Add realtime consult thinking and fast mode overrides#72237
Add realtime consult thinking and fast mode overrides#72237VACInc wants to merge 1 commit intoopenclaw:mainfrom
Conversation
Greptile SummaryThis PR adds optional Confidence Score: 5/5Safe to merge — changes are additive, backward-compatible, and covered by targeted tests. No logic errors, data loss, or security issues found. Both override paths (Voice Call plugin and Control UI Talk) are independently tested. Schema validation is consistent between Zod, Typebox, and the generated JSON schema. The fastModeOverride/thinkingLevelOverride priority insertion in get-reply-directives.ts is correct and one-shot (not persisted). All edge cases (unset values, false booleans, 'off' think level) are handled correctly. No files require special attention. Reviews (1): Last reviewed commit: "add talk realtime consult overrides" | Re-trigger Greptile |
|
Tested and verified. Running live in my production env |
|
Codex review: needs real behavior proof before merge. Summary Reproducibility: yes. for the missing capability: current-main source shows Voice Call, Talk tool-call, chat.send, and protocol paths do not carry these consult override values. This is source-reproducible feature absence rather than a failing runtime bug. Real behavior proof Next step before merge Security Review findings
Review detailsBest possible solution: Land the additive one-shot override path after moving optional Talk values out of the defaults example, adding a single-line changelog entry, and supplying inspectable live proof. Do we have a high-confidence way to reproduce the issue? Yes for the missing capability: current-main source shows Voice Call, Talk tool-call, chat.send, and protocol paths do not carry these consult override values. This is source-reproducible feature absence rather than a failing runtime bug. Is this the best way to solve the issue? Yes for the feature direction: optional config plus one-shot runtime handoff is narrower than changing persistent session defaults. It is not merge-ready until proof, docs-defaults, and changelog issues are handled. Full review comments:
Overall correctness: patch is correct What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning high; reviewed against 2daf3d332ff0. |
d7ccfc6 to
5842758
Compare
008f33e to
ebc77cd
Compare
ebc77cd to
e33753f
Compare
|
Closing as superseded by #79385, which landed the combined realtime consult run-control surface and follow-up maintainer fixes on |
Summary
realtime.consultThinkingLevelandrealtime.consultFastModeto the Voice Call plugin config and manifest schema.talk.consultThinkingLevelandtalk.consultFastModefor Control UI Talk realtime consults.openclaw_agent_consultcalls when set.Behavior
Unset configs keep existing behavior. Voice Call keeps resolving the consult thinking level from the selected response model default and leaves fast mode unset. Control UI Talk realtime omits the one-shot
chat.sendoverrides unless the new Talk consult config is set.Validation
pnpm test extensions/voice-call/src/config.test.ts extensions/voice-call/src/runtime.test.ts src/realtime-voice/agent-consult-runtime.test.tspnpm test src/config/zod-schema.talk.test.ts src/config/talk.normalize.test.ts src/config/schema.help.quality.test.ts src/gateway/server-methods/talk.test.ts ui/src/ui/chat/realtime-talk.test.ts src/gateway/protocol/index.test.tspnpm check:base-config-schemapnpm config:docs:checkpnpm tsgo:corepnpm tsgo:core:testgit diff --checkNote: a broader
pnpm check:changedrerun hit unrelated e2e failures after the focused lanes above passed; the earlier run also hit a Vitest worker heap limit.