mirror of
https://github.com/bilawalsidhu/gods-eye-view.git
synced 2026-11-05 19:04:02 +00:00
feat(camera): add globe rotation (#1225)
Add continuous globe rotation with direction and speed controls, viewport fit checks, camera ownership handoff, and voice refusal guards.
This commit is contained in:
1 parent
8cdd184a90
commit
b2fb4a261e
25 files changed
+3407
-36
No files matched your search
@@ -729,7 +729,7 @@ const schemas = [
|
||||
properties: {
|
||||
motion: {
|
||||
type: 'string',
|
||||
enum: ['orbit', 'pan', 'tilt', 'rotate', 'stop'],
|
||||
enum: ['orbit', 'pan', 'tilt', 'rotate', 'rotate_globe', 'stop'],
|
||||
},
|
||||
direction: {
|
||||
type: 'string',
|
||||
|
||||
@@ -1,3 +1,4 @@
|
||||
import { realtimeInstructions } from '../../server/providers/openai/instructions.js';
|
||||
import test from 'node:test';
|
||||
import assert from 'node:assert/strict';
|
||||
import { createHash } from 'node:crypto';
|
||||
@@ -38,7 +39,10 @@ test('the complete Realtime tool payload pins the manifest-generated layer relea
|
||||
// requested radius and center authority in descriptions only.
|
||||
// Street Level and GNSS extend the manifest; tracked and analyst aircraft describe broadcasts.
|
||||
// Air quality (ECCC AQHI) extends the manifest.
|
||||
'775ba323384b4c1b62fb8415bac52aecc3fc668e20bedcfa57e426dcd1da3969',
|
||||
// Globe rotation adds one motion; visual keyhole fit and explicit
|
||||
// reset guidance/defaults update only move_camera and mode descriptions.
|
||||
// Restore base descriptions; append only globe-rotation defaults and fit guidance.
|
||||
'967b5a043b812e2f8c1e190fc571cf12b960c9d4486df736902221aab4d99cf6',
|
||||
);
|
||||
});
|
||||
|
||||
@@ -106,6 +110,17 @@ test('all legacy action arguments are byte-identical after removing the delibera
|
||||
delete property('show_data_layers_menu').layerId.enum;
|
||||
delete property('get_entity_context').layerId.enum;
|
||||
delete property('analyst_query').layers.items.enum;
|
||||
// Earth-centered rotation is additive; local camera yaw keeps its contract.
|
||||
const motion = property('move_camera').motion;
|
||||
assert.deepEqual(motion.enum, [
|
||||
'orbit',
|
||||
'pan',
|
||||
'tilt',
|
||||
'rotate',
|
||||
'rotate_globe',
|
||||
'stop',
|
||||
]);
|
||||
motion.enum = motion.enum.filter((value) => value !== 'rotate_globe');
|
||||
const panels = property('set_panel_open').panelId;
|
||||
panels.enum = panels.enum.filter(
|
||||
(id) => !['weather-panel', 'recent-imagery-panel'].includes(id),
|
||||
@@ -145,6 +160,24 @@ test('all legacy action arguments are byte-identical after removing the delibera
|
||||
);
|
||||
});
|
||||
|
||||
test('globe rotation is an additive camera motion with the existing control arguments', () => {
|
||||
const schema = GEV_ACTION_SCHEMAS.find(
|
||||
(tool) => tool.name === 'move_camera',
|
||||
).parameters;
|
||||
for (const motion of ['rotate_globe', 'rotate', 'orbit', 'pan', 'tilt']) {
|
||||
assert.deepEqual(
|
||||
validateArgs(schema, {
|
||||
motion,
|
||||
direction: 'right',
|
||||
speed: 'slow',
|
||||
mode: 'continuous',
|
||||
}),
|
||||
[],
|
||||
);
|
||||
}
|
||||
assert.ok(validateArgs(schema, { motion: 'spin_camera' }).length > 0);
|
||||
});
|
||||
|
||||
test('track_entity accepts a referent alone but rejects an empty target', () => {
|
||||
const schema = GEV_ACTION_SCHEMAS.find(
|
||||
(tool) => tool.name === 'track_entity',
|
||||
@@ -183,3 +216,15 @@ test('the benchmark still rejects an empty track_entity although the model-facin
|
||||
[],
|
||||
);
|
||||
});
|
||||
|
||||
test('globe reset guidance requires a separate explicit command after refusal', () => {
|
||||
const paragraph = realtimeInstructions()
|
||||
.split('\n')
|
||||
.find((line) => line.includes('globe-view-required'));
|
||||
assert.ok(paragraph);
|
||||
assert.match(paragraph, /calls only move_camera motion rotate_globe/);
|
||||
assert.match(paragraph, /explicit user Reset\/Show Globe View/);
|
||||
assert.match(paragraph, /never from your own plan/);
|
||||
assert.match(paragraph, /bare yes\/ok/);
|
||||
assert.match(paragraph, /Do not ask a question/);
|
||||
});
|
||||
@@ -21,6 +21,7 @@ import {
|
||||
shouldSendViewportImage,
|
||||
shouldStopVoiceAfterRadioTool,
|
||||
responseInstructionForToolResult,
|
||||
isGlobeViewRefusal,
|
||||
} from './realtimeProtocol.js';
|
||||
|
||||
const SOCKET_PATH =
|
||||
@@ -182,6 +183,8 @@ export class GeminiSessionController {
|
||||
this.model = null;
|
||||
this.epoch = 0;
|
||||
this.turn = 0;
|
||||
this.globeToolsBlocked = false;
|
||||
this.globeResetFailed = false;
|
||||
this.inputGeneration = 0;
|
||||
this.turnAbort = new AbortController();
|
||||
this.disposed = false;
|
||||
@@ -659,6 +662,8 @@ export class GeminiSessionController {
|
||||
return;
|
||||
if (!pushToTalk || activityCaptureId === metadata.captureId) return;
|
||||
this.captureOwner.admitted = true;
|
||||
this.globeToolsBlocked = false;
|
||||
this.globeResetFailed = false;
|
||||
if (
|
||||
!this.interrupt('push-to-talk', {
|
||||
keepPlayback: !this.pttPlaybackSuppressed,
|
||||
@@ -1294,6 +1299,8 @@ export class GeminiSessionController {
|
||||
this.replyPending = false;
|
||||
if (!this.interrupt()) return;
|
||||
} else if (!this.retiredReplyPending && !this.interrupt()) return;
|
||||
this.globeToolsBlocked = false;
|
||||
this.globeResetFailed = false;
|
||||
// Local activity ownership is only for expiry liveness, not tools.
