niahere 0.2.91 → 0.3.0

package/README.md

@@ -36,13 +36,14 @@ nia start               # starts daemon + registers OS service
 
 - **Telegram** — message your agent from your phone, typing indicator while processing
 - **Slack** — Socket Mode bot with thread awareness, thinking emoji, watch channels for proactive monitoring
+- **Phone (voice)** — Twilio + OpenAI Realtime. Inbound calls from allowlisted contacts and outbound calls via `place_call` MCP tool. Scheduled jobs can dial you (morning standup, evening retro, escalation). See `/nia-phone` skill.
 - **Terminal chat** — REPL with session resume support
 - **Scheduled jobs** — recurring jobs and crons that run Claude and can message you back. Stateful by default (working memory), per-job model routing for cost savings
 - **Persona system** — customizable identity, soul, owner profile, rules, and memory (preloaded every session)
 - **Agents** — domain specialists (marketer, senior-dev) via Claude Agent SDK subagents
 - **Skills** — loads skills from multiple directories, invokable as slash commands
 - **Cross-platform service** — launchd (macOS), systemd (Linux), service-aware restart
-- **MCP tools** — 20 tools for job management, messaging, memory, rules, and channel control
+- **MCP tools** — 21 tools for job management, messaging, memory, rules, channel control, and outbound phone calls
 - **Background memory consolidation** — stages memory candidates from conversations automatically
 - **Session summaries** — optional handoff notes between sessions for continuity
 - **Backups** — `nia backup` with auto-backup before updates

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "niahere",
-  "version": "0.2.91",
+  "version": "0.3.0",
   "description": "A personal AI assistant daemon — chat, scheduled jobs, persona system, extensible via skills.",
   "type": "module",
   "scripts": {

package/skills/nia-phone/SKILL.md

@@ -26,26 +26,43 @@ Voice to the OpenAI Realtime API. It exposes:
 Transcripts persist to the `messages` table with `channel = 'phone'` and
 `room = phone-<callSid>`.
 
