pi-io-provider 1.1.57 → 1.1.59

package/AGENTS.md

@@ -19,6 +19,7 @@ When a model needs overrides, new properties, or corrections, edit the appropria
 | `patch.json` | Per-model overrides keyed by model ID. Add reasoning flags, compat settings, pricing corrections, thinking level maps, etc. Applied on top of `models.json` at runtime and for README generation. |
 | `custom-models.json` | Models that don't exist in the provider API (hidden models, router endpoints, cross-provider aliases). Merged after patch. Format: array of full model objects (same schema as `models.json` entries). |
 | `index.ts` | Provider extension code. |
+| `sync-policy.ts` | Quiet startup sync policy (TTL, key fingerprint, cross-process lock). Provider-agnostic and shared verbatim across pi-*-provider repos — keep it free of provider specifics. |
 | `scripts/update-models.js` | The sync script itself (edit only if changing how models are fetched/transformed). |
 
 ## Data Flow

package/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## 1.1.59
+
+- Quiet startup sync: background `/api/v1/models` revalidation runs only with an API key, only when the disk cache is older than 1 hour (`PI_PROVIDER_SYNC_TTL_MS`, `0` = always), missing, or fetched with a different key, and only in one pi process at a time (lock file next to the cache). Failed refreshes back off for 5 minutes; sessions that skip adopt a cache refreshed by another process.
+- Never store the API key: key changes are detected via a salted sha256 fingerprint in `cache/io-intelligence-models.sync.json`. The cache file format is unchanged.
+- Write the model cache atomically and swallow API key resolution failures in `session_start`.
+- Add vitest coverage for the sync policy and the session_start integration.
+
 ## 1.1.48
 
 - Pin Pi SDK development dependencies to 1.0.0 while retaining wildcard host peers.

package/README.md

@@ -11,50 +11,51 @@ A [pi](https://github.com/badlogic/pi-mono) extension that adds [IO Intelligence
 - **Prompt caching** — Cache read/write support on most models
 - **Confidential inference** — Verifiable TEE inference with attestation (via /private/ endpoints)
 - **Streaming** — Real-time token streaming
+- **Quiet live model sync** — Models refresh from the IO Intelligence API in the background, at most hourly and by one pi process at a time ([details](#model-sync))
 
