@onereach/step-voice 7.0.26 → 7.0.27
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dst/Voice Agent Function Result.js +10 -14
- package/dst/Voice Agent.d.ts +31 -0
- package/dst/Voice Agent.js +158 -5
- package/dst/voice.js +1 -2
- package/package.json +1 -1
|
@@ -4,31 +4,27 @@ const tslib_1 = require("tslib");
|
|
|
4
4
|
const voice_1 = tslib_1.__importDefault(require("./voice"));
|
|
5
5
|
class VoiceAgentFunctionResult extends voice_1.default {
|
|
6
6
|
async runStep() {
|
|
7
|
-
const ctx = await this.thread.get('__voiceAgentContext') ?? {};
|
|
8
|
-
const { callId, callCallback, name: functionName, functionId } = ctx;
|
|
9
7
|
const { result, stopAgent } = this.data;
|
|
10
|
-
|
|
8
|
+
const ctx = await this.thread.get(`__voiceAgentContext_${this.thread.id}`);
|
|
9
|
+
if (!ctx?.callId) {
|
|
11
10
|
throw new Error('Voice Agent Function Result: missing call context (must be placed on a Voice Agent function exit)');
|
|
12
11
|
}
|
|
13
|
-
|
|
12
|
+
const { callId, callCallback, name: functionName, functionId } = ctx;
|
|
13
|
+
this.log.info("Voice Agent Function Result", { functionId, callId, threadId: this.thread.id });
|
|
14
14
|
const params = {
|
|
15
|
-
functionName
|
|
16
|
-
functionId
|
|
17
|
-
result
|
|
18
|
-
options: {
|
|
19
|
-
stop: stopAgent
|
|
20
|
-
}
|
|
15
|
+
functionName,
|
|
16
|
+
functionId,
|
|
17
|
+
result,
|
|
18
|
+
options: { stop: stopAgent }
|
|
21
19
|
};
|
|
22
|
-
|
|
20
|
+
await this.thread.eventManager.emit({
|
|
23
21
|
target: this.helpers.providersAccountId,
|
|
24
22
|
name: 'out/voice/voicer',
|
|
25
23
|
params: {
|
|
26
24
|
id: callId,
|
|
27
25
|
commands: [{ name: 'agentFunctionResult', params }]
|
|
28
26
|
}
|
|
29
|
-
}
|
|
30
|
-
this.log.info("Voice Agent Function Result", event);
|
|
31
|
-
await this.thread.eventManager.emit(event, {
|
|
27
|
+
}, {
|
|
32
28
|
target: callCallback,
|
|
33
29
|
invocationType: 'async',
|
|
34
30
|
timeout: 5000
|
package/dst/Voice Agent.d.ts
CHANGED
|
@@ -1,8 +1,34 @@
|
|
|
1
1
|
import VoiceStep from './voice';
|
|
2
|
+
type PreToolSpeechMode = 'auto' | 'static' | 'off';
|
|
3
|
+
type ToolCallSound = 'typing' | 'none';
|
|
4
|
+
type ToolCallSoundBehavior = 'always' | 'with_pre_speech';
|
|
5
|
+
type SoftTimeoutMode = 'static' | 'auto';
|
|
6
|
+
interface ToolCallFeedbackConfig {
|
|
7
|
+
enabled: boolean;
|
|
8
|
+
pre_tool_speech: PreToolSpeechMode;
|
|
9
|
+
pre_tool_speech_phrases: string[];
|
|
10
|
+
tool_call_sound: ToolCallSound;
|
|
11
|
+
tool_call_sound_behavior: ToolCallSoundBehavior;
|
|
12
|
+
delay_seconds: number;
|
|
13
|
+
}
|
|
14
|
+
interface PerToolConfig {
|
|
15
|
+
enabled?: boolean | null;
|
|
16
|
+
pre_tool_speech?: PreToolSpeechMode | null;
|
|
17
|
+
pre_tool_speech_phrases?: string[] | null;
|
|
18
|
+
tool_call_sound?: ToolCallSound | null;
|
|
19
|
+
tool_call_sound_behavior?: ToolCallSoundBehavior | null;
|
|
20
|
+
delay_seconds?: number | null;
|
|
21
|
+
}
|
|
22
|
+
interface SoftTimeoutConfig {
|
|