-## Required env vars
+## Configuration
 
-```bash
-TWILIO_SID            # Account SID (AC…) — or an API Key SID (SK…)
-TWILIO_SECRET         # Auth Token (if SID is AC) — or API Key Secret (if SID is SK)
-TWILIO_AUTH_TOKEN     # Required when SID is an API Key — signs webhooks.
-                      # Omit if TWILIO_SECRET is already the Auth Token.
-PRIMARY_PHONE_USER    # Owner's number in E.164 (e.g. +917667078414).
-PHONE_FROM_NUMBER     # Your Twilio number in E.164 (e.g. +13025480697).
-PUBLIC_BASE_URL       # https://<your-tunnel-hostname>  — NO trailing slash.
-OPENAI_API_KEY        # For the Realtime voice loop.
-
-# Optional
-PHONE_PORT=7079       # Local port the webhook server binds to.
-PHONE_ALLOWLIST=+12025550100,+14155551234   # Extra allowed inbound callers.
-PHONE_VOICE=marin     # Realtime voice (marin | cedar | shimmer | coral | alloy | ash | …).
-PHONE_REALTIME_MODEL=gpt-realtime   # Override if you want a specific model.
+Phone config lives in `~/.niahere/config.yaml` under `channels.phone` —
+same place as `channels.telegram` and `channels.slack`. Every field is
+overridable by the matching env var if you prefer `.env` for secrets.
+
+```yaml
+# ~/.niahere/config.yaml
+channels:
+  phone:
+    twilio_sid: AC... # Account SID — or an API Key SID (SK…)
+    twilio_secret: ... # Auth Token if SID is AC, API Key Secret if SID is SK
+    twilio_auth_token:
+      ... # Required when twilio_sid is an API Key (SK…); signs webhooks.
+      # Omit if twilio_secret is already the Auth Token.
+    from_number: "+1..." # Your Twilio number (E.164)
+    owner_number: "+91..." # Owner's phone (E.164) — highest-trust caller
+    public_base_url: https://nia.example.com # No trailing slash
+    openai_api_key: sk-proj-... # For the Realtime voice loop
+
+    # Optional
+    port: 7079 # Local port the webhook server binds to
+    allowlist: ["+12025550100"] # Extra allowed inbound callers (E.164)
+    voice: marin # Realtime voice (marin | cedar | shimmer | coral | alloy | ash | …)
+    realtime_model: gpt-realtime
+```
+
+Env overrides (use these if you'd rather keep secrets in `.env`):
+
+```
+TWILIO_SID, TWILIO_SECRET, TWILIO_AUTH_TOKEN
+PHONE_FROM_NUMBER, PRIMARY_PHONE_USER
+PUBLIC_BASE_URL, OPENAI_API_KEY
+PHONE_PORT, PHONE_ALLOWLIST (comma-separated), PHONE_VOICE, PHONE_REALTIME_MODEL
 ```
 
-`nia phone status` prints which vars are set / missing.
+`nia phone status` prints which fields are set / missing.
 
 ## Cloudflared named tunnel (production)
 

package/src/channels/phone/index.ts

@@ -370,17 +370,20 @@ class PhoneChannel implements Channel {
   private requireCreds(): { accountSid: string; authToken: string } {
     const sid = this.cfg.twilio_sid;
     const secret = this.cfg.twilio_secret;
-    if (!sid || !secret) throw new Error("phone: TWILIO_SID/TWILIO_SECRET not configured");
+    if (!sid || !secret)
+      throw new Error("phone: channels.phone.twilio_sid and twilio_secret not set (config.yaml or env)");
     return { accountSid: sid, authToken: secret };
   }
 
   private requirePublicBaseUrl(): string {
-    if (!this.cfg.public_base_url) throw new Error("phone: PUBLIC_BASE_URL not configured");
+    if (!this.cfg.public_base_url)
+      throw new Error("phone: channels.phone.public_base_url not set (config.yaml or PUBLIC_BASE_URL env)");
     return this.cfg.public_base_url;
   }
 
   private requireFromNumber(): string {
-    if (!this.cfg.from_number) throw new Error("phone: PHONE_FROM_NUMBER not configured");
+    if (!this.cfg.from_number)
+      throw new Error("phone: channels.phone.from_number not set (config.yaml or PHONE_FROM_NUMBER env)");
     return this.cfg.from_number;
   }
 

package/src/cli/phone.ts

@@ -45,7 +45,7 @@ async function phoneCallCommand(): Promise<void> {
   const channel = createPhoneChannel();
   if (!channel) {