|
||||
// Assign after retiring observed output so its boundary cannot clear
|
||||
// a newer detected activity that has not received a reply yet.
|
||||
@@ -1573,6 +1580,7 @@ export class GeminiSessionController {
|
||||
.then(async () => {
|
||||
if (!this.owns(epoch) || turn !== this.turn) return;
|
||||
if (this.cancelledCalls.has(call.id)) {
|
||||
if (call.name === 'zoom_to_globe') this.globeResetFailed = true;
|
||||
if (this.tools.get(call.id) === action) this.tools.delete(call.id);
|
||||
this.pendingTools--;
|
||||
if (!this.pendingTools) {
|
||||
@@ -1618,8 +1626,28 @@ export class GeminiSessionController {
|
||||
(authority !== 'playback' ||
|
||||
handoffEpoch === this.radio.radioHandoffEpoch);
|
||||
let result;
|
||||
let skippedGlobeRotation = false;
|
||||
let blockedGlobeTool = false;
|
||||
try {
|
||||
if (!actionCurrent()) throw cancelled();
|
||||
// Live has no response-scoped tool_choice. Recheck here so queued siblings
|
||||
// cannot reset after a rotation refusal.
|
||||
if (this.globeToolsBlocked) {
|
||||
blockedGlobeTool = true;
|
||||
throw new Error(
|
||||
'Wait for fresh user input before calling another tool.',
|
||||
);
|
||||
}
|
||||
if (
|
||||
this.globeResetFailed &&
|
||||
call.name === 'move_camera' &&
|
||||
String(args.motion || '').toLowerCase() === 'rotate_globe'
|
||||
) {
|
||||
skippedGlobeRotation = true;
|
||||
throw new Error(
|
||||
'Not run: the Globe View reset/rotation sequence failed or was cancelled.',
|
||||
);
|
||||
}
|
||||
if (stronger)
|
||||
reservation = this.radio.reserveRadioToolHandoff({ abortScope });
|
||||
handoffEpoch = this.radio.radioHandoffEpoch;
|
||||
@@ -1689,7 +1717,12 @@ export class GeminiSessionController {
|
||||
} catch (error) {
|
||||
result = {
|
||||
ok: false,
|
||||
tool: call.name,
|
||||
...(skippedGlobeRotation
|
||||
? { action: call.name, reason: 'globe-reset-failed', skipped: true }
|
||||
: {
|
||||
tool: call.name,
|
||||
...(blockedGlobeTool ? { skipped: true } : {}),
|
||||
}),
|
||||
...(radioVisibility ? { layerId: 'radio' } : {}),
|
||||
error:
|
||||
error?.name === 'AbortError'
|
||||
@@ -1716,6 +1749,11 @@ export class GeminiSessionController {
|
||||
this.noteRadioStrongerCommit(action, abortScope, turn);
|
||||
}
|
||||
if (!this.owns(epoch) || turn !== this.turn) return;
|
||||
if (
|
||||
call.name === 'zoom_to_globe' &&
|
||||
(result?.ok !== true || result.cancelled)
|
||||
)
|
||||
this.globeResetFailed = true;
|
||||
const playbackObsolete =
|
||||
authority === 'playback' && handoffEpoch !== this.radio.radioHandoffEpoch;
|
||||
if (radioControl || radioVisibility) {
|
||||
@@ -1744,6 +1782,8 @@ export class GeminiSessionController {
|
||||
this.stopAfterReply = true;
|
||||
// Native call cancellation suppresses its output, without canceling siblings.