 ## Available Models
 
 | Model | ID | Context | Max Output | Vision | Reasoning | Cache | Input $/M | Output $/M |
 |-------|----|---------|------------|--------|-----------|-------|-----------|------------|
-| DeepSeek V4.1 Flash | `deepseek-ai/DeepSeek-V4.1-Flash` | 1.0M | 131K | ✅ | ✅ | ✅ | $0.29 | $1.19 |
-| Gemma 4 26B A4B | `google/gemma-4-26b-a4b-it` | 262K | 262K | ✅ | ✅ | ✅ | $0.09 | $0.33 |
+| DeepSeek V4.1 Flash | `deepseek-ai/DeepSeek-V4.1-Flash` | 1.0M | 131K | ✅ | ✅ | ✅ | $0.31 | $1.26 |
+| Gemma 4 26B A4B | `google/gemma-4-26b-a4b-it` | 262K | 262K | ✅ | ✅ | ✅ | $0.09 | $0.32 |
 | Gemma 4 31B Instruct | `google/gemma-4-31b-it` | 262K | 16K | ✅ | ✅ | ✅ | $0.38 | $1.15 |
 | Kimi K2.5 | `moonshotai/Kimi-K2.5` | 262K | 262K | ✅ | ✅ | ✅ | $0.53 | $2.79 |
-| Kimi K2.6 | `moonshotai/Kimi-K2.6` | 262K | 262K | ✅ | ✅ | ✅ | $0.80 | $3.68 |
-| Kimi K2.7 Code | `moonshotai/Kimi-K2.7-Code` | 262K | 262K | ✅ | ✅ | ✅ | $1.08 | $4.67 |
+| Kimi K2.6 | `moonshotai/Kimi-K2.6` | 262K | 262K | ✅ | ✅ | ✅ | $0.85 | $3.78 |
+| Kimi K2.7 Code | `moonshotai/Kimi-K2.7-Code` | 262K | 262K | ✅ | ✅ | ✅ | $1.03 | $4.54 |
 | Kimi K3 | `moonshotai/Kimi-K3` | 1.0M | 1.0M | ✅ | ✅ | ✅ | $3.66 | $17.80 |
-| Qwen3.6 27B | `Qwen/Qwen3.6-27B` | 33K | 33K | ✅ | ✅ | ✅ | $0.34 | $3.12 |
-| Qwen3.6 35B A3B | `Qwen/Qwen3.6-35B-A3B` | 262K | 262K | ✅ | ✅ | ✅ | $0.15 | $1.06 |
-| Qwen3.8 27B | `Qwen/Qwen3.8-27B` | 66K | 66K | ✅ | ✅ | ✅ | $0.50 | $3.09 |
+| Qwen3.6 27B | `Qwen/Qwen3.6-27B` | 33K | 33K | ✅ | ✅ | ✅ | $0.34 | $2.87 |
+| Qwen3.6 35B A3B | `Qwen/Qwen3.6-35B-A3B` | 262K | 262K | ✅ | ✅ | ✅ | $0.16 | $1.04 |
+| Qwen3.8 27B | `Qwen/Qwen3.8-27B` | 66K | 66K | ✅ | ✅ | ✅ | $0.54 | $3.19 |
 | MiMo-V2.6-Flash | `XiaomiMiMo/MiMo-V2.6-Flash` | 1.0M | 131K | ✅ | ✅ | ✅ | $0.14 | $0.29 |
 | GLM 5.3 Flash | `zai-org/GLM-5.3-Flash` | 262K | 131K | ✅ | ✅ | ✅ | $0.17 | $0.57 |
 | DeepSeek R1 0528 | `deepseek-ai/DeepSeek-R1-0528` | 128K | 128K | ❌ | ✅ | ✅ | $0.57 | $2.28 |
 | DeepSeek V3.2 | `deepseek-ai/DeepSeek-V3.2` | 164K | 164K | ❌ | ✅ | ✅ | $1.38 | $2.18 |
 | DeepSeek V4 Flash | `deepseek-ai/DeepSeek-V4-Flash` | 33K | 33K | ❌ | ✅ | ✅ | $0.19 | $0.74 |
 | DeepSeek V4 Flash 0731 | `deepseek-ai/DeepSeek-V4-Flash-0731` | 262K | 66K | ❌ | ✅ | ✅ | $0.19 | $0.48 |
-| DeepSeek V4 Pro | `deepseek-ai/DeepSeek-V4-Pro` | 1.0M | 600K | ❌ | ✅ | ✅ | $1.09 | $4.85 |
+| DeepSeek V4 Pro | `deepseek-ai/DeepSeek-V4-Pro` | 1.0M | 600K | ❌ | ✅ | ✅ | $1.25 | $4.91 |
 | DeepSeek V4 Pro 0813 | `deepseek-ai/DeepSeek-V4-Pro-0813` | 1.0M | 393K | ❌ | ✅ | ✅ | $1.13 | $3.97 |
-| MiniMax M2.5 | `MiniMaxAI/MiniMax-M2.5` | 197K | 197K | ❌ | ✅ | ✅ | $0.35 | $1.42 |
-| MiniMax M2.7 | `MiniMaxAI/MiniMax-M2.7` | 262K | 66K | ❌ | ✅ | ✅ | $0.40 | $1.46 |
+| MiniMax M2.5 | `MiniMaxAI/MiniMax-M2.5` | 197K | 197K | ❌ | ✅ | ✅ | $0.36 | $1.44 |
