fix(codex): preserve semantic native threads across compaction by 100yenadmin · Pull Request #86160 · openclaw/openclaw

100yenadmin · 2026-05-24T18:17:03Z

Summary

This is PR #4 in the Codex native-thread stack, after #85978, #86069, and #86094. It turns the diagnostics foundation into the long-term cache/lifecycle fix for compatible context-engine thread_bootstrap bindings:

preserve Codex thread_bootstrap bindings across successful context-engine-owned compaction;
copy preserved bindings to successor session files during transcript rollover, then clear the archived/original binding;
reread the active Codex binding after forced compaction before rebuilding the Codex prompt;
if optimistic semantic reuse rotates during startup, rebuild the prompt for the fresh thread before turn/start so the fresh thread receives projected context;
keep legacy, ownerless, non-bootstrap, mismatch, and app-server rejection paths rotating/clearing as before;
hash sensitive MCP/environment comparison tokens before persisting or emitting lifecycle diagnostics;
add context-engine-compaction-preserved-binding lifecycle diagnostics and docs explaining the preserved vs invalidated paths.

Why this is not “raise 70k to 258k/272k”

The default 70000 native reuse guard remains a legacy/stale-binding safety rail. The healthy Codex + LCM path is semantic ownership:

Context engine returns contextProjection.mode = "thread_bootstrap" with a stable epoch/fingerprint.
OpenClaw injects large assembled context once into the Codex native thread.
Later turns resume that native thread and send only the current prompt when the engine/policy/projection/tool identity still matches.
Compaction or rollover preserves the binding only when that semantic identity remains safe.
Actual app-server overflow/rejection still clears the binding before fresh-thread retry.

The important distinction is model capacity vs warm-thread lifecycle policy:

Value	Meaning	Share of 70k guard
`70000`	OpenClaw proactive native-thread reuse guard for legacy/stale bindings	100%
`272000`	OpenClaw Codex runtime input cap used for GPT-5.5 forward-compat rows	70k is 25.7%
`400000`	GPT-5.5 native/context window in current OpenClaw catalog/test fixtures	70k is 17.5%

With a 30% post-compaction target, even a compacted transcript can land above the guard:

30% of 258k = 77.4k
30% of 272k = 81.6k

So repeatedly applying a hard 70k warm-thread rotation rule to otherwise compatible Codex native threads makes long-running sessions take the cold path right after compaction. This PR keeps the 70k guard for legacy/non-semantic bindings, but lets semantically-owned thread_bootstrap bindings survive compaction/rollover.

Local bootstrap size scan

I ran a bounded local scan over named bootstrap candidates (AGENTS.md, USER.md, MEMORY.md, SOUL.md, IDENTITY.md, TOOLS.md, BOOTSTRAP.md) under /Volumes/LEXAR/repos, /Volumes/LEXAR/Codex, ~/.openclaw, and ~/.codex, using ceil(bytes / 4) as a conservative token estimate.

Metric	Result
Candidate files	3,235
Aggregate directories	2,032
Aggregate p50	~637 tokens
Aggregate p90	~3.8k tokens
Aggregate p99	~13.5k tokens
Max aggregate	~117.7k tokens (`~/.codex/memories/MEMORY.md`)
Largest repeated AGENTS candidate	~23.6k tokens

Interpretation: the bug is not “every AGENTS.md is over 70k.” The failure mode is total rendered pressure: compacted history + context-engine projection + workspace bootstrap/developer instructions + native rollout growth. Once that total has created a warm native thread, the efficient Codex path is to preserve the semantically valid native thread rather than reproject the large stable payload every turn.

Flow

flowchart TD
  A["Long-running Codex session"] --> B{"Saved native binding?"}
  B -- "Legacy / ownerless / non-bootstrap" --> C{"session/native tokens >= guard?"}
  C -- "yes" --> D["Rotate cold: thread/start + reproject context"]
  C -- "no" --> E["Try normal resume"]
  B -- "context-engine thread_bootstrap" --> F{"Engine id + policy + epoch/fingerprint + tools match?"}
  F -- "no" --> D
  F -- "yes" --> G["Resume warm native thread"]
  G --> H["Send current prompt only"]
  H --> I{"Context-engine compaction?"}
  I -- "same file, identity stable" --> J["Preserve binding in place"]
  I -- "successor rollover, identity stable" --> K["Copy binding to successor; clear archived original"]
  I -- "identity changes or app-server rejects" --> D
  J --> G
  K --> G