|
||||
if (!this.cancelledCalls.has(call.id)) {
|
||||
if (isGlobeViewRefusal(result) && !skippedGlobeRotation)
|
||||
this.globeToolsBlocked = true;
|
||||
this.postToolOutput = false;
|
||||
this.modelComplete = false;
|
||||
this.replyPending = true;
|
||||
@@ -1751,7 +1791,9 @@ export class GeminiSessionController {
|
||||
const mapEvents = mapSnapshot.map((text) => JSON.parse(text));
|
||||
const response = {
|
||||
...(result ?? { ok: false, error: 'Command returned no result' }),
|
||||
responseGuidance: responseInstructionForToolResult(result),
|
||||
responseGuidance: blockedGlobeTool
|
||||
? 'That action did not run. Wait for fresh user input; do not claim any change or call another tool.'
|
||||
: responseInstructionForToolResult(result),
|
||||
};
|
||||
this.send(
|
||||
{
|
||||
@@ -1902,6 +1944,8 @@ export class GeminiSessionController {
|
||||
{
|
||||
onSent: () => {
|
||||
if (!owned() || this.typedIntent !== typedIntent) return;
|
||||
this.globeToolsBlocked = false;
|
||||
this.globeResetFailed = false;
|
||||
this.replyPending = true;
|
||||
this.bindTextIntent(typedIntent);
|
||||
if (!owned() || this.typedIntent !== typedIntent) return;
|
||||
@@ -2455,6 +2499,8 @@ export class GeminiSessionController {
|
||||
this.cancelledCalls.clear();
|
||||
this.pendingMapEvents = [];
|
||||
this.userTurnPending = false;
|
||||
this.globeToolsBlocked = false;
|
||||
this.globeResetFailed = false;
|
||||
this.replyPending = false;
|
||||
this.unrequestedReply = false;
|
||||
this.automaticTurnOpen = false;
|
||||
|
||||
@@ -3398,3 +3398,215 @@ test('a narration callback that submits a successor keeps the old packet tools f
|
||||
assert.deepEqual(h.calls, []);
|
||||
h.c.stop();
|
||||
});
|
||||
|
||||
for (const reply of ['typed', 'push-to-talk', 'open-mic']) {
|
||||
test(`globe refusal blocks queued and later tools until fresh ${reply} input`, async () => {
|
||||
const names = [];
|
||||
const h = harness({
|
||||
runAction: async (name) => {
|
||||
names.push(name);
|
||||
return name === 'move_camera'
|
||||
? { ok: false, action: 'move_camera', reason: 'globe-view-required' }
|
||||
: { ok: true, action: name };
|
||||
},
|
||||
});
|
||||
try {
|
||||
await h.c.start({ pushToTalk: reply === 'push-to-talk' });
|
||||
establishToolInput(h, 'rotate the globe');
|
||||
const ws = h.sockets[0];
|
||||
ws.message({
|
||||
toolCall: {
|
||||
functionCalls: [
|
||||
{
|
||||
id: 'rotate',
|
||||
name: 'move_camera',
|
||||
args: { motion: 'rotate_globe' },
|
||||
},
|
||||
{ id: 'queued-reset', name: 'zoom_to_globe', args: {} },
|
||||
],
|
||||
},
|
||||
});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(names, ['move_camera']);
|
||||
const blocked = ws.sent
|
||||
.flatMap((entry) => entry.toolResponse?.functionResponses || [])
|
||||
.find((call) => call.id === 'queued-reset');
|
||||
assert.equal(blocked.response.ok, false);
|
||||
assert.match(blocked.response.error, /fresh user input/i);
|
||||
assert.equal(blocked.response.skipped, true);
|
||||
assert.match(blocked.response.responseGuidance, /did not run/i);
|
||||
assert.doesNotMatch(
|
||||
blocked.response.responseGuidance,
|
||||
/confirm.*completed/i,
|
||||
);
|
||||
requestTool(ws, 'repeated-rotation', 'move_camera', {
|
||||
motion: 'rotate_globe',
|
||||
});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(names, ['move_camera']);
|
||||
const repeated = toolResponses(ws).find(
|
||||
(call) => call.id === 'repeated-rotation',
|
||||
).response;
|
||||
assert.equal(repeated.ok, false);
|
||||
assert.equal(repeated.skipped, true);
|
||||
assert.match(repeated.responseGuidance, /did not run/i);
|
||||
assert.doesNotMatch(
|
||||
repeated.responseGuidance,
|
||||
/confirm.*completed/i,
|
||||
);
|
||||
|
||||
if (reply === 'typed') {
|
||||
const stimuli = [
|
||||
() => {},
|
||||
() => ws.message({ serverContent: { interrupted: true } }),
|
||||
() =>
|
||||
ws.message({
|
||||
serverContent: {
|
||||
inputTranscription: { text: 'reset to globe view' },
|
||||
},
|
||||
}),
|
||||
() => {
|
||||
h.c.input.setMicrophoneEnabled(false);
|
||||
h.c.input.setMicrophoneEnabled(true);
|
||||
},
|
||||
];
|
||||
for (const [i, stimulus] of stimuli.entries()) {
|
||||
stimulus();
|
||||
requestTool(ws, `unrequested-reset-${i}`, 'zoom_to_globe', {});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(names, ['move_camera']);
|
||||
}
|
||||
}
|
||||
// Finish the refusal reply before admitting the actual next user input.
