You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix(web): use session flavor label in voice context formatters
Replaces hardcoded 'Claude Code' strings in voice context injections with
the active session's flavor label (Cursor, Codex, Gemini, etc.) via
getFlavorLabel() from @hapi/protocol. Falls back to 'coding agent' for
unknown or missing flavors.
Threads an agentLabel param through formatMessage, formatPermissionRequest,
formatReadyEvent, formatNewMessages, formatHistory, and formatSessionFull.
voiceHooks resolves the label once per call via session.metadata.flavor.
Fixestiann#680
via [HAPI](https://hapi.run)
Co-Authored-By: HAPI <noreply@hapi.run>
Users control these agents through the Hapi web interface or Telegram Mini App. You serve as the voice interface to whichever agent is currently active.
25
27
@@ -39,7 +41,7 @@ These updates appear as system messages. You do NOT need to poll or ask for upda
39
41
Send user requests to the active coding agent.
40
42
41
43
When to use:
42
-
- User says "ask Claude to..." or "have it..."
44
+
- User says "ask the agent to...", "have it...", or any similar delegation
43
45
- Any coding, file, or development request
44
46
- User wants to continue a task
45
47
@@ -88,7 +90,7 @@ You will receive a context update when the agent responds or finishes.
88
90
89
91
## Request Routing
90
92
- Direct address ("Assistant, explain...") -> Answer yourself
91
-
- Explicit delegation ("Have Claude...") -> Use messageCodingAgent
93
+
- Explicit delegation ("Have it...", "Ask the agent...") -> Use messageCodingAgent
92
94
- Coding/file tasks -> Use messageCodingAgent
93
95
- General questions you can answer -> Answer yourself
94
96
@@ -109,7 +111,7 @@ Stay silent when:
109
111
110
112
## Permission Requests
111
113
When you see a permission request, immediately inform the user:
112
-
"Claude wants to run a bash command. Should I allow it?"
114
+
"The agent wants to run a bash command. Should I allow it?"
113
115
Then wait for their response and use processPermissionRequest.
return`Claude Code done working in session: ${sessionId}. The previous message(s) are the summary of the work done. Report this to the human immediately.`
return`${agentLabel} done working in session: ${sessionId}. The previous message(s) are the summary of the work done. Report this to the human immediately.`
0 commit comments