     fail(
-      "Phone channel not configured. Need TWILIO_SID, TWILIO_SECRET, PHONE_FROM_NUMBER in .env (plus OPENAI_API_KEY and PUBLIC_BASE_URL for the realtime voice loop).",
+      "Phone channel not configured. Set channels.phone.{twilio_sid,twilio_secret,from_number} in ~/.niahere/config.yaml (also channels.phone.{openai_api_key,public_base_url} for the realtime voice loop). Env vars TWILIO_SID / TWILIO_SECRET / PHONE_FROM_NUMBER / OPENAI_API_KEY / PUBLIC_BASE_URL override if you prefer .env.",
     );
   }
 
@@ -53,13 +53,13 @@ async function phoneCallCommand(): Promise<void> {
   const cfg = getConfig().channels.phone;
   console.log(`${ICON_PASS} phone server up on :${cfg.port}`);
   if (!cfg.public_base_url) {
-    console.log(`${ICON_WARN} PUBLIC_BASE_URL not set — Twilio cannot reach this server.`);
-    console.log(`         Start cloudflared (or your tunnel) and set PUBLIC_BASE_URL in .env first.`);
+    console.log(`${ICON_WARN} public_base_url not set — Twilio cannot reach this server.`);
+    console.log(`         Start cloudflared (or your tunnel) and set channels.phone.public_base_url in config.yaml.`);
     await channel!.stop();
     process.exit(1);
   }
   if (!cfg.openai_api_key) {
-    console.log(`${ICON_WARN} OPENAI_API_KEY not set — realtime voice loop will fall back to TwiML <Say>.`);
+    console.log(`${ICON_WARN} openai_api_key not set — realtime voice loop will fall back to TwiML <Say>.`);
   }
 
   console.log(`  dialing ${number} ...`);
@@ -119,9 +119,13 @@ function helpText(): string {
     "                                phone server, dials, waits, prints transcript.",
     "  status                        Show phone channel configuration.",
     "",
-    "Required env:",
-    "  TWILIO_SID, TWILIO_SECRET, PHONE_FROM_NUMBER",
-    "  OPENAI_API_KEY (for realtime voice loop)",
-    "  PUBLIC_BASE_URL (cloudflared/ngrok tunnel pointing at PHONE_PORT)",
+    "Config lives in ~/.niahere/config.yaml under channels.phone:",
+    "  twilio_sid, twilio_secret, from_number          (required)",
+    "  openai_api_key, public_base_url                 (required for realtime voice loop)",
+    "  twilio_auth_token                               (required if twilio_sid is an API Key SID)",
+    "  port, voice, realtime_model, allowlist          (optional)",
+    "",
+    "Each field can be overridden by the matching env var (TWILIO_SID, OPENAI_API_KEY, etc.)",
+    "if you prefer .env. See the nia-phone skill for full deploy walkthrough.",
   ].join("\n");
 }

package/src/commands/init.ts

@@ -215,6 +215,60 @@ export async function runInit(): Promise<void> {
       }
     }
 
+    // Phone (Twilio Voice + OpenAI Realtime)
+    const exPh = (exCh.phone || {}) as Record<string, unknown>;
+    let phoneTwilioSid = (exPh.twilio_sid as string) || "";
+    let phoneTwilioSecret = (exPh.twilio_secret as string) || "";
+    let phoneTwilioAuthToken = (exPh.twilio_auth_token as string) || "";
+    let phoneFromNumber = (exPh.from_number as string) || "";
+    let phoneOwnerNumber = (exPh.owner_number as string) || "";
+    let phonePublicBaseUrl = (exPh.public_base_url as string) || "";
+    let phoneOpenAiKey = (exPh.openai_api_key as string) || "";
+    let phoneVoice = (exPh.voice as string) || "";
+
+    const existingPhoneSid = phoneTwilioSid;
+    if (existingPhoneSid) {
+      const masked = `...${existingPhoneSid.slice(-6)}`;
+      const reconfigure = await ask(rl, `\nPhone (Twilio + Realtime): configured (${masked}). Reconfigure? (y/n)`, "n");
+      if (reconfigure.toLowerCase() === "y") {
+        phoneTwilioSid = (await ask(rl, "Twilio SID (AC… or SK…)", phoneTwilioSid)) || phoneTwilioSid;
+        phoneTwilioSecret =
+          (await ask(rl, "Twilio Secret (Auth Token if AC, API Key Secret if SK)", phoneTwilioSecret)) ||