|
||||
ws.message({ serverContent: { turnComplete: true } });
|
||||
if (reply === 'typed')
|
||||
assert.equal(h.c.sendTextCommand('reset to globe view'), true);
|
||||
else if (reply === 'push-to-talk') {
|
||||
h.c.input.pushToTalkKeyHeld = true;
|
||||
h.c.input.spaceKeyHeld = true;
|
||||
h.c.input.setMicrophoneEnabled(true);
|
||||
h.c.input.setMicrophoneEnabled(false);
|
||||
} else ws.message({ voiceActivity: { type: 'ACTIVITY_START' } });
|
||||
requestTool(ws, 'accepted-reset', 'zoom_to_globe', {});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(names, ['move_camera', 'zoom_to_globe']);
|
||||
} finally {
|
||||
h.c.stop();
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
test('a cancelled globe reset skips queued rotation, preserves Stop and recovers on fresh input', async () => {
|
||||
const calls = [];
|
||||
let resets = 0;
|
||||
const h = harness({
|
||||
runAction: async (name, args) => {
|
||||
calls.push(name === 'move_camera' ? args.motion : name);
|
||||
return name === 'zoom_to_globe' && resets++ === 0
|
||||
? { ok: false, action: name, cancelled: true }
|
||||
: { ok: true, action: name };
|
||||
},
|
||||
});
|
||||
try {
|
||||
await h.c.start();
|
||||
establishToolInput(h, 'reset to globe view and rotate');
|
||||
const ws = h.sockets[0];
|
||||
ws.message({
|
||||
toolCall: {
|
||||
functionCalls: [
|
||||
{ id: 'reset-failed', name: 'zoom_to_globe', args: {} },
|
||||
{
|
||||
id: 'rotate-skipped',
|
||||
name: 'move_camera',
|
||||
args: { motion: 'rotate_globe' },
|
||||
},
|
||||
{ id: 'stop-allowed', name: 'move_camera', args: { motion: 'stop' } },
|
||||
],
|
||||
},
|
||||
});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(calls, ['zoom_to_globe', 'stop']);
|
||||
const skipped = toolResponses(ws).find(
|
||||
(call) => call.id === 'rotate-skipped',
|
||||
).response;
|
||||
assert.equal(skipped.ok, false);
|
||||
assert.equal(skipped.reason, 'globe-reset-failed');
|
||||
assert.equal(skipped.skipped, true);
|
||||
assert.equal(
|
||||
toolResponses(ws).find((call) => call.id === 'stop-allowed').response.ok,
|
||||
true,
|
||||
);
|
||||
|
||||
ws.message({ serverContent: { turnComplete: true } });
|
||||
establishToolInput(h, 'reset to globe view and rotate');
|
||||
ws.message({
|
||||
toolCall: {
|
||||
functionCalls: [
|
||||
{ id: 'reset-success', name: 'zoom_to_globe', args: {} },
|
||||
{
|
||||
id: 'rotate-success',
|
||||
name: 'move_camera',
|
||||
args: { motion: 'rotate_globe' },
|
||||
},
|
||||
],
|
||||
},
|
||||
});
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(calls, [
|
||||
'zoom_to_globe',
|
||||
'stop',
|
||||
'zoom_to_globe',
|
||||
'rotate_globe',
|
||||
]);
|
||||
} finally {
|
||||
h.c.stop();
|
||||
}
|
||||
});
|
||||
|
||||
test('cancelling a queued globe reset also skips its dependent rotation', async () => {
|
||||
const blocker = deferred();
|
||||
const calls = [];
|
||||
const h = harness({
|
||||
runAction: async (name, args) => {
|
||||
calls.push(name === 'move_camera' ? args.motion : name);
|
||||
return args.motion === 'pan'
|
||||
? blocker.promise
|
||||
: { ok: true, action: name };
|
||||
},
|
||||
});
|
||||
try {
|
||||
await h.c.start();
|
||||
establishToolInput(h, 'pan, reset to globe view and rotate');
|
||||
const ws = h.sockets[0];
|
||||
ws.message({
|
||||
toolCall: {
|
||||
functionCalls: [
|
||||
{
|
||||
id: 'blocker',
|
||||
name: 'move_camera',
|
||||
args: { motion: 'pan', direction: 'left' },
|
||||
},
|
||||
{ id: 'reset-cancelled', name: 'zoom_to_globe', args: {} },
|
||||
{
|
||||
id: 'rotate-after-cancel',
|
||||
name: 'move_camera',
|
||||
args: { motion: 'rotate_globe' },
|
||||
},
|
||||
],
|
||||
},
|
||||
});
|
||||
await tick();
|
||||
ws.message({ toolCallCancellation: { ids: ['reset-cancelled'] } });
|
||||
blocker.resolve({ ok: true, action: 'move_camera' });
|
||||
await h.c.toolQueue;
|
||||
assert.deepEqual(calls, ['pan']);
|
||||
const skipped = toolResponses(ws).find(
|
||||
(call) => call.id === 'rotate-after-cancel',
|
||||
).response;
|
||||
assert.equal(skipped.reason, 'globe-reset-failed');
|
||||
assert.equal(skipped.skipped, true);
|
||||
} finally {
|
||||
blocker.resolve({ ok: true });
|
||||
h.c.stop();
|
||||
}
|
||||
});
|
||||
@@ -1,3 +1,4 @@
|
||||
import { interruptGlobeRotation } from '../cameraVerbs.js';
|
||||
import {
|
||||
layerSnapshot,
|
||||
feedProvenanceEnvelope,
|
||||
@@ -262,7 +263,9 @@ export function createGevActionRunner({
|
||||
}),
|
||||
));
|
||||
installViewTargetPrewarm(viewer);
|
||||
initCameraVerbs(viewer, getViewTargetCartesian);
|
||||
initCameraVerbs(viewer, getViewTargetCartesian, {
|
||||
isCockpitActive: () => Boolean(styleManager?.getCockpitState?.()?.active),
|
||||
});
|
||||
// Nested calls (select_nearest_aircraft composing layer + flight) use the
|
||||
// raw runner; only the outer call gets a voice envelope.