Real behavior proof

Behavior or issue addressed: Compatible Codex context-engine thread_bootstrap native bindings are preserved across successful context-engine compaction and successor session-file rollover, while legacy/non-bootstrap, mismatch, and app-server rejection paths still rotate/clear.
Real environment tested: Local OpenClaw worktree on the Lexar-backed checkout at /Volumes/LEXAR/repos/worktrees/openclaw-codex-semantic-reuse-guard, branch codex/semantic-native-thread-preservation, current head 8b405b4d5c59432b4f3218b570609d3d43f5a4ad. The proof used temp session files under /Volumes/LEXAR/Codex/openclaw-pr-86160-proof-8b405b4d5c and the actual exported reconcileContextEngineCompactedCodexBinding(...), writeCodexAppServerBinding(...), readCodexAppServerBinding(...), clearCodexAppServerBinding(...), and startOrResumeThread(...) helpers from this checkout.
Exact steps or command run after this patch: Ran a one-off pnpm exec tsx proof probe, not Vitest/tsgo/build. The probe wrote matching thread_bootstrap bindings, simulated a compactor clearing the sidecar before same-file compaction, simulated successor rollover copy/archived clear, and seeded a legacy raw user-MCP binding to prove the mismatch path rotates through thread/start and persists a hashed comparison token. I also verified the patched worktree identity and whitespace state with pwd, git rev-parse HEAD, git branch --show-current, git status --porcelain, and git diff --check HEAD^ HEAD.
Evidence after fix: Copied terminal output from the current patched worktree proof probe:

[agent/embedded] context-engine-owned Codex app-server compaction preserved native thread-bootstrap binding
[agent/embedded] context-engine-owned Codex app-server compaction preserved native thread-bootstrap binding
{
  "head": "8b405b4d5c59432b4f3218b570609d3d43f5a4ad",
  "proofDir": "/Volumes/LEXAR/Codex/openclaw-pr-86160-proof-8b405b4d5c",
  "sameFileThreadBootstrap": {
    "outcome": {
      "invalidated": false,
      "preserved": true,
      "successorInvalidated": false
    },
    "restoredAfterCompactorClearedSidecar": true,
    "preservedThreadId": "thread-same-file-warm",
    "projectionMode": "thread_bootstrap",
    "projectionEpoch": "epoch-proof",
    "projectionFingerprint": "projection-fp-proof"
  },
  "successorRolloverThreadBootstrap": {
    "outcome": {
      "invalidated": false,
      "preserved": true,
      "successorInvalidated": false
    },
    "archivedBindingPresent": false,
    "successorThreadId": "thread-rollover-warm",
    "successorSessionFileBasename": "rollover-successor.jsonl",
    "projectionMode": "thread_bootstrap",
    "projectionEpoch": "epoch-rollover",
    "projectionFingerprint": "projection-fp-rollover"
  },
  "mcpConfigMismatchRotation": {
    "lifecycleAction": "started",
    "methods": [
      "thread/start"
    ],
    "previousThreadRotated": true,
    "savedThreadId": "thread-mcp-new",
    "userMcpServersFingerprintPrefix": "sha256:",
    "leakedMcpSecretInBinding": false,
    "leakedMcpPathInBinding": false
  }
}
$ pwd
/Volumes/LEXAR/repos/worktrees/openclaw-codex-semantic-reuse-guard
$ git rev-parse HEAD
8b405b4d5c59432b4f3218b570609d3d43f5a4ad
$ git branch --show-current
codex/semantic-native-thread-preservation
$ git status --porcelain
$ git diff --check HEAD^ HEAD

git status --porcelain and git diff --check HEAD^ HEAD produced no output.

