includes transcriptMirror, and deliverReplies mirrors every delivered payload when that callback is present. Sending standalone tool/progress updates through this helper will append non-final status text to the assistant transcript, so future turns can see tool labels or progress chatter as model context; clear transcriptMirror for non-final progress or use a non-transcript delivery path.
    Confidence: 0.96
  • [P1] Preserve label-only progress mode — extensions/telegram/src/bot-message-dispatch.ts:681
    With streaming.progress.toolProgress: false, streamToolProgressVisible is false and this branch returns before displaying anything. Current docs/config hints and tests keep a label-only progress draft for that setting, so existing users would lose their visible progress indicator instead of only hiding tool lines.
    Confidence: 0.93
  • [P1] Do not carry stale context windows across model switches — src/agents/session-context-tokens.ts:72
    The helper always includes fallbackContextTokens in the Math.max candidate set, and callers pass the active session entry's context window as that fallback. After switching from a larger old model to a smaller runtime-reported model, the old larger value still wins even when provider/model no longer match, which can overstate context and compaction state.
    Confidence: 0.91

Overall correctness: patch is incorrect
Overall confidence: 0.92

Acceptance criteria:

  • node scripts/run-vitest.mjs src/agents/session-context-tokens.test.ts src/agents/command/session-store.test.ts src/auto-reply/reply/session.test.ts extensions/telegram/src/bot-message-dispatch.test.ts
  • Real Telegram proof for progress-mode delivery after fixes, with private IDs/tokens/endpoints redacted

What I checked:

  • PR patch: The fetched PR patch for head dd6b47b adds standalone Telegram progress delivery and a new resolvePreservedSessionContextTokens helper across session persistence paths. (dd6b47b97e68)
  • Maintainer Telegram decision: The Telegram maintainer note says streaming owns one visible preview message and Telegram transport/streaming PRs need real Telegram proof; that makes this progress-mode behavior change upgrade-sensitive rather than a routine refactor. (.agents/maintainer-notes/telegram.md:7, 8284c035a096)
  • Current documented behavior: Current Telegram docs say progress mode keeps one editable status draft for tool progress, clears it at completion, and sends the final answer as a normal message. Public docs: docs/channels/telegram.md. (docs/channels/telegram.md:314, 8284c035a096)
  • Transcript mirror path: Current dispatch options carry transcriptMirror in deliveryBaseOptions, and deliverReplies mirrors delivered content whenever that callback is supplied; the PR's standalone progress helper spreads those options without clearing transcriptMirror. (extensions/telegram/src/bot-message-dispatch.ts:1015, 8284c035a096)
  • Existing label-only progress coverage: Current tests and config hints cover progress.toolProgress=false as label-only progress draft behavior; the PR deletes those expectations and suppresses output instead. (extensions/telegram/src/bot-message-dispatch.test.ts:1859, 8284c035a096)
  • Session context provenance: Current context-token selection in agent-runner was introduced in the provider-qualified session context limits fix, so fallback context values are part of the relevant session-state history. (src/auto-reply/reply/agent-runner.ts:1599, 2645ed154b42)

Likely related people:

  • steipete: git blame and release-base history tie the current Telegram streaming/progress code, docs, and tests to commits c8a953a and a185ca2. (role: recent area contributor; confidence: high; commits: c8a953af9371, a185ca283a74; files: extensions/telegram/src/bot-message-dispatch.ts, extensions/telegram/src/bot-message-dispatch.test.ts, docs/channels/telegram.md)
  • neeravmakwana: The provider-qualified session context limit work in 2645ed1 touched the same agent-runner fallback context-token path affected by this PR. (role: session context limits contributor; confidence: medium; commits: 2645ed154b42; files: src/auto-reply/reply/agent-runner.ts, src/auto-reply/reply/followup-runner.ts)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 8284c035a096.

@clawsweeper clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask. mantis: telegram-visible-proof Mantis should capture Telegram visible proof. P2 Normal backlog priority with limited blast radius. merge-risk: 🚨 compatibility 🚨 May break existing users, config, migrations, defaults, or upgrade paths. merge-risk: 🚨 message-delivery 🚨 May drop, duplicate, misroute, suppress, or wrongly target messages. merge-risk: 🚨 session-state 🚨 May lose, corrupt, stale, or mis-associate session, agent, or context state. labels May 21, 2026
@clawsweeper

clawsweeper Bot commented May 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

ClawSweeper PR egg

🔥 Warming up: real-behavior proof passed; findings, security review, or rank-up moves are still in progress.

Hatch command

Comment @clawsweeper hatch when this PR is hatchable.

Hatchability rules:

  • Merged PRs are hatchable.
  • Open PRs are hatchable when they are status: 👀 ready for maintainer look, status: 🚀 automerge armed, or labeled clawsweeper:automerge.
  • Closed unmerged PRs are hatchable only when one of those hatchable labels is still present in the durable record.
What is this egg doing here?
  • Eggs appear after the PR passes real-behavior proof. It is here for vibes, not verdicts: it does not change labels, ratings, merge decisions, or automation.
  • The shell reacts to review momentum: open follow-up work warms it up, re-review makes it wobble, and a clean final review lets it hatch.
  • Hatchability usually comes from sufficient real-behavior proof, no blocking P0/P1/P2 findings, no security attention needed, and clean correctness. A merged PR is already final, so merge makes the egg hatchable independently.
  • The hatch is seeded from this repository and PR number, so the same PR keeps the same creature; the reviewed head SHA can only change safe visual details.
  • Rarity is just collectible sparkle: 🥚 common, 🌱 uncommon, 💎 rare, ✨ glimmer, and 🌈 legendary.

@ooiuuii ooiuuii force-pushed the agent/xiaozhua/yoyou-519-telegram-context branch from d4b2793 to dd6b47b Compare May 21, 2026 06:02
@openclaw-barnacle openclaw-barnacle Bot added proof: supplied External PR includes structured after-fix real behavior proof. and removed triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 21, 2026
@clawsweeper clawsweeper Bot added proof: sufficient ClawSweeper judged the real behavior proof convincing. status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action. and removed status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask. labels May 21, 2026
@ooiuuii

ooiuuii commented May 21, 2026

Copy link
Copy Markdown
Contributor Author

Closing this for now: this branch is intended as a personal fork customization for durable Telegram progress behavior rather than a proposed upstream default. Thanks.

@ooiuuii ooiuuii closed this May 21, 2026
@ooiuuii ooiuuii mentioned this pull request Jun 4, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

agents Agent runtime and tooling channel: telegram Channel integration: telegram mantis: telegram-visible-proof Mantis should capture Telegram visible proof. merge-risk: 🚨 compatibility 🚨 May break existing users, config, migrations, defaults, or upgrade paths. merge-risk: 🚨 message-delivery 🚨 May drop, duplicate, misroute, suppress, or wrongly target messages. merge-risk: 🚨 session-state 🚨 May lose, corrupt, stale, or mis-associate session, agent, or context state. P2 Normal backlog priority with limited blast radius. proof: sufficient ClawSweeper judged the real behavior proof convincing. proof: supplied External PR includes structured after-fix real behavior proof. rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. size: M status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@ooiuuii