+| MiniMax M2.7 | `MiniMaxAI/MiniMax-M2.7` | 262K | 66K | ❌ | ✅ | ✅ | $0.40 | $1.49 |
 | Kimi K2 Thinking | `moonshotai/Kimi-K2-Thinking` | 262K | 262K | ❌ | ✅ | ✅ | $0.60 | $2.50 |
-| Nemotron 3.5 Lightning | `nvidia/nemotron-3.5-lightning` | 262K | 131K | ❌ | ✅ | ✅ | $0.06 | $0.18 |
-| gpt-oss-120b | `openai/gpt-oss-120b` | 131K | 131K | ❌ | ✅ | ✅ | $0.17 | $0.63 |
-| gpt-oss-20b | `openai/gpt-oss-20b` | 64K | 64K | ❌ | ✅ | ✅ | $0.06 | $0.20 |
+| Nemotron 3.5 Lightning | `nvidia/nemotron-3.5-lightning` | 262K | 131K | ❌ | ✅ | ✅ | $0.06 | $0.19 |
+| gpt-oss-120b | `openai/gpt-oss-120b` | 131K | 131K | ❌ | ✅ | ✅ | $0.18 | $0.68 |
+| gpt-oss-20b | `openai/gpt-oss-20b` | 64K | 64K | ❌ | ✅ | ✅ | $0.06 | $0.19 |
 | MiMo-V2.5 | `XiaomiMiMo/MiMo-V2.5` | 1.0M | 1.0M | ❌ | ✅ | ✅ | $0.20 | $0.64 |
 | GLM-4.5-Air | `zai-org/GLM-4.5-Air` | 131K | 131K | ❌ | ✅ | ✅ | $0.16 | $0.94 |
 | GLM 4.6 | `zai-org/GLM-4.6` | 131K | 131K | ❌ | ✅ | ✅ | $0.52 | $2.04 |
 | GLM 4.7 | `zai-org/GLM-4.7` | 203K | 203K | ❌ | ✅ | ✅ | $0.53 | $2.07 |
 | GLM 4.7 Flash | `zai-org/GLM-4.7-Flash` | 200K | 200K | ❌ | ✅ | ✅ | $0.06 | $0.40 |
-| GLM 5 | `zai-org/GLM-5` | 203K | 203K | ❌ | ✅ | ✅ | $0.93 | $2.88 |
-| GLM 5.1 | `zai-org/GLM-5.1` | 203K | 33K | ❌ | ✅ | ✅ | $1.37 | $4.31 |
-| GLM 5.2 | `zai-org/GLM-5.2` | 262K | 131K | ❌ | ✅ | ✅ | $1.60 | $7.52 |
-| GLM 5.3 | `zai-org/GLM-5.3` | 262K | 131K | ❌ | ✅ | ✅ | $1.25 | $4.71 |
+| GLM 5 | `zai-org/GLM-5` | 203K | 203K | ❌ | ✅ | ✅ | $0.86 | $2.75 |
+| GLM 5.1 | `zai-org/GLM-5.1` | 203K | 33K | ❌ | ✅ | ✅ | $1.39 | $4.36 |
+| GLM 5.2 | `zai-org/GLM-5.2` | 262K | 131K | ❌ | ✅ | ✅ | $1.95 | $6.40 |
+| GLM 5.3 | `zai-org/GLM-5.3` | 262K | 131K | ❌ | ✅ | ✅ | $1.33 | $4.73 |
 | Llama 4 Maverick 17B 128E Instruct FP8 | `meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8` | 430K | 430K | ✅ | ❌ | ✅ | $0.27 | $0.83 |
-| Qwen3 Coder 480B A35B Instruct INT4 Mixed AR | `Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar` | 106K | 106K | ❌ | ❌ | ✅ | $0.45 | $2.15 |
-| Llama 3.3 70B Instruct | `meta-llama/Llama-3.3-70B-Instruct` | 128K | 128K | ❌ | ❌ | ✅ | $0.44 | $0.98 |
+| Qwen3 Coder 480B A35B Instruct INT4 Mixed AR | `Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar` | 106K | 106K | ❌ | ❌ | ✅ | $0.46 | $2.29 |
+| Llama 3.3 70B Instruct | `meta-llama/Llama-3.3-70B-Instruct` | 128K | 128K | ❌ | ❌ | ✅ | $0.55 | $1.02 |
 | Mistral Nemo Instruct 2407 | `mistralai/Mistral-Nemo-Instruct-2407` | 128K | 128K | ❌ | ❌ | ✅ | $0.05 | $0.08 |
 | Kimi K2 Instruct 0905 | `moonshotai/Kimi-K2-Instruct-0905` | 262K | 262K | ❌ | ❌ | ✅ | $0.57 | $2.30 |
-| Qwen3 Next 80B A3B Instruct | `Qwen/Qwen3-Next-80B-A3B-Instruct` | 262K | 262K | ❌ | ❌ | ✅ | $0.12 | $1.14 |
+| Qwen3 Next 80B A3B Instruct | `Qwen/Qwen3-Next-80B-A3B-Instruct` | 262K | 262K | ❌ | ❌ | ✅ | $0.11 | $1.04 |
 
