PR #66594: fix(gateway): register chat run on agent start and enable provider hooks during model warmup

• Repository: openclaw/openclaw
• Author: Angfr95
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/66594

Description (problem / solution / changelog)

Summary

• Problem: With codex/gpt-5.4, the TUI/webchat remains in pondering for ~39s after the backend has already written the assistant reply. Additionally, gateway startup logs Unknown model: codex/gpt-5.4 during warmup.
• Why it matters: Interactive Codex sessions appear hung or much slower than actual completion time.
• What changed:
  1. chat.send now calls context.addChatRun() inside onAgentRunStart — agent.ts already did this, chat.send did not.
  2. server-startup-post-attach.ts sets skipProviderRuntimeHooks: false so provider-registered models (like codex) resolve correctly at boot.
  3. Test mock in chat.directive-tags.test.ts updated to expose addChatRun: vi.fn() — without it, onAgentRunStart threw a TypeError before reaching registerToolEventRecipient and the audio dispatch loop.

Closes #66470 Supersedes #66337

Root Cause

1. chat.send never called addChatRun() to register the agent run ID. Without this registration, finalizeLifecycleEvent cannot map the agent's internal run ID back to the client-facing idempotency key → missing chat.final events → TUI hangs until timeout (~39s).
2. prewarmConfiguredPrimaryModel passed skipProviderRuntimeHooks: true, skipping hooks that register dynamic model IDs → Unknown model warning at startup.

Changes

Diagram

Before: [chat.send] → [agent writes reply] → [TUI waits for final event] → [timeout ~39s] → [render] After: [chat.send] → [onAgentRunStart: addChatRun()] → [agent writes reply] → [TUI receives final event] → [render immediately]

Security Impact

None. No new permissions, no secrets handling changes, no new network calls, no command execution surface changes.

Tests

49/49 passing in chat.directive-tags.test.ts. 6/6 startup tests passing.

Changed files

• src/gateway/server-methods/chat.directive-tags.test.ts (modified, +2/-0)
• src/gateway/server-methods/chat.ts (modified, +4/-0)
• src/gateway/server-startup-post-attach.ts (modified, +1/-1)
• src/gateway/server-startup.test.ts (modified, +1/-1)

PR #66337: fix(gateway): use provider runtime hooks in startup warmup

• Repository: openclaw/openclaw
• Author: yidianyiko
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/66337

Description (problem / solution / changelog)

Summary

• Problem: gateway startup warmup forced resolveModel(..., { skipProviderRuntimeHooks: true }), so provider-owned dynamic models such as codex/gpt-5.4 were misreported as Unknown model during startup.
• Why it matters: the warning suggests the model is unsupported even when the normal runtime path resolves and uses it successfully.
• What changed: startup warmup now uses the normal resolveModel(...) path, removes the stale static-only error text, and adds regression coverage for provider-runtime-resolved Codex models.
• What did NOT change (scope boundary): this PR does not claim to fix broader model catalog/listing or harness-activation issues tracked elsewhere.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes #
• Related #64938
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: prewarmConfiguredPrimaryModel() in src/gateway/server-startup-post-attach.ts forced the startup warmup through resolveModel(..., { skipProviderRuntimeHooks: true }), which swaps in STATIC_PROVIDER_RUNTIME_HOOKS and disables runProviderDynamicModel. Provider-owned dynamic models like codex/* depend on that hook to resolve.
• Missing detection / guardrail: src/gateway/server-startup.test.ts asserted the static-only warmup behavior instead of locking in parity with the normal runtime resolution path.
• Contributing context (if known): the old error text explicitly documented warmup as “static model resolution” only, which baked the degraded behavior into startup diagnostics.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: src/gateway/server-startup.test.ts
• Scenario the test should lock in: startup warmup should not warn when a configured primary model is resolved via provider runtime hooks, using codex/gpt-5.4 as the regression case.
• Why this is the smallest reliable guardrail: the bug is isolated to the startup warmup call-site and can be reproduced by observing whether it passes the skip flag into resolveModel.
• Existing test that already covers this (if any): none
• If no new test is added, why not: N/A

User-visible / Behavior Changes

• Gateway startup no longer emits a false Unknown model: codex/gpt-5.4 warning when the model is provider-runtime-resolved.

Diagram (if applicable)

Before:
[startup warmup] -> resolveModel(skipProviderRuntimeHooks=true) -> static-only resolution -> false Unknown model warning

After:
[startup warmup] -> resolveModel() -> provider runtime resolution -> no false warning

Security Impact (required)

• New permissions/capabilities? (No)
• Secrets/tokens handling changed? (No)
• New/changed network calls? (No)
• Command/tool execution surface changed? (No)
• Data access scope changed? (No)
• If any Yes, explain risk + mitigation:

Repro + Verification

Environment

• OS: Linux
• Runtime/container: Node 22 / pnpm workspace checkout
• Model/provider: codex/gpt-5.4
• Integration/channel (if any): Gateway startup warmup
• Relevant config (redacted): agents.defaults.model.primary = "codex/gpt-5.4"

Steps

1. Configure agents.defaults.model.primary to codex/gpt-5.4.
2. Start the gateway and let startup warmup run.
3. Observe whether startup emits startup model warmup failed for codex/gpt-5.4: Error: Unknown model: codex/gpt-5.4.

Expected

• Startup warmup should resolve the configured primary model the same way the normal runtime path does and should not emit a false Unknown model warning for provider-runtime-resolved Codex models.

Actual

• Before this patch, startup warmup forced static-only resolution and emitted the false warning.

Evidence

Attach at least one:

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording
• Perf numbers (if relevant)

Human Verification (required)

• Verified scenarios:
  • pnpm test src/gateway/server-startup.test.ts
  • pnpm test src/gateway/server-startup-post-attach.test.ts
  • pnpm test extensions/codex/provider.test.ts -t "resolves arbitrary Codex app-server model ids through the codex provider"
  • pnpm build
• Edge cases checked:
  • explicit openai-codex/gpt-5.4 warmup still prewarms successfully
  • provider-runtime-resolved codex/gpt-5.4 no longer warns in the startup unit test path
• What you did not verify:
  • live gateway startup logs on a real Codex-configured install in this branch
  • repo-wide pnpm check is currently red on unrelated pre-existing tsgo failures outside the touched surface

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? (Yes)
• Config/env changes? (No)
• Migration needed? (No)
• If yes, exact upgrade steps:

Risks and Mitigations

• Risk: startup warmup now allows provider runtime hooks for the configured primary model.
  • Mitigation: the warmup path still uses synchronous resolveModel() (not resolveModelAsync()), so this change re-enables provider runtime model resolution without introducing prepareProviderDynamicModel() async prefetch behavior; regression coverage locks the intended call path.

Changed files

• src/gateway/server-startup-post-attach.ts (modified, +2/-7)
• src/gateway/server-startup.test.ts (modified, +50/-4)