Observed result after fix: Same-file compaction returned preserved: true and restored thread-same-file-warm after the sidecar was cleared. Successor rollover returned preserved: true, cleared the archived/original binding (archivedBindingPresent: false), and copied thread-rollover-warm with the same thread_bootstrap epoch/fingerprint to the successor session file. The MCP mismatch path did not resume the stale thread; it called thread/start, saved thread-mcp-new, persisted a sha256: comparison fingerprint, and did not leak the seeded MCP email/path into the saved binding.
What was not tested: No local Vitest/tsgo suite, package build, or live Gateway/Discord/Codex app-server replay was run locally, to avoid the local resource issue called out in AGENTS.md. GitHub Actions remains the validation gate for type, package-boundary, and test execution.
Before evidence: The motivating production-style logs reported repeated codex app-server native transcript exceeded active token limit; starting a fresh thread at native token counts above the 70k reuse guard; issue Codex long-running sessions should use semantic thread/bootstrap cache ownership #86023 and this PR body include the budget math and local bootstrap-size scan explaining that failure mode.

Validation

Local validation intentionally skipped heavy suites per AGENTS.md / user instruction. GitHub Actions is the validation gate.

Local:

one-off pnpm exec tsx lifecycle proof probe, as captured above
git diff --check HEAD^ HEAD

GitHub Actions:

current head 8b405b4d5c59432b4f3218b570609d3d43f5a4ad is green, including the previously failing logging support-export slice after the stability-bundle mode parser fix and the rerun after the unrelated model-catalog timeout.

Not run locally:

Vitest focused slices
pnpm tsgo:extensions:test
live Gateway/Discord/Codex app-server replay

Stack / Review Notes

This PR is stacked on #86094. The diff to review is the top commit:

8b405b4d5c fix(codex): preserve semantic native threads across compaction

Refs #86023.

clawsweeper · 2026-05-24T18:19:29Z

Codex review: needs real behavior proof before merge. Reviewed June 24, 2026, 12:24 PM ET / 16:24 UTC.

Summary
The PR changes Codex app-server lifecycle/config/diagnostics to preserve compatible context-engine thread_bootstrap bindings across native compaction and session rollover, adds token-guard config/docs, and expands tests.

PR surface: Source +1743, Tests +2123, Docs +81, Generated 0. Total +3947 across 30 files.

Reproducibility: yes. at source level: current main clears the stored thread_bootstrap projection before native compaction, and upstream Codex compact keeps operating on the same thread without returning replacement binding identity.

Review metrics: 3 noteworthy metrics.

PR surface: 30 files, +4119/-172. The branch spans Codex runtime, diagnostics, config, docs, logging, and tests, so maintainers need a clean merge result before landing.
Config/default surfaces: 1 added. agents.defaults.compaction.maxActiveTranscriptTokens changes operator-visible Codex native-thread reuse policy and needs upgrade/default proof.
Trusted diagnostic surfaces: 1 event family added. codex.native_thread.lifecycle carries thread/session/runtime identity fields, so redaction and support-export behavior matter before merge.

Stored data model
Persistent data-model change detected: serialized state: extensions/codex/src/app-server/native-thread-diagnostics.test.ts. Confirm migration or upgrade compatibility proof before merge.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #86023
Summary: This PR is a candidate implementation slice for the canonical Codex long-running native-thread/cache ownership tracker, with diagnostics and liveness issues still related but not replacing it.

Members:

canonical: Codex long-running sessions should use semantic thread/bootstrap cache ownership #86023 - Canonical open tracker for semantic Codex thread/bootstrap cache ownership and lifecycle diagnostics.
same_root_cause: feat(codex): diagnose native thread lifecycle #86094 - Open diagnostics foundation in the same Codex native-thread lifecycle stack.
partial_overlap: fix(codex): make native thread token guard configurable #86069 - Merged configurable token-guard slice that explicitly left semantic compaction preservation out of scope.
partial_overlap: [Bug] Orphaned/oversized native Codex thread wedges a session permanently — chat.send returns started but no run executes, silently dropping messages #86963 - Open Codex native-thread/session-state liveness report with message-loss symptoms adjacent to this lifecycle work.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🦪 silver shellfish
Proof: 🦪 silver shellfish
Patch quality: 🦐 gold shrimp
Result: blocked until stronger real behavior proof is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

Rebase or restack onto current main and refresh the merge result.
Update the PR body with redacted current-head proof for preservation, rollover, mismatch rotation, prompt rebuild, and diagnostic redaction.
Run focused Codex app-server compaction/run-attempt tests and the maintainer-selected extension type/test lane after rebase.

