@fastagent-sh/voicenote 0.19.0 → 0.21.0

package/README.md

@@ -113,6 +113,8 @@ Optional settings:
   "VOICENOTE_RECORD_DIR": "/Volumes/VTR6500/RECORD",
   "VOICENOTE_MAX_AGE_HOURS": "48",
   "VOICENOTE_PI_BIN": "pi",
+  "VOICENOTE_PI_MODEL": "openai-codex/gpt-5.6-sol",
+  "PI_CODING_AGENT_DIR": "$HOME/.config/voicenote/pi-agent",
   "VOICENOTE_PI_THINKING": "high",
   "VOICENOTE_PI_SUMMARY_TOOLS": "read,grep",
   "VOICENOTE_CONTEXT_DIR": "/Users/you/vault"
@@ -121,12 +123,29 @@ Optional settings:
 
 ### Which model writes the notes
 
-voicenote does not choose one. It runs `pi -p` with no `--provider`/`--model`, so
-the provider, model and credentials are whatever pi itself is configured to use
-(`pi` → `/login <provider>`, pi's settings, or a provider API key in the
-environment). Change the model in pi, not here. There is no fallback to a second
-provider: if pi fails, the transcript is kept and the summary can be retried with
-`vn run --latest`.
+`VOICENOTE_PI_MODEL` is passed straight to pi as `--model`. pi accepts
+`provider/id`, so one value pins both (`openai-codex/gpt-5.6-sol`). Leave it unset
+and pi's own configured default model is used.
+
+Credentials always belong to pi (`pi` → `/login <provider>`, or a provider API key
+in the environment); voicenote never picks a provider and never falls back to a
+second one. If pi fails, the transcript is kept and the summary can be retried
+with `vn run --latest`.
+
+### Credentials of their own
+
+`PI_CODING_AGENT_DIR` relocates pi's config directory, which is where it keeps
+`auth.json`. Point it at a voicenote-owned directory and `vn login` writes there,
+pi refreshes the tokens there, and an interactive pi session cannot clobber
+them — it rewrites its own `auth.json` wholesale on exit, which has silently
+dropped providers before:
+
+```bash
+echo '{"env":{"PI_CODING_AGENT_DIR":"$HOME/.config/voicenote/pi-agent"}}' | vn config set
+vn login   # signs in and stores credentials in that directory
+```
+
+`vn doctor` prints the path it will read (`pi.auth=...`).
 
 `DEEPSEEK_API_KEY` and `OPENAI_API_KEY` in the config are only forwarded to pi's
 environment for providers that read them.

package/README.zh-CN.md

@@ -113,6 +113,8 @@ brew install ffmpeg
   "VOICENOTE_RECORD_DIR": "/Volumes/VTR6500/RECORD",
   "VOICENOTE_MAX_AGE_HOURS": "48",
   "VOICENOTE_PI_BIN": "pi",
+  "VOICENOTE_PI_MODEL": "openai-codex/gpt-5.6-sol",
+  "PI_CODING_AGENT_DIR": "$HOME/.config/voicenote/pi-agent",
   "VOICENOTE_PI_THINKING": "high",
   "VOICENOTE_PI_SUMMARY_TOOLS": "read,grep",
   "VOICENOTE_CONTEXT_DIR": "/Users/you/vault"
@@ -121,10 +123,26 @@ brew install ffmpeg
 
 ### 纪要用哪个模型
 
-voicenote 不选。它执行 `pi -p`, 不传 `--provider`/`--model`, 所以 provider、模型和凭证
-全部来自 pi 自己的配置(`pi` → `/login <provider>`、pi 的设置, 或环境里的 provider
-API key)。要换模型就去 pi 里改。也不会回退到第二个 provider: pi 失败时 transcript 会
-保留, 用 `vn run --latest` 重试纪要即可。
+`VOICENOTE_PI_MODEL` 会原样传给 pi 的 `--model`。pi 支持 `provider/id` 写法, 所以
+一个值就能同时定 provider 和模型(`openai-codex/gpt-5.6-sol`)。不设就用 pi 自己配置
+的默认模型。
+
+凭证始终属于 pi(`pi` → `/login <provider>`, 或环境里的 provider API key); voicenote
+不选 provider, 也不会回退到第二个。pi 失败时 transcript 会保留, 用 `vn run --latest`
+重试纪要即可。
+
+### 给 voicenote 一份独立凭证
+
+`PI_CODING_AGENT_DIR` 可以改掉 pi 的配置目录, `auth.json` 就存在那里。指向一个
+voicenote 自己的目录后, `vn login` 写入这里, pi 也在这里刷新 token, 交互式 pi 会话
+碰不到它 —— 那边退出时会整份覆写自己的 `auth.json`, 之前就把 provider 条目悄悄弄丢过:
+
+```bash
+echo '{"env":{"PI_CODING_AGENT_DIR":"$HOME/.config/voicenote/pi-agent"}}' | vn config set
+vn login   # 登录后凭证就存在该目录
+```
+
+`vn doctor` 的 `pi.auth=` 一行会打印实际读取的路径。
 