 *Costs are per million tokens. Cache read/write pricing available on most models.*
 
@@ -122,6 +123,7 @@ Get your API key from [io.net](https://io.net).
 | Variable | Required | Description |
 |----------|----------|-------------|
 | `IOINTELLIGENCE_API_KEY` | No | Your IO Intelligence API key (fallback if not in auth.json) |
+| `PI_PROVIDER_SYNC_TTL_MS` | No | Model sync freshness window in ms (default `3600000`; `0` = revalidate every session start) |
 
 ## Configuration
 
@@ -145,6 +147,20 @@ Once loaded, select a model with:
 
 Or use `/models` to browse all available IO Intelligence models.
 
+## Model sync
+
+Startup never waits on the network. pi starts with the cached catalog (`~/.pi/agent/cache/io-intelligence-models.json`,
+falling back to the bundled `models.json`) and revalidates against `/api/v1/models` in the background — only when:
+
+- an IO Intelligence API key is configured (no key, no request),
+- the cache is older than 1 hour, missing, or was fetched with a different API key, and
+- no other pi process is already refreshing it (an `io-intelligence-models.lock` file next to the cache).
+
+So `/new`, `/reload`, `pi -p` and fan-out child agents share one cache instead of each calling the API; a
+session that skips the fetch still picks up a cache another process refreshed. A failed refresh is retried after
+5 minutes at the earliest. `PI_PROVIDER_SYNC_TTL_MS` changes the window (`0` revalidates on every session start).
+Only a hash fingerprint of the key is stored (`io-intelligence-models.sync.json`), never the key itself.
+
 ## API Documentation
 
 - IO Intelligence Docs: https://io.net/docs/guides/confidential-inference/quick-start

package/index.ts

@@ -7,6 +7,9 @@
  * Model resolution strategy: Stale-While-Revalidate
  *   1. Serve stale immediately: disk cache → embedded models.json (zero-latency)
  *   2. Revalidate in background: live API /models → merge with embedded → cache → hot-swap
+ *      Quiet startup sync (sync-policy.ts): only with an API key, only when the
+ *      disk cache is older than the TTL (1h, PI_PROVIDER_SYNC_TTL_MS) or the key
+ *      changed, and only one pi process at a time (lock file next to the cache).
  *   3. patch.json + custom-models.json applied on top of whichever source won
  *
  * Merge order: [live|cache|embedded] → apply patch.json → merge custom-models.json
@@ -43,6 +46,7 @@ import patchData from "./patch.json" with { type: "json" };
 import deprecatedData from "./deprecated-models.json" with { type: "json" };
 import fs from "fs";
 import path from "path";
+import { acquireModelSync, cacheChangedSinceLoad, markCacheLoaded, writeFileAtomic } from "./sync-policy";
 