23
|
+
timeout_seconds: number;
|
|
24
|
+
phrases: string[];
|
|
25
|
+
mode: SoftTimeoutMode;
|
|
26
|
+
}
|
|
2
27
|
interface Handler {
|
|
3
28
|
name: string;
|
|
4
29
|
description: string;
|
|
5
30
|
parameters: string;
|
|
31
|
+
feedback?: PerToolConfig;
|
|
6
32
|
}
|
|
7
33
|
interface AgentServiceConfig {
|
|
8
34
|
provider: string;
|
|
@@ -35,14 +61,19 @@ interface INPUT {
|
|
|
35
61
|
llm: Partial<AgentServiceConfig>;
|
|
36
62
|
tts: Partial<AgentServiceConfig>;
|
|
37
63
|
realtime: Partial<AgentServiceConfig>;
|
|
64
|
+
tool_feedback?: ToolCallFeedbackConfig;
|
|
65
|
+
soft_timeout?: SoftTimeoutConfig;
|
|
38
66
|
}
|
|
39
67
|
export default class VoiceAgent extends VoiceStep<Partial<INPUT>> {
|
|
40
68
|
runStep(): Promise<void>;
|
|
41
69
|
private normalizeIdlePrompts;
|
|
42
70
|
private findExitByLabel;
|
|
71
|
+
private validateToolFeedback;
|
|
43
72
|
private isStepFlagEnabled;
|
|
44
73
|
private isValidAgentUrl;
|
|
45
74
|
private buildAgentServiceConfig;
|
|
75
|
+
private validatePerToolFeedback;
|
|
76
|
+
private validateSoftTimeout;
|
|
46
77
|
private parseAgentServiceOptions;
|
|
47
78
|
exitToThread(): Promise<void>;
|
|
48
79
|
}
|
package/dst/Voice Agent.js
CHANGED
|
@@ -17,7 +17,15 @@ class VoiceAgent extends voice_1.default {
|
|
|
17
17
|
const functionId = event.params.functionId ?? '';
|
|
18
18
|
const exitId = this.findExitByLabel(name) ?? this.getExitStepId('function_call');
|
|
19
19
|
if (exitId) {
|
|
20
|
-
this.exitStep(exitId, {
|
|
20
|
+
this.exitStep(exitId, {
|
|
21
|
+
[name]: params,
|
|
22
|
+
name,
|
|
23
|
+
functionId,
|
|
24
|
+
__voicecb: cb,
|
|
25
|
+
callId: call.id,
|
|
26
|
+
callType: call.type,
|
|
27
|
+
callCallback: call.callback
|
|
28
|
+
}, true);
|
|
21
29
|
}
|
|
22
30
|
else {
|
|
23
31
|
this.log.warn('no exit for function_call', { name });
|
|
@@ -34,7 +42,7 @@ class VoiceAgent extends voice_1.default {
|
|
|
34
42
|
}
|
|
35
43
|
});
|
|
36
44
|
this.triggers.otherwise(async () => {
|
|
37
|
-
const { handlers = [], system_instruction = '', developer_messages = [], greeting_message = '', prevent_greeting_interruption = false, idle_detection_enabled = false, idle_timeout_secs = 8, idle_max_retries = 3, idle_prompts = [], idle_goodbye_message = '', dtmf_enabled = false, dtmf_timeout = 2, dtmf_termination_digit = '#', dtmf_prefix = 'DTMF: ', custom_config_enabled = false, agent_url_overwritten = false, agent_url = '', mode = '', stt = {}, llm = {}, tts = {}, realtime = {} } = this.data;
|
|
45
|
+
const { handlers = [], system_instruction = '', developer_messages = [], greeting_message = '', prevent_greeting_interruption = false, idle_detection_enabled = false, idle_timeout_secs = 8, idle_max_retries = 3, idle_prompts = [], idle_goodbye_message = '', dtmf_enabled = false, dtmf_timeout = 2, dtmf_termination_digit = '#', dtmf_prefix = 'DTMF: ', custom_config_enabled = false, agent_url_overwritten = false, tool_feedback, soft_timeout, agent_url = '', mode = '', stt = {}, llm = {}, tts = {}, realtime = {} } = this.data;