|
||||
const runGevAction = async function runGevAction(
|
||||
@@ -298,6 +301,7 @@ export function createGevActionRunner({
|
||||
// Zoom during an active orbit adjusts the orbit RADIUS (spiral in/out) —
|
||||
// a straight camera move would be snapped back by the per-frame lookAt.
|
||||
if (name === 'adjust_camera_zoom') {
|
||||
interruptGlobeRotation('voice-zoom-intent');
|
||||
const zoomOut = String(args.direction || '').toLowerCase() === 'out';
|
||||
const amt = String(args.amount || 'medium').toLowerCase();
|
||||
const factor = { little: 1.25, medium: 1.6, lot: 2.4 }[amt] || 1.6;
|
||||
|
||||
File diff suppressed because it is too large.
Load diff
@@ -99,6 +99,18 @@ export function hasStructuredViewIdentity(result) {
|
||||
);
|
||||
}
|
||||
|
||||
export function isGlobeViewRefusal(result) {
|
||||
return (
|
||||
result?.action === 'move_camera' &&
|
||||
result.ok === false &&
|
||||
[
|
||||
'globe-view-required',
|
||||
'globe-reset-unavailable',
|
||||
'globe-reset-failed',
|
||||
].includes(result.reason)
|
||||
);
|
||||
}
|
||||
|
||||
export function responseInstructionForToolResult(result) {
|
||||
if (result?.action === 'control_radio' && result.radioPlaybackSuppressed) {
|
||||
if (result.audioState === 'paused') {
|
||||
@@ -123,6 +135,20 @@ export function responseInstructionForToolResult(result) {
|
||||
: 'Be concise: useful names, place, layer and one or two notable properties.',
|
||||
].join(' ');
|
||||
}
|
||||
if (isGlobeViewRefusal(result)) {
|
||||
// A refusal is not a completed action. Keep tool text out of instructions.
|
||||
if (result.reason === 'globe-reset-failed')
|
||||
return 'The globe did not rotate after the Globe View reset attempt. Say so briefly. Do not retry, offer or perform another reset for this request.';
|
||||
if (result.reason === 'globe-reset-unavailable')
|
||||
return 'The globe did not rotate. Ask the user to zoom out further; the Globe View preset does not fit this viewport. Do not offer or perform a reset.';
|
||||
return 'The globe did not rotate. Say "Globe rotation isn’t available from this view. You can say ‘Reset to Globe View and rotate the globe.’" The suggested command is not a user request. Do not ask a question, perform a reset, or call tools in this reply.';
|
||||
}
|
||||
if (
|
||||
(result?.action ?? result?.tool) === 'zoom_to_globe' &&
|
||||
(result.ok !== true || result.cancelled)
|
||||
) {
|
||||
return 'The Globe View reset did not finish. Say so briefly; do not claim Globe View was reached. Do not retry the reset or rotate the globe for this request.';
|
||||
}
|
||||
if (result?.action === 'adjust_camera_zoom') {
|
||||
return result.ok
|
||||
? `Confirm once that the camera zoomed ${result.direction}. Do not claim any other change.`
|
||||
|
||||
+229
-10
@@ -6,6 +6,7 @@ import {
|
||||
SUPERSEDED_RESPONSE_MEMORY,
|
||||
shouldStopVoiceAfterRadioTool,
|
||||
responseInstructionForToolResult,
|
||||
isGlobeViewRefusal,
|
||||
extractFunctionCalls,
|
||||
callDedupeKeys,
|
||||
parseArguments,
|
||||
@@ -66,6 +67,11 @@ export class RealtimeTurns {
|
||||
this.pendingUserTextResponse = false;
|
||||
this.activeResponseId = null;
|
||||
this.supersededResponseIds = new Set();
|
||||
this.globeRefusals = new Map();
|
||||
this.globeAdmissionGates = new Map();
|
||||
this.globeResets = new Map();
|
||||
this.globeResponseRequests = new Map();
|
||||
this.globeRequest = { failed: false, pending: new Set() };
|
||||
this.activeToolAbortControllers = new Set();
|
||||
this.narration = null;
|
||||
this.narrationSequence = 0;
|
||||
@@ -138,6 +144,7 @@ export class RealtimeTurns {
|
||||
},
|
||||
};
|
||||
if (!this.sendRealtimeEvent(itemEvent, 'client.user_text')) return;
|
||||
this.globeRequest = { failed: false, pending: new Set() };
|
||||
this.requestUserTextResponse();
|
||||
}
|
||||
|
||||
@@ -330,6 +337,7 @@ export class RealtimeTurns {
|
||||
this.userTurnPending = true;
|
||||
this.pendingResponseInstructions = null;
|
||||
this.cancelRadioHandoff({ abortTools: true });
|
||||
this.globeRequest = { failed: false, pending: new Set() };
|
||||
this.setVoiceSpeaker('user');
|
||||
// Point-and-ask: read the pointer as speech starts (open mic), or reuse
|
||||
// the push-to-talk keydown snapshot, before the user audio commits.