Proof guidance:

[P1] Needs stronger real behavior proof before merge: Terminal proof is relevant but stale: it demonstrates old head 8b405b4, while the live PR head is c99a46c with later behavior commits. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The live PR branch is conflicting/dirty, so maintainers need a refreshed merge result before judging final code and CI.
[P1] The current proof covers old head 8b405b4, not live head c99a46c.
[P1] Preserving and copying native Codex bindings across compaction and rollover can mis-associate future turns if the equivalence key is incomplete.
[P1] The PR adds trusted diagnostics and config/default surface around sensitive runtime identity, so redaction and upgrade/default behavior need current-head proof.

Maintainer options:

Refresh branch and proof before merge (recommended)
Resolve the dirty merge state and update the PR body with redacted current-head terminal or log proof for same-file preservation, successor copy/clear, mismatch rotation, prompt rebuild, and hashed diagnostics.
Land or replace diagnostics foundation first
Maintainers can require the lifecycle diagnostic surface to land or be replaced before accepting semantic preservation so the policy is observable and reviewable.
Open a narrower current-main replacement
If rebasing discards most of this stale stack, close it and open a focused current-main PR for compaction and rollover preservation only.

Next step before merge

[P1] Maintainers need a refreshed merge result, current-head proof, and a Codex lifecycle/config policy decision; this is not a safe automated repair candidate.

Security
Cleared: No concrete supply-chain, workflow, dependency, or secret-handling defect was found; security-sensitive diagnostics remain a merge-risk proof requirement.

Review details

Best possible solution:

Rebase or replace the branch on current main after the diagnostics stack is settled, refresh current-head redacted proof, and land only after maintainers accept the Codex lifecycle policy and privacy boundaries.

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

Yes at source level: current main clears the stored thread_bootstrap projection before native compaction, and upstream Codex compact keeps operating on the same thread without returning replacement binding identity.

Is this the best way to solve the issue?

Unclear as a landing vehicle: semantic preservation matches the existing binding contract, but this conflicting stacked branch needs current-main rebase, current-head proof, and maintainer policy acceptance before it can be called the best fix.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 3217165be770.

Label changes

Label justifications:

P2: The PR targets a real Codex long-session correctness/performance issue but is bounded to Codex lifecycle behavior with maintainer-review blockers.
merge-risk: 🚨 compatibility: The diff includes config/default and native-thread reuse behavior changes that can affect existing long-running Codex sessions on upgrade.
merge-risk: 🚨 session-state: The PR preserves and migrates native thread bindings across compaction and rollover, which can mis-associate future turns if semantic identity is incomplete.
merge-risk: 🚨 security-boundary: The PR adds trusted diagnostics and fingerprints around auth, MCP, environment, and thread/session identity, making redaction boundaries merge-critical.
rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🦐 gold shrimp.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs stronger real behavior proof before merge: Terminal proof is relevant but stale: it demonstrates old head 8b405b4, while the live PR head is c99a46c with later behavior commits. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +1743, Tests +2123, Docs +81, Generated 0. Total +3947 across 30 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	13	1866	123	+1743
Tests	13	2170	47	+2123
Docs	3	81	0	+81
Config	0	0	0	0
Generated	1	2	2	0
Other	0	0	0	0
Total	30	4119	172	+3947

Acceptance criteria:

[P1] After rebase, run focused Codex app-server compact/run-attempt tests through node scripts/run-vitest.mjs.
[P1] Run the relevant Codex extension type/test lane or maintainer-selected Crabbox/Testbox proof before landing.
[P1] Refresh redacted current-head terminal/log proof for same-file preservation, successor copy/clear, mismatch rotation, prompt rebuild, and diagnostic redaction.

What I checked:

Repository policy applied: Root AGENTS.md requires full OpenClaw PR review, scoped guides, compatibility-risk handling, and direct sibling Codex source inspection before a Codex verdict. (AGENTS.md:1, e9720c27fa69)
Live PR state blocks merge readiness: GitHub reports the PR open at head c99a46c with mergeable=false and mergeable_state=dirty. (c99a46c76230)
Current main still clears projection before native compaction: Current main calls clearContextEngineProjectionBeforeNativeCompaction before thread/compact/start, so the semantic-preservation behavior is not already implemented. (extensions/codex/src/app-server/compact.ts:249, e9720c27fa69)
Current main still clears oversized startup bindings: Current main clears the Codex app-server binding when native/session tokens exceed the computed startup fuse. (extensions/codex/src/app-server/startup-binding.ts:398, e9720c27fa69)
Codex compact API does not return replacement binding identity: Upstream Codex thread/compact/start accepts only a thread_id and returns an empty response, so OpenClaw must manage its sidecar binding lifecycle itself. (../codex/codex-rs/app-server-protocol/src/protocol/v2/thread.rs:924, 390b73133b07)
Codex compact operation acts on the same loaded thread: Upstream Codex loads the requested thread and submits Op::Compact, then returns an empty ThreadCompactStartResponse. (../codex/codex-rs/app-server/src/request_processors/thread_processor.rs:1755, 390b73133b07)

Likely related people:

vincentkoc: Current-main blame for the central Codex compaction/startup-binding files and many live PR branch commits point to Vincent Koc as recent adjacent area maintainer. (role: recent area contributor; confidence: high; commits: e9720c27fa69, c2e52db52b5e, c99a46c76230; files: extensions/codex/src/app-server/compact.ts, extensions/codex/src/app-server/startup-binding.ts, extensions/codex/src/app-server/run-attempt.ts)
joshavant: Authored merged context-engine compaction ownership work that now owns adjacent projection-clearing behavior and tests. (role: recent area contributor; confidence: high; commits: a8d33f23a09d; files: extensions/codex/src/app-server/compact.ts, extensions/codex/src/app-server/compact.test.ts)
fuller-stack-dev: Prior review and commit history tie this handle to the merged native-thread overflow/startup-binding mitigation that this PR extends. (role: adjacent owner; confidence: medium; commits: 81505ada186f; files: extensions/codex/src/app-server/startup-binding.ts, extensions/codex/src/app-server/run-attempt.ts)
steipete: Authored neighboring stuck Codex compaction recovery work and appears in the merged configurable-guard slice history. (role: adjacent recovery contributor; confidence: medium; commits: 44d5330993ce, 0177987431ff; files: extensions/codex/src/app-server/compact.ts, extensions/codex/src/app-server/run-attempt.ts, extensions/codex/src/app-server/startup-binding.ts)
100yenadmin: Authored the merged configurable native-token guard PR and owns the open diagnostics and semantic-preservation PRs in this stack, beyond only opening this PR. (role: follow-up implementer; confidence: medium; commits: 7a147419dbbd, f623fc0a7dd1, c99a46c76230; files: extensions/codex/src/app-server/run-attempt.ts, extensions/codex/src/app-server/startup-binding.ts, extensions/codex/src/app-server/compact.ts)

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.

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.

clawsweeper · 2026-05-24T18:27:41Z

ClawSweeper PR egg

🎁 Pass real behavior proof to wake the egg and unlock a hatchable treat.

Where did the egg go?

The egg game starts only after the PR passes the real-behavior proof check.
Before that, no creature or rarity is rolled. The treat waits for real proof.
This is still just collectible flavor: proof affects review readiness, not creature quality.

100yenadmin · 2026-05-24T18:33:44Z

@clawsweeper please re-review this PR against the latest head 996ebfe6c4 and the now-passing Real behavior proof check.

100yenadmin · 2026-05-24T18:38:34Z

/review

clawsweeper · 2026-05-24T18:38:36Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

State: Superseded
Detail: A newer re-review for this item started before this run finished, so GitHub cancelled this older run. Check the latest ClawSweeper run for the current result.
Run: https://github.com/openclaw/clawsweeper/actions/runs/26369555394
Updated: 2026-05-24T18:42:03.032Z

Add Vincent Koc as a co-author for the PR context and review trail. Co-authored-by: Vincent Koc <[email protected]>

openclaw-barnacle · 2026-06-21T05:25:34Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

openclaw-barnacle · 2026-06-22T05:05:11Z

This assigned pull request has been automatically marked as stale after being open for 27 days.
Please add updates or it will be closed.