|
|
38
46
|
if (lodash_1.default.isEmpty(system_instruction)) {
|
|
39
47
|
this.throwError(new Error('Voice Agent: system instruction is required'));
|
|
40
48
|
}
|
|
@@ -48,7 +56,19 @@ class VoiceAgent extends voice_1.default {
|
|
|
48
56
|
parameters = {};
|
|
49
57
|
}
|
|
50
58
|
}
|
|
51
|
-
|
|
59
|
+
const tool = {
|
|
60
|
+
type: 'function',
|
|
61
|
+
name: h.name,
|
|
62
|
+
description: h.description,
|
|
63
|
+
parameters: parameters || {}
|
|
64
|
+
};
|
|
65
|
+
if (h.feedback != null) {
|
|
66
|
+
const validatedFeedback = this.validatePerToolFeedback(h.feedback, h.name);
|
|
67
|
+
if (validatedFeedback) {
|
|
68
|
+
tool.feedback = validatedFeedback;
|
|
69
|
+
}
|
|
70
|
+
}
|
|
71
|
+
return tool;
|
|
52
72
|
});
|
|
53
73
|
const config = {
|
|
54
74
|
system_instruction,
|
|
@@ -102,6 +122,18 @@ class VoiceAgent extends voice_1.default {
|
|
|
102
122
|
this.throwError(new Error(`Voice Agent: unknown mode "${mode}"`));
|
|
103
123
|
}
|
|
104
124
|
}
|
|
125
|
+
if (this.isStepFlagEnabled(tool_feedback?.enabled) && tools.length > 0) {
|
|
126
|
+
const validatedFeedback = this.validateToolFeedback(tool_feedback);
|
|
127
|
+
if (validatedFeedback) {
|
|
128
|
+
config.tool_feedback = validatedFeedback;
|
|
129
|
+
}
|
|
130
|
+
}
|
|
131
|
+
if (soft_timeout != null) {
|
|
132
|
+
const validatedSoftTimeout = this.validateSoftTimeout(soft_timeout);
|
|
133
|
+
if (validatedSoftTimeout) {
|
|
134
|
+
config.soft_timeout = validatedSoftTimeout;
|
|
135
|
+
}
|
|
136
|
+
}
|
|
105
137
|
const params = { config };
|
|
106
138
|
if (agentCustomUrl) {
|
|
107
139
|
params.url = agentCustomUrl;
|
|
@@ -119,6 +151,57 @@ class VoiceAgent extends voice_1.default {
|
|
|
119
151
|
const exit = lodash_1.default.find(this.step.exits, (e) => e.label === label);
|
|
120
152
|
return exit?.id;
|
|
121
153
|
}
|
|
154
|
+
validateToolFeedback(config) {
|
|
155
|
+
if (config == null || typeof config !== 'object')
|
|
156
|
+
return null;
|
|
157
|
+
const cfg = config;
|
|
158
|
+
// If enabled is not boolean true, silently return undefined (disabled case)
|
|
159
|
+
if (typeof cfg.enabled !== 'boolean' || cfg.enabled !== true)
|
|
160
|
+
return null;
|
|
161
|
+
// Validate pre_tool_speech
|
|
162
|
+
const validPreToolSpeech = ['auto', 'static', 'off'];
|
|
163
|
+
if (!validPreToolSpeech.includes(cfg.pre_tool_speech)) {
|
|
164
|
+
this.log.warn('Voice Agent: invalid tool_feedback.pre_tool_speech — feedback config excluded');
|
|
165
|
+
return null;
|
|
166
|
+
}
|
|
167
|
+
// Validate tool_call_sound
|
|
168
|
+
const validToolCallSound = ['typing', 'none'];
|
|
169
|
+
if (!validToolCallSound.includes(cfg.tool_call_sound)) {
|
|
170
|
+
this.log.warn('Voice Agent: invalid tool_feedback.tool_call_sound — feedback config excluded');
|
|
171
|
+
return null;
|
|
172
|
+
}
|
|
173
|
+
// Validate tool_call_sound_behavior
|
|
174
|
+
const validToolCallSoundBehavior = ['always', 'with_pre_speech'];
|
|