+          phoneTwilioSecret;
+        if (phoneTwilioSid.startsWith("SK")) {
+          phoneTwilioAuthToken =
+            (await ask(rl, "Twilio Auth Token (account-level — needed for webhook signing)", phoneTwilioAuthToken)) ||
+            phoneTwilioAuthToken;
+        }
+        phoneFromNumber =
+          (await ask(rl, "Twilio number to dial from (E.164, e.g. +13025551234)", phoneFromNumber)) || phoneFromNumber;
+        phoneOwnerNumber = (await ask(rl, "Your phone (E.164)", phoneOwnerNumber)) || phoneOwnerNumber;
+        phonePublicBaseUrl =
+          (await ask(rl, "Public base URL (cloudflared/ngrok https://…)", phonePublicBaseUrl)) || phonePublicBaseUrl;
+        phoneOpenAiKey = (await ask(rl, "OpenAI API key (for Realtime voice loop)", phoneOpenAiKey)) || phoneOpenAiKey;
+        phoneVoice =
+          (await ask(rl, "Realtime voice (marin, cedar, shimmer, coral, alloy…)", phoneVoice || "marin")) || phoneVoice;
+      }
+    } else {
+      const setupPhone = await ask(rl, "\nSet up phone (Twilio + OpenAI Realtime voice calls)? (y/n)", "n");
+      if (setupPhone.toLowerCase() === "y") {
+        console.log("  You'll need: a Twilio voice number, your phone number, an OpenAI API key, and a public tunnel.");
+        console.log("  See /nia-phone skill for the full deploy walkthrough.\n");
+        phoneTwilioSid = await ask(rl, "Twilio SID (AC… or SK…)", "");
+        if (phoneTwilioSid) {
+          phoneTwilioSecret = await ask(rl, "Twilio Secret (Auth Token if AC, API Key Secret if SK)", "");
+          if (phoneTwilioSid.startsWith("SK")) {
+            phoneTwilioAuthToken = await ask(rl, "Twilio Auth Token (account-level — for webhook signing)", "");
+          }
+          phoneFromNumber = await ask(rl, "Twilio number to dial from (E.164, e.g. +13025551234)", "");
+          phoneOwnerNumber = await ask(rl, "Your phone (E.164)", "");
+          phonePublicBaseUrl = await ask(rl, "Public base URL (cloudflared/ngrok https://…)", "");
+          phoneOpenAiKey = await ask(rl, "OpenAI API key", "");
+          phoneVoice = await ask(rl, "Realtime voice", "marin");
+        }
+      }
+    }
+
     // Gemini API key (for image generation)
     let geminiApiKey = "";
     const existingGemini = (existing.gemini_api_key as string) || "";
@@ -430,6 +484,19 @@ export async function runInit(): Promise<void> {
     if (slackBotToken && !telegramToken) {
       channels.default = "slack";
     }
+    if (phoneTwilioSid && phoneTwilioSecret && phoneFromNumber) {
+      const ph: Record<string, unknown> = {
+        twilio_sid: phoneTwilioSid,
+        twilio_secret: phoneTwilioSecret,
+        from_number: phoneFromNumber,
+      };
+      if (phoneTwilioAuthToken) ph.twilio_auth_token = phoneTwilioAuthToken;
+      if (phoneOwnerNumber) ph.owner_number = phoneOwnerNumber;
+      if (phonePublicBaseUrl) ph.public_base_url = phonePublicBaseUrl.replace(/\/$/, "");
+      if (phoneOpenAiKey) ph.openai_api_key = phoneOpenAiKey;
+      if (phoneVoice && phoneVoice !== "marin") ph.voice = phoneVoice;
+      channels.phone = ph;
+    }
     if (Object.keys(channels).length > 0) {
       config.channels = channels;
     }

package/src/mcp/tools.ts

@@ -480,7 +480,7 @@ export async function placeCall(args: {
   const { getPhoneChannel } = await import("../channels/phone");
   const phone = getPhoneChannel();
   if (!phone) {
-    return "Phone channel is not configured. Set TWILIO_SID, TWILIO_SECRET, PHONE_FROM_NUMBER, PUBLIC_BASE_URL, OPENAI_API_KEY in .env and restart the daemon.";
+    return "Phone channel is not configured. Add a channels.phone block to ~/.niahere/config.yaml with twilio_sid, twilio_secret, from_number, public_base_url, openai_api_key (or set the matching env vars in .env), then restart the daemon.";
   }
   try {
     const result = await phone.placeCall({