|
||||
@@ -384,6 +392,27 @@ export class RealtimeTurns {
|
||||
}
|
||||
|
||||
const toolResponseId = payload.response_id || payload.response?.id || null;
|
||||
// Follow-up responses retain reset ownership until fresh admitted input.
|
||||
const arrivalRequest = this.globeRequest;
|
||||
const globeRequest =
|
||||
this.globeResponseRequests.get(toolResponseId) ?? arrivalRequest;
|
||||
// Later siblings must wait for globe admission before they can mutate state.
|
||||
// Other tool responses keep their existing concurrency.
|
||||
let heldForAdmission = false;
|
||||
while (toolResponseId && this.globeAdmissionGates.has(toolResponseId)) {
|
||||
// Drop claimed duplicates before every wait, before their keys can expire.
|
||||
if (
|
||||
calls.every((call) =>
|
||||
callDedupeKeys(call).some((key) => this.processedCalls.has(key)),
|
||||
)
|
||||
)
|
||||
return;
|
||||
heldForAdmission = true;
|
||||
await this.globeAdmissionGates.get(toolResponseId).promise;
|
||||
if (!this.ownsConversation(eventChannel) || this.isSessionEnding())
|
||||
return;
|
||||
}
|
||||
|
||||
// A newer typed command superseded the response these calls belong to.
|
||||
// They are stale intent — dispatching one would let the old turn mutate
|
||||
// the map after the operator asked for something else.
|
||||
@@ -394,7 +423,10 @@ export class RealtimeTurns {
|
||||
// hazard `callDedupeKeys` is written to avoid). So each refused call gets
|
||||
// a terminal output saying plainly that the turn moved on. No
|
||||
// `response.create` follows — the deferred typed turn is the single answer.
|
||||
if (this.isSupersededResponse(toolResponseId)) {
|
||||
if (
|
||||
this.isSupersededResponse(toolResponseId) ||
|
||||
(heldForAdmission && arrivalRequest !== this.globeRequest)
|
||||
) {
|
||||
this.pruneProcessedCalls();
|
||||
for (const call of calls) {
|
||||
const keys = callDedupeKeys(call);
|
||||
@@ -429,7 +461,19 @@ export class RealtimeTurns {
|
||||
if (keys.some((key) => this.processedCalls.has(key))) continue;
|
||||
keys.forEach((key) => this.processedCalls.set(key, performance.now()));
|
||||
|
||||
if (this.globeRefusals.get(toolResponseId)?.blocking) {
|
||||
this.sendToolOutput(call.call_id || call.id, {
|
||||
ok: false,
|
||||
action: call.name,
|
||||
skipped: true,
|
||||
error: 'Not run: this response must wait for the Globe View reply.',
|
||||
});
|
||||
continue;
|
||||
}
|
||||
|
||||
let result;
|
||||
let globeAdmissionGate = null;
|
||||
let globeResetGate = null;
|
||||
const resultChannel = this.dc;
|
||||
let radioHandoffEpochAtStart = this.radio.radioHandoffEpoch;
|
||||
let toolController = null;
|
||||
@@ -443,6 +487,80 @@ export class RealtimeTurns {
|
||||
this.narration?.preamble();
|
||||
try {
|
||||
const parsedArguments = parseArguments(call.arguments);
|
||||
if (
|
||||
toolResponseId &&
|
||||
call.name === 'move_camera' &&
|
||||
String(parsedArguments.motion || '').toLowerCase() === 'rotate_globe'
|
||||
) {
|
||||
let resolve;
|
||||
const promise = new Promise((settle) => {
|
||||
resolve = settle;
|
||||
});
|
||||
globeAdmissionGate = { promise, resolve };
|
||||
this.globeAdmissionGates.set(toolResponseId, globeAdmissionGate);
|
||||
// OpenAI delivers sibling calls concurrently. A reset dispatched
|
||||
// first must finish before rotation evaluates the camera's fit.
|
||||
// Hold later siblings at admission while this rotation waits.