175
|
+
if (!validToolCallSoundBehavior.includes(cfg.tool_call_sound_behavior)) {
|
|
176
|
+
this.log.warn('Voice Agent: invalid tool_feedback.tool_call_sound_behavior — feedback config excluded');
|
|
177
|
+
return null;
|
|
178
|
+
}
|
|
179
|
+
// Validate delay_seconds
|
|
180
|
+
const delay = Number(cfg.delay_seconds);
|
|
181
|
+
if (delay <= 0 || delay > 30) {
|
|
182
|
+
this.log.warn('Voice Agent: invalid tool_feedback.delay_seconds — feedback config excluded');
|
|
183
|
+
return null;
|
|
184
|
+
}
|
|
185
|
+
// Validate pre_tool_speech_phrases (only required when mode is 'static')
|
|
186
|
+
const phrases = cfg.pre_tool_speech_phrases;
|
|
187
|
+
if (cfg.pre_tool_speech === 'static') {
|
|
188
|
+
if (!Array.isArray(phrases) ||
|
|
189
|
+
phrases.length < 1 ||
|
|
190
|
+
phrases.length > 20 ||
|
|
191
|
+
!phrases.every((p) => typeof p === 'string' && p.trim().length > 0)) {
|
|
192
|
+
this.log.warn('Voice Agent: invalid tool_feedback.pre_tool_speech_phrases — feedback config excluded');
|
|
193
|
+
return null;
|
|
194
|
+
}
|
|
195
|
+
}
|
|
196
|
+
return {
|
|
197
|
+
enabled: true,
|
|
198
|
+
pre_tool_speech: cfg.pre_tool_speech,
|
|
199
|
+
pre_tool_speech_phrases: (Array.isArray(phrases) ? phrases : []),
|
|
200
|
+
tool_call_sound: cfg.tool_call_sound,
|
|
201
|
+
tool_call_sound_behavior: cfg.tool_call_sound_behavior,
|
|
202
|
+
delay_seconds: delay
|
|
203
|
+
};
|
|
204
|
+
}
|
|
122
205
|
isStepFlagEnabled(value) {
|
|
123
206
|
return value === true || value === 'true';
|
|
124
207
|
}
|
|
@@ -138,6 +221,75 @@ class VoiceAgent extends voice_1.default {
|
|
|
138
221
|
options: this.parseAgentServiceOptions(label, service.options)
|
|
139
222
|
};
|
|
140
223
|
}
|
|
224
|
+
validatePerToolFeedback(config, toolName) {
|
|
225
|
+
if (config == null || typeof config !== 'object' || Array.isArray(config)) {
|
|
226
|
+
return null;
|
|
227
|
+
}
|
|
228
|
+
const raw = config;
|
|
229
|
+
const result = {};
|
|
230
|
+
if ('delay_seconds' in raw && raw.delay_seconds != null) {
|
|
231
|
+
const delay = Number(raw.delay_seconds);
|
|
232
|
+
if (delay <= 0 || delay > 30) {
|
|
233
|
+
this.log.warn(`Voice Agent: invalid feedback.delay_seconds for tool "${toolName}" — per-tool feedback excluded`);
|
|
234
|
+
return null;
|
|
235
|
+
}
|
|
236
|
+
result.delay_seconds = delay;
|
|
237
|
+
}
|
|
238
|
+
if ('pre_tool_speech' in raw && raw.pre_tool_speech != null) {
|
|
239
|
+
const valid = ['auto', 'static', 'off'];
|
|
240
|
+
if (!valid.includes(raw.pre_tool_speech)) {
|
|
241
|
+
this.log.warn(`Voice Agent: invalid feedback.pre_tool_speech for tool "${toolName}" — per-tool feedback excluded`);
|
|
242
|
+
return null;
|
|
243
|
+
}
|
|
244
|
+
result.pre_tool_speech = raw.pre_tool_speech;
|
|
245
|
+
}
|
|
246
|
+
if ('tool_call_sound' in raw && raw.tool_call_sound != null) {
|
|
247
|
+
const valid = ['typing', 'none'];
|
|
248
|
+
if (!valid.includes(raw.tool_call_sound)) {
|
|
249
|
+