100yenadmin · 2026-07-07T04:16:36Z

closing to make PR space

100yenadmin mentioned this pull request May 24, 2026

Codex long-running sessions should use semantic thread/bootstrap cache ownership #86023

Open

100yenadmin force-pushed the codex/semantic-native-thread-preservation branch from 66906a1 to 996ebfe Compare May 24, 2026 18:22

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 24, 2026

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. labels May 24, 2026

100yenadmin force-pushed the codex/semantic-native-thread-preservation branch from 996ebfe to 2ead386 Compare May 24, 2026 19:02

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 24, 2026

clawsweeper Bot added status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask. and removed status: ⏳ waiting on author ClawSweeper has contributor-facing work open and is waiting for author action. labels May 24, 2026

100yenadmin force-pushed the codex/semantic-native-thread-preservation branch from 2ead386 to e496a20 Compare May 24, 2026 19:20

vincentkoc and others added 15 commits May 27, 2026 13:05

fix(codex): preserve successor thread bootstrap reuse

14f7a7e

fix(codex): mirror compacted successor turns

3436ab8

fix(codex): register compacted active session

81cdf8f

fix(codex): return compacted session file

8b94837

fix(codex): preserve rotated active run controls

e4c7a47

docs(config): refresh native token guard baseline

3ede929

fix(logging): redact Codex thread ids from stability

8810fc9

fix(codex): surface pre-start compaction metadata

f623fc0

fix(codex): preserve semantic native threads across compaction

8d96276

chore: credit vincentkoc on PR context

5fc45a0

Add Vincent Koc as a co-author for the PR context and review trail. Co-authored-by: Vincent Koc <[email protected]>

fix(codex): preserve compacted successor binding

611f208

fix(codex): defer post-start context compaction

aca942f

fix(codex): verify successor compaction bindings

4354e89

fix(codex): clear unsent bootstrap bindings

3e0d151

fix(codex): clean up early bootstrap returns

c99a46c

vincentkoc force-pushed the codex/semantic-native-thread-preservation branch from 1055c44 to c99a46c Compare May 27, 2026 11:09

clawsweeper Bot mentioned this pull request May 27, 2026

feat(codex): diagnose native thread lifecycle #86094

Closed

This comment was marked as spam.

Sign in to view

clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. and removed rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. labels Jun 14, 2026

eldar702 mentioned this pull request Jun 16, 2026

fix(codex): route OpenClaw runtime context via turn-scoped developer instructions [AI-assisted] #93762

Closed

clawsweeper Bot added rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. and removed rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. labels Jun 19, 2026

clawsweeper Bot mentioned this pull request Jun 20, 2026

[Bug] Orphaned/oversized native Codex thread wedges a session permanently — chat.send returns started but no run executes, silently dropping messages #86963

Open

openclaw-barnacle Bot added the stale Marked as stale due to inactivity label Jun 21, 2026

clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. and removed rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. labels Jun 21, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(codex): preserve semantic native threads across compaction#86160

fix(codex): preserve semantic native threads across compaction#86160
100yenadmin wants to merge 21 commits into
openclaw:mainfrom
100yenadmin:codex/semantic-native-thread-preservation

100yenadmin commented May 24, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented May 24, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented May 24, 2026 •

edited

Loading

Uh oh!

100yenadmin commented May 24, 2026

Uh oh!

100yenadmin commented May 24, 2026

Uh oh!

clawsweeper Bot commented May 24, 2026 •

edited

Loading

Uh oh!

This comment was marked as spam.

openclaw-barnacle Bot commented Jun 21, 2026

Uh oh!

openclaw-barnacle Bot commented Jun 22, 2026

Uh oh!

100yenadmin commented Jul 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

100yenadmin commented May 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Why this is not “raise 70k to 258k/272k”

Local bootstrap size scan

Flow

Real behavior proof

Validation

Stack / Review Notes

Uh oh!

clawsweeper Bot commented May 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented May 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

100yenadmin commented May 24, 2026

Uh oh!

100yenadmin commented May 24, 2026

Uh oh!

clawsweeper Bot commented May 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as spam.

openclaw-barnacle Bot commented Jun 21, 2026

Uh oh!

openclaw-barnacle Bot commented Jun 22, 2026

Uh oh!

100yenadmin commented Jul 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

100yenadmin commented May 24, 2026 •

edited

Loading

clawsweeper Bot commented May 24, 2026 •

edited

Loading

clawsweeper Bot commented May 24, 2026 •

edited

Loading

clawsweeper Bot commented May 24, 2026 •

edited

Loading