|
||||
const resets = this.globeResets.get(toolResponseId);
|
||||
if (
|
||||
resets ||
|
||||
globeRequest.failed ||
|
||||
globeRequest.pending.size ||
|
||||
arrivalRequest !== this.globeRequest
|
||||
) {
|
||||
while (resets?.pending.size || globeRequest.pending.size)
|
||||
await Promise.all(
|
||||
[
|
||||
...new Set([
|
||||
...(resets?.pending || []),
|
||||
...globeRequest.pending,
|
||||
]),
|
||||
].map((gate) => gate.promise),
|
||||
);
|
||||
if (!this.ownsConversation(eventChannel) || this.isSessionEnding())
|
||||
return;
|
||||
keys.forEach((key) =>
|
||||
this.processedCalls.set(key, performance.now()),
|
||||
);
|
||||
const superseded =
|
||||
this.isSupersededResponse(toolResponseId) ||
|
||||
arrivalRequest !== this.globeRequest;
|
||||
if (superseded || resets?.failed || globeRequest.failed) {
|
||||
result = {
|
||||
ok: false,
|
||||
action: call.name,
|
||||
reason: 'globe-reset-failed',
|
||||
skipped: true,
|
||||
...(superseded ? { superseded: true } : {}),
|
||||
error: superseded
|
||||
? 'Superseded by a newer command from the operator — this call was not run.'
|
||||
: 'Not run: the Globe View reset/rotation sequence failed or was cancelled.',
|
||||
};
|
||||
}
|
||||
}
|
||||
}
|
||||
if (toolResponseId && call.name === 'zoom_to_globe') {
|
||||
let resolve;
|
||||
const promise = new Promise((settle) => {
|
||||
resolve = settle;
|
||||
});
|
||||
let resets = this.globeResets.get(toolResponseId);
|
||||
if (!resets) {
|
||||
resets = { pending: new Set(), failed: false };
|
||||
this.globeResets.set(toolResponseId, resets);
|
||||
// Retain settled outcomes for later siblings, without evicting an
|
||||
// active reset that another response still needs to await.
|
||||
for (const [id, record] of this.globeResets) {
|
||||
if (this.globeResets.size <= SUPERSEDED_RESPONSE_MEMORY) break;
|
||||
if (!record.pending.size && id !== toolResponseId)
|
||||
this.globeResets.delete(id);
|
||||
}
|
||||
}
|
||||
globeResetGate = { promise, resolve, resets, request: globeRequest };
|
||||
resets.pending.add(globeResetGate);
|
||||
globeRequest.pending.add(globeResetGate);
|
||||
}
|
||||
|
||||
const isRadioControlCall = call.name === 'control_radio';
|
||||
const isRadioVisibilityCall =
|
||||
call.name === 'set_layer_visibility' &&
|
||||
@@ -499,7 +617,7 @@ export class RealtimeTurns {
|
||||
authorityDomain: radioAuthorityDomain,
|
||||
});
|
||||
}
|
||||
result = await (this.actionExecutor || this.runner)(
|
||||
result ??= await (this.actionExecutor || this.runner)(
|
||||
call.name,
|
||||
parsedArguments,
|
||||
{
|
||||
@@ -515,7 +633,28 @@ export class RealtimeTurns {
|
||||
radioHandoffEpochAtStart === this.radio.radioHandoffEpoch),
|
||||
},
|
||||
);
|
||||
if (
|
||||
globeAdmissionGate &&
|
||||
this.globeResets.has(toolResponseId) &&
|
||||
result?.reason === 'globe-view-required'
|
||||
)
|
||||
result = { ...result, reason: 'globe-reset-failed' };
|
||||
if (!this.ownsConversation(eventChannel)) return;
|
||||
if (
|
||||
isGlobeViewRefusal(result) &&
|
||||
toolResponseId &&
|
||||
arrivalRequest === this.globeRequest &&
|
||||
!this.isSupersededResponse(toolResponseId)
|
||||
) {
|
||||
this.globeRefusals.set(toolResponseId, {
|
||||
result,
|
||||
followupQueued: false,
|
||||
blocking: !result.skipped,
|
||||
});
|
||||
while (this.globeRefusals.size > SUPERSEDED_RESPONSE_MEMORY) {
|
||||
this.globeRefusals.delete(this.globeRefusals.keys().next().value);
|
||||
}
|
||||
}
|
||||
if (result?.ok && result.radioPlaybackRequested) {
|
||||
const sessionIsCurrent =
|
||||
this.activeToolAbortControllers.has(toolController) &&
|
||||
@@ -595,6 +734,23 @@ export class RealtimeTurns {
|
||||
: {}),
|
||||
};
|
||||
} finally {
|
||||
if (globeResetGate) {
|
||||
const { resets, request } = globeResetGate;
|
||||
if (result?.ok !== true || result.cancelled) {
|
||||
resets.failed = true;
|
||||
request.failed = true;
|
||||
}
|
||||
resets.pending.delete(globeResetGate);
|
||||
request.pending.delete(globeResetGate);
|
||||
globeResetGate.resolve();
|
||||
}
|
||||
if (globeAdmissionGate) {
|
||||
if (
|
||||
this.globeAdmissionGates.get(toolResponseId) === globeAdmissionGate
|
||||
)
|
||||
this.globeAdmissionGates.delete(toolResponseId);
|
||||
globeAdmissionGate.resolve();
|
||||
}
|
||||
if (toolController) {
|
||||
this.activeToolAbortControllers.delete(toolController);
|
||||
this.radio.releaseTool(toolController);
|
||||
@@ -671,13 +827,38 @@ export class RealtimeTurns {
|
||||
});
|
||||
}
|
||||
if (!this.ownsConversation(eventChannel)) return;
|
||||
// Keep the Radio handoff wording authoritative even when another tool
|
||||
// result follows Radio in the same multi-intent response.