 配置里的 `DEEPSEEK_API_KEY` / `OPENAI_API_KEY` 只是透传给 pi 的环境变量。
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fastagent-sh/voicenote",
-  "version": "0.19.0",
+  "version": "0.21.0",
   "description": "Voice recordings → diarized transcripts → integrated semantic Markdown notes. Currently optimized for the PHILIPS VTR6500 recorder, but the workflow is generic.",
   "type": "module",
   "license": "MIT",

package/src/cli.ts

@@ -12,7 +12,7 @@ import { fileURLToPath, pathToFileURL } from 'node:url'
 import { spawn, spawnSync } from 'node:child_process'
 import os from 'node:os'
 
-const VERSION = '0.19.0'
+const VERSION = '0.21.0'
 const LAUNCH_AGENT_LABEL = 'sh.fastagent.voicenote'
 const LAUNCH_AGENT_LABEL_LEGACY = 'com.kid7st.voicenote' // pre-fastagent installs; cleaned up on install
 const TASK_NAME = 'VoiceNote'   // Windows Task Scheduler name (mac uses LAUNCH_AGENT_LABEL)
@@ -31,7 +31,15 @@ const LOG_DIR = join(STATE_DIR, 'logs')
 const LOCK_PATH = join(STATE_DIR, 'run.lock')
 const SPEAKERS_PATH = join(CONFIG_DIR, 'speakers.json')
 const CONFIG_ENV_PATH = join(CONFIG_DIR, 'config.json')
-const PI_AUTH_PATH = join(os.homedir(), '.pi', 'agent', 'auth.json')
+// pi keeps credentials in its config dir, which PI_CODING_AGENT_DIR relocates.
+// Point it at a voicenote-owned directory to get an auth.json that only the
+// pipeline reads and refreshes: an interactive pi session rewrites its own
+// auth.json wholesale on exit and has already dropped entries that way.
+// Resolved per call — the env is hydrated from config.json after module load.
+function piAuthPath(): string {
+  const dir = process.env.PI_CODING_AGENT_DIR || join(os.homedir(), '.pi', 'agent')
+  return join(expandHome(dir), 'auth.json')
+}
 
 const AUDIO_EXTENSIONS = new Set(['.mp3', '.wav', '.m4a', '.wma', '.aac', '.flac'])
 
@@ -125,7 +133,9 @@ const ENV_KEYS = [
   'VOLCANO_TOS_KEEP',
   'VOICENOTE_PI_BIN',
   'VOICENOTE_PI_CLI',
+  'PI_CODING_AGENT_DIR',
   'VOICENOTE_FFPROBE_BIN',
+  'VOICENOTE_PI_MODEL',
   'VOICENOTE_PI_THINKING',
   'VOICENOTE_PI_SUMMARY_TOOLS',
   'VOICENOTE_CONTEXT_DIR',
@@ -174,6 +184,14 @@ function systemProxyUrl(): string | null {
   } catch { return null }
 }
 