 // ─── Types ────────────────────────────────────────────────────────────────────
 
@@ -227,7 +231,8 @@ function loadCachedModels(): JsonModel[] | null {
 function cacheModels(models: JsonModel[]): void {
   try {
     fs.mkdirSync(CACHE_DIR, { recursive: true });
-    fs.writeFileSync(CACHE_PATH, JSON.stringify(models, null, 2) + "\n");
+    // Atomic (tmp + rename): other pi processes may be reading the cache right now.
+    writeFileAtomic(CACHE_PATH, JSON.stringify(models, null, 2) + "\n");
   } catch {
     // Cache write failure is non-fatal
   }
@@ -300,6 +305,7 @@ function withDeprecated(models: JsonModel[]): JsonModel[] {
 }
 
 function loadStaleModels(embeddedModels: JsonModel[]): JsonModel[] {
+  markCacheLoaded(CACHE_PATH); // stat before read: lets session_start spot newer caches
   const cached = loadCachedModels();
   if (!cached || cached.length === 0) return embeddedModels;
 
@@ -313,13 +319,26 @@ function loadStaleModels(embeddedModels: JsonModel[]): JsonModel[] {
   return cached;
 }
 
-async function revalidateModels(apiKey: string | undefined, embeddedModels: JsonModel[], signal?: AbortSignal): Promise<JsonModel[] | null> {
-  if (!apiKey) return null;
-  const liveModels = await fetchLiveModels(apiKey, signal);
-  if (!liveModels || liveModels.length === 0) return null;
-  const merged = mergeWithEmbedded(liveModels, embeddedModels);
-  cacheModels(merged);
-  return merged;
+/**
+ * Background revalidation, gated by the quiet startup sync policy
+ * (sync-policy.ts): no key → no fetch; fresh cache (TTL) → no fetch; another
+ * process already fetching → no fetch. When skipped, a cache that another
+ * process refreshed since we loaded it is adopted from disk instead.
+ * `force` bypasses the TTL (e.g. after /login); it is still single-flight.
+ */
+async function revalidateModels(apiKey: string | undefined, embeddedModels: JsonModel[], signal?: AbortSignal, force = false): Promise<JsonModel[] | null> {
+  const lease = acquireModelSync({ cachePath: CACHE_PATH, apiKey, force });
+  if (!lease) return cacheChangedSinceLoad(CACHE_PATH) ? loadStaleModels(embeddedModels) : null;
+  let merged: JsonModel[] | null = null;
+  try {
+    const liveModels = apiKey ? await fetchLiveModels(apiKey, signal) : null;
+    if (!liveModels || liveModels.length === 0) return null;
+    merged = mergeWithEmbedded(liveModels, embeddedModels);
+    cacheModels(merged);
+    return merged;
+  } finally {
+    lease.release(merged ? "fetched" : signal?.aborted ? "aborted" : "failed");
+  }
 }
 
 // ─── API Key Resolution (via ModelRegistry) ────────────────────────────────────
@@ -348,12 +367,15 @@ export default function (pi: ExtensionAPI) {
     models: staleModels,
   });
 
+  // Fire-and-forget: never await network here — startup must not block on the
+  // provider API. Nothing is awaited and every rejection is swallowed.
   pi.on("session_start", async (_event, ctx) => {
     revalidateAbort?.abort();
     revalidateAbort = new AbortController();
     const signal = revalidateAbort.signal;
-    resolveApiKey(ctx.modelRegistry).then(() => {
-      revalidateModels(cachedApiKey, embeddedModels, signal).then((freshBase) => {
+    void resolveApiKey(ctx.modelRegistry)
+      .then(() => revalidateModels(cachedApiKey, embeddedModels, signal))
+      .then((freshBase) => {
         if (freshBase && !signal.aborted) {
           pi.registerProvider("io-intelligence", {
             baseUrl: BASE_URL,
@@ -362,8 +384,8 @@ export default function (pi: ExtensionAPI) {
             models: buildModels(freshBase, customModels, patches),
           });
         }
-      });
-    });
+      })
+      .catch(() => { /* keep serving the stale catalog */ });
   });
 