|
||||
this.queueResponseCreate(
|
||||
responseInstructionForToolResult(
|
||||
this.radio.pendingRadioPlaybackResult || lastResult,
|
||||
),
|
||||
);
|
||||
const refusal = this.globeRefusals.get(toolResponseId);
|
||||
// A newer turn may arrive during admission or the viewport-image await.
|
||||
// Preserve its reply; neither the refusal map nor lastResult may revive
|
||||
// an old refusal follow-up after supersession.
|
||||
if (
|
||||
(this.isSupersededResponse(toolResponseId) ||
|
||||
arrivalRequest !== this.globeRequest) &&
|
||||
(refusal || isGlobeViewRefusal(lastResult) || lastResult?.superseded)
|
||||
) {
|
||||
// Complete an idle turn below, but leave a newer running tool's status
|
||||
// authoritative while suppressing this obsolete refusal follow-up.
|
||||
if (this.activeToolAbortControllers.size) return;
|
||||
} else if (refusal) {
|
||||
// Even an earlier sibling finishing late cannot replace the reset instruction.
|
||||
if (!refusal.followupQueued) {
|
||||
refusal.followupQueued = true;
|
||||
this.queueResponseCreate({
|
||||
instructions: responseInstructionForToolResult(refusal.result),
|
||||
tool_choice: 'none',
|
||||
});
|
||||
}
|
||||
} else {
|
||||
// Keep Radio handoff wording authoritative for ordinary multi-tool replies.
|
||||
const followupResult =
|
||||
this.radio.pendingRadioPlaybackResult || lastResult;
|
||||
const instructions = responseInstructionForToolResult(followupResult);
|
||||
this.queueResponseCreate(
|
||||
isGlobeViewRefusal(followupResult)
|
||||
? { instructions, tool_choice: 'none' }
|
||||
: instructions,
|
||||
);
|
||||
}
|
||||
}
|
||||
this.setStatus('listening', t('Ask or command'));
|
||||
}
|
||||
@@ -966,6 +1147,22 @@ export class RealtimeTurns {
|
||||
this.userTurnPending = false;
|
||||
this.activeResponseId =
|
||||
payload.response?.id || payload.response_id || null;
|
||||
// Late calls retain only their globe reset/request ownership. Ordinary
|
||||
// tool admission and concurrency continue to use their existing policy.
|
||||
if (
|
||||
this.activeResponseId &&
|
||||
!this.globeResponseRequests.has(this.activeResponseId)
|
||||
) {
|
||||
this.globeResponseRequests.set(
|
||||
this.activeResponseId,
|
||||
this.globeRequest,
|
||||
);
|
||||
while (this.globeResponseRequests.size > SUPERSEDED_RESPONSE_MEMORY)
|
||||
this.globeResponseRequests.delete(
|
||||
this.globeResponseRequests.keys().next().value,
|
||||
);
|
||||
}
|
||||
|
||||
this.setVoiceSpeaker('ai');
|
||||
return;
|
||||
}
|
||||
@@ -1045,7 +1242,8 @@ export class RealtimeTurns {
|
||||
const sent = this.sendRealtimeEvent(
|
||||
{
|
||||
type: 'response.create',
|
||||
response: { instructions },
|
||||
response:
|
||||
typeof instructions === 'string' ? { instructions } : instructions,
|
||||
},
|
||||
'client.response_create.tool_followup',
|
||||
);
|
||||
@@ -1057,6 +1255,20 @@ export class RealtimeTurns {
|
||||
controller.abort();
|
||||
this.activeToolAbortControllers.clear();
|
||||
this.radio.clearTools();
|
||||
this.releaseGlobeResets();
|
||||
}
|
||||
|
||||
releaseGlobeResets() {
|
||||
for (const resets of this.globeResets.values()) {
|
||||
if (!resets.pending.size) continue;
|
||||
resets.failed = true;
|
||||
for (const gate of resets.pending) {
|
||||
gate.request.failed = true;
|
||||
gate.request.pending.delete(gate);
|
||||
gate.resolve();
|
||||
}
|
||||
resets.pending.clear();
|
||||
}
|
||||
}
|
||||
|
||||
reset() {
|
||||
@@ -1068,6 +1280,13 @@ export class RealtimeTurns {
|
||||
this.pendingUserTextResponse = false;
|
||||
this.activeResponseId = null;
|
||||
this.supersededResponseIds.clear();
|
||||
this.globeRefusals.clear();
|
||||
for (const gate of this.globeAdmissionGates.values()) gate.resolve();
|
||||
this.globeAdmissionGates.clear();
|
||||
this.releaseGlobeResets();
|
||||
this.globeResets.clear();
|
||||
this.globeResponseRequests.clear();
|
||||
this.globeRequest = { failed: false, pending: new Set() };
|
||||
this.narration?.cancel();
|
||||
this.assistantAudioPlaying = false;
|
||||
this.speechResponseIds.clear();
|
||||
|
||||
Reference in new issue
Block a user