+// Bun's child_process does NOT hand a child the proxy variables this process set
+// on process.env (http_proxy/https_proxy/no_proxy and their uppercase forms are
+// special-cased internally; only all_proxy survives). Every spawn that must reach
+// the network through the proxy has to pass them explicitly, so record them here.
+// Without this, a scheduler run whose proxy comes from config.json rather than a
+// real shell env leaves pi with no proxy at all — it fails with `fetch failed`.
+const childProxyEnv: Record<string, string> = {}
+
 function applyDerivedProxy(): void {
   const host = process.env.LOCAL_PROXY_HOST
   const port = process.env.LOCAL_PROXY_PORT
@@ -190,6 +208,7 @@ function applyDerivedProxy(): void {
     const needs = (k: string) => !process.env[k] || process.env[k]!.includes('${')
     for (const k of ['http_proxy', 'https_proxy', 'all_proxy', 'HTTP_PROXY', 'HTTPS_PROXY', 'ALL_PROXY']) {
       if (needs(k)) { process.env[k] = url; hydratedEnvKeys.add(k) } // derived, not real env
+      childProxyEnv[k] = process.env[k]!
     }
   }
   // no_proxy/NO_PROXY: seed a base when a proxy is active, then always merge
@@ -200,6 +219,7 @@ function applyDerivedProxy(): void {
   for (const k of ['no_proxy', 'NO_PROXY'] as const) {
     const r = deriveNoProxy(process.env[k], premergeRealNoProxy[k], hydratedEnvKeys.has(k), !!url, baseNoProxy, VOLCANO_NO_PROXY_HOSTS)
     process.env[k] = r.runtime
+    childProxyEnv[k] = r.runtime
     if (r.hydrate) hydratedEnvKeys.add(k)
     if (r.capture !== undefined) premergeRealNoProxy[k] = r.capture
   }
@@ -1308,15 +1328,16 @@ function piInvocation(args: string[]): { bin: string; args: string[] } {
 // ───────────────────────────────────────────────────────────────────────
 
 async function persistPiOAuth(providerId: string, creds: Record<string, unknown>): Promise<void> {
-  await mkdir(dirname(PI_AUTH_PATH), { recursive: true })
+  const authPath = piAuthPath()
+  await mkdir(dirname(authPath), { recursive: true })
   let existing: Json = {}
-  if (existsSync(PI_AUTH_PATH)) {
-    try { existing = JSON.parse(await readFile(PI_AUTH_PATH, 'utf8')) as Json } catch (e) { warnSideEffect(`parse ${PI_AUTH_PATH}`, e) }
+  if (existsSync(authPath)) {
+    try { existing = JSON.parse(await readFile(authPath, 'utf8')) as Json } catch (e) { warnSideEffect(`parse ${authPath}`, e) }
   }
   existing[providerId] = { type: 'oauth', ...creds }
-  const tmp = `${PI_AUTH_PATH}.tmp-${process.pid}`
+  const tmp = `${authPath}.tmp-${process.pid}`
   await writeFile(tmp, JSON.stringify(existing, null, 2) + '\n', { mode: 0o600 })
-  await rename(tmp, PI_AUTH_PATH)
+  await rename(tmp, authPath)
 }
 
 async function loginChatGPT(opts: { json?: boolean; deviceCode?: boolean; emit?: (o: Record<string, unknown>) => void }): Promise<void> {
@@ -1366,7 +1387,7 @@ async function loginChatGPT(opts: { json?: boolean; deviceCode?: boolean; emit?:
     }
     await persistPiOAuth(oauth.openaiCodexOAuthProvider.id, creds)
     if (json) emit({ event: 'success', provider: oauth.openaiCodexOAuthProvider.id })
-    else console.log(`\n✓ Signed in. Credentials saved to ${PI_AUTH_PATH}. Verify with: vn doctor`)
+    else console.log(`\n✓ Signed in. Credentials saved to ${piAuthPath()}. Verify with: vn doctor`)
   } catch (e: any) {
     let message = String(e?.message || e)
     if (/unsupported_country_region_territory|\b403\b/.test(message)) {
@@ -1503,20 +1524,23 @@ async function runPi(opts: {
   appendSystemPrompt?: string
   cwd?: string  // agent working dir: the knowledge base, so read/grep/find default there
 }): Promise<string> {
-  // No --provider/--model: pi's own configuration picks the model and credentials.
   const args = [
     '-p',
     '--mode', 'text',
     '--no-extensions', '--no-skills', '--no-context-files', '--no-session', '--no-prompt-templates', '--no-themes',
     '--system-prompt', opts.systemPrompt,
   ]
+  // Unset means pi's own default model and provider. There is no second
+  // provider to fall back to either way.
+  const model = piSummaryModel()
+  if (model) args.push('--model', model)
   if (opts.thinking) args.push('--thinking', opts.thinking)
   if (opts.tools && opts.tools.trim()) args.push('--tools', opts.tools.trim())
   else args.push('--no-tools')
   if (opts.appendSystemPrompt) args.push('--append-system-prompt', opts.appendSystemPrompt)
   return new Promise<string>((resolve, reject) => {
     const inv = piInvocation(args)
-    const child = spawn(inv.bin, inv.args, { stdio: ['pipe', 'pipe', 'pipe'], cwd: opts.cwd, windowsHide: true })
+    const child = spawn(inv.bin, inv.args, { stdio: ['pipe', 'pipe', 'pipe'], cwd: opts.cwd, windowsHide: true, env: { ...process.env, ...childProxyEnv } })
     let stdout = '', stderr = ''
     const timer = opts.timeoutMs ? setTimeout(() => child.kill('SIGKILL'), opts.timeoutMs) : null
     child.stdout.on('data', d => stdout += String(d))
@@ -1562,6 +1586,12 @@ async function chatCompleteViaPi(opts: Parameters<typeof runPi>[0]): Promise<str
   }
 }
 