this.log.warn(`Voice Agent: invalid feedback.tool_call_sound for tool "${toolName}" — per-tool feedback excluded`);
|
|
250
|
+
return null;
|
|
251
|
+
}
|
|
252
|
+
result.tool_call_sound = raw.tool_call_sound;
|
|
253
|
+
}
|
|
254
|
+
if ('tool_call_sound_behavior' in raw && raw.tool_call_sound_behavior != null) {
|
|
255
|
+
const valid = ['always', 'with_pre_speech'];
|
|
256
|
+
if (!valid.includes(raw.tool_call_sound_behavior)) {
|
|
257
|
+
this.log.warn(`Voice Agent: invalid feedback.tool_call_sound_behavior for tool "${toolName}" — per-tool feedback excluded`);
|
|
258
|
+
return null;
|
|
259
|
+
}
|
|
260
|
+
result.tool_call_sound_behavior = raw.tool_call_sound_behavior;
|
|
261
|
+
}
|
|
262
|
+
return result;
|
|
263
|
+
}
|
|
264
|
+
validateSoftTimeout(config) {
|
|
265
|
+
if (config == null || typeof config !== 'object' || Array.isArray(config))
|
|
266
|
+
return null;
|
|
267
|
+
const cfg = config;
|
|
268
|
+
const timeoutSeconds = Number(cfg.timeout_seconds);
|
|
269
|
+
if (timeoutSeconds <= 0 || timeoutSeconds > 30) {
|
|
270
|
+
this.log.warn('Voice Agent: invalid soft_timeout.timeout_seconds — soft timeout config excluded');
|
|
271
|
+
return null;
|
|
272
|
+
}
|
|
273
|
+
const validModes = ['static', 'auto'];
|
|
274
|
+
if (!validModes.includes(cfg.mode)) {
|
|
275
|
+
this.log.warn('Voice Agent: invalid soft_timeout.mode — soft timeout config excluded');
|
|
276
|
+
return null;
|
|
277
|
+
}
|
|
278
|
+
// Validate phrases
|
|
279
|
+
const phrases = cfg.phrases;
|
|
280
|
+
if (!Array.isArray(phrases) ||
|
|
281
|
+
phrases.length < 1 ||
|
|
282
|
+
phrases.length > 20 ||
|
|
283
|
+
!phrases.every((p) => typeof p === 'string' && p.trim().length > 0)) {
|
|
284
|
+
this.log.warn('Voice Agent: invalid soft_timeout.phrases — soft timeout config excluded');
|
|
285
|
+
return null;
|
|
286
|
+
}
|
|
287
|
+
return {
|
|
288
|
+
timeout_seconds: timeoutSeconds,
|
|
289
|
+
phrases: phrases,
|
|
290
|
+
mode: cfg.mode
|
|
291
|
+
};
|
|
292
|
+
}
|
|
141
293
|
parseAgentServiceOptions(label, options) {
|
|
142
294
|
if (options == null || options === '')
|
|
143
295
|
return undefined;
|
|
@@ -155,8 +307,9 @@ class VoiceAgent extends voice_1.default {
|
|
|
155
307
|
return options;
|
|
156
308
|
}
|
|
157
309
|
async exitToThread() {
|
|
158
|
-
|
|
159
|
-
this.thread.
|
|
310
|
+
const result = this.state.result;
|
|
311
|
+
await this.thread.set(`__voiceAgentContext_${this.thread.id}`, result);
|
|
312
|
+
this.thread.exitStep(this.state.exitStep, result);
|
|
160
313
|
}
|
|
161
314
|
}
|
|
162
315
|
exports.default = VoiceAgent;
|
package/dst/voice.js
CHANGED
|
@@ -18,8 +18,7 @@ class VoiceStep extends step_1.default {
|
|
|
18
18
|
if (this.cache != null) {
|
|
19
19
|
if (this.thread.id !== this.dataThreadId)
|
|
20
20
|
this.thread.refThread(this.dataThreadId);
|
|
21
|
-
|
|
22
|
-
await this.handleHeartbeat(this.cache);
|
|
21
|
+
await this.handleHeartbeat(this.cache);
|
|
23
22
|
}
|
|
24
23
|
}
|
|
25
24
|
exitStep(exitId, data, byThread = false) {
|