+// pi's --model accepts "provider/id" (e.g. openai-codex/gpt-5.6-sol), so this one
+// setting pins both. Empty/unset = whatever pi is configured to use.
+function piSummaryModel(): string {
+  return (process.env.VOICENOTE_PI_MODEL || '').trim()
+}
+
 function piThinkingLevel(): string {
   return process.env.VOICENOTE_PI_THINKING || 'high'
 }
@@ -1758,7 +1788,7 @@ async function processRecording(config: Config, rec: Recording, opts: any): Prom
 
   let summaryError: any = null
   if (needsNotes) {
-    progressStep(nextStep(), totalSteps, 'Generate integrated semantic notes', 'via pi (pi\'s own provider/model config)')
+    progressStep(nextStep(), totalSteps, 'Generate integrated semantic notes', `via pi, model=${piSummaryModel() || "pi's own default"}`)
     try {
       meta = await withHeartbeat('generate integrated semantic notes', () => summarizeTranscript(config, transcript, rec, files.audio), 60)
     } catch (e: any) {
@@ -2688,8 +2718,8 @@ async function collectDoctor() {
       : { configured: false as const },
     // Provider/model/credentials are pi's own configuration; `pi.available` is
     // all we can honestly report about whether a summary can run.
-    summary: { backend: 'pi', thinking: piThinkingLevel(), tools: tools || null, contextDir: tools ? summaryContextDir(config) : null },
-    pi: { bin: piCodexBin(), version: piCheck.code === 0 ? (piCheck.stdout.trim() || piCheck.stderr.trim() || null) : null, available: piCheck.code === 0, auth: existsSync(PI_AUTH_PATH) },
+    summary: { backend: 'pi', model: piSummaryModel() || null, thinking: piThinkingLevel(), tools: tools || null, contextDir: tools ? summaryContextDir(config) : null },
+    pi: { bin: piCodexBin(), version: piCheck.code === 0 ? (piCheck.stdout.trim() || piCheck.stderr.trim() || null) : null, available: piCheck.code === 0, auth: existsSync(piAuthPath()), authPath: piAuthPath() },
     // Outbound proxy for HTTPS endpoints (updater/GitHub): honor the standard
     // env chain, not just lowercase http_proxy — an https_proxy-only setup must
     // still route the updater.
@@ -2755,15 +2785,15 @@ async function doctor(opts: { json?: boolean } = {}): Promise<void> {
   } else {
     console.log(`volcano=not configured`)
   }
-  console.log(`summaryBackend=${s.summary.backend} (provider/model come from pi's own config)`)
-  console.log(`pi.bin=${s.pi.bin}`)
+  console.log(`summaryBackend=${s.summary.backend}`)
+  console.log(`pi.bin=${s.pi.bin} model=${s.summary.model || "<pi's own default>"}`)
   console.log(`pi.thinking=${s.summary.thinking}`)
   console.log(`pi.summaryTools=${s.summary.tools || '<disabled>'}`)
   if (s.summary.contextDir) console.log(`pi.contextDir=${s.summary.contextDir} (summary agent cwd + read/grep cross-reference root)`)
   console.log(`pi.version=${s.pi.version || 'missing'}`)
   // Neutral fact, not an instruction: an API-key user has no auth.json and needs
   // nothing fixed.
-  console.log(`pi.auth=${s.pi.auth ? 'logged-in (auth.json present)' : 'no ~/.pi/agent/auth.json (fine if a provider API key is set)'}`)
+  console.log(`pi.auth=${s.pi.authPath} ${s.pi.auth ? '(present)' : '(missing — fine if a provider API key is set)'}`)
   console.log(`defaultMode=notes`)
   console.log(`proxy=${s.proxy.url || '<unset>'}`)
   console.log(`speakers.self=${s.identity.self || '<unset>'}`)