   pi.on("session_shutdown", () => {

package/models.json

@@ -31,9 +31,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.294,
-      "output": 1.19,
-      "cacheRead": 0.147,
+      "input": 0.313,
+      "output": 1.26,
+      "cacheRead": 0.1565,
       "cacheWrite": 0
     },
     "contextWindow": 1048320,
@@ -76,9 +76,9 @@
       "text"
     ],
     "cost": {
-      "input": 1.246,
-      "output": 4.708,
-      "cacheRead": 0.623,
+      "input": 1.32848,
+      "output": 4.72954,
+      "cacheRead": 0.66424,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
@@ -99,9 +99,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.4978,
-      "output": 3.088,
-      "cacheRead": 0.2489,
+      "input": 0.5378,
+      "output": 3.188,
+      "cacheRead": 0.2689,
       "cacheWrite": 0
     },
     "contextWindow": 65536,
@@ -188,9 +188,9 @@
       "text"
     ],
     "cost": {
-      "input": 1.6004,
-      "output": 7.52,
-      "cacheRead": 0.8002,
+      "input": 1.95,
+      "output": 6.4,
+      "cacheRead": 0.975,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
@@ -211,9 +211,9 @@
       "image"
     ],
     "cost": {
-      "input": 1.0844,
-      "output": 4.668,
-      "cacheRead": 0.5422,
+      "input": 1.02864,
+      "output": 4.538,
+      "cacheRead": 0.51432,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
@@ -234,9 +234,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.1472,
-      "output": 1.06275,
-      "cacheRead": 0.0736,
+      "input": 0.1572,
+      "output": 1.04275,
+      "cacheRead": 0.0786,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
@@ -257,9 +257,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.343,
-      "output": 3.12,
-      "cacheRead": 0.1715,
+      "input": 0.341,
+      "output": 2.87,
+      "cacheRead": 0.1705,
       "cacheWrite": 0
     },
     "contextWindow": 32768,
@@ -279,9 +279,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.396,
-      "output": 1.464,
-      "cacheRead": 0.198,
+      "input": 0.402,
+      "output": 1.488,
+      "cacheRead": 0.201,
       "cacheWrite": 0
     },
     "contextWindow": 262100,
@@ -323,9 +323,9 @@
       "text"
     ],
     "cost": {
-      "input": 1.08814,
-      "output": 4.8536,
-      "cacheRead": 0.54407,
+      "input": 1.25206,
+      "output": 4.9124,
+      "cacheRead": 0.62603,
       "cacheWrite": 0
     },
     "contextWindow": 1048576,
@@ -346,9 +346,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.8,
-      "output": 3.682,
-      "cacheRead": 0.4,
+      "input": 0.85,
+      "output": 3.782,
+      "cacheRead": 0.425,
       "cacheWrite": 0
     },
     "contextWindow": 262142,
@@ -368,9 +368,9 @@
       "text"
     ],
     "cost": {
-      "input": 1.37228,
-      "output": 4.31288,
-      "cacheRead": 0.68614,
+      "input": 1.38628,
+      "output": 4.35688,
+      "cacheRead": 0.69314,
       "cacheWrite": 0
     },
     "contextWindow": 202750,
@@ -390,9 +390,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.354,
-      "output": 1.416,
-      "cacheRead": 0.177,
+      "input": 0.359,
+      "output": 1.44,
+      "cacheRead": 0.1795,
       "cacheWrite": 0
     },
     "contextWindow": 196600,
@@ -435,9 +435,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.93,
-      "output": 2.878,
-      "cacheRead": 0.465,
+      "input": 0.86,
+      "output": 2.752,
+      "cacheRead": 0.43,
       "cacheWrite": 0
     },
     "contextWindow": 202752,
@@ -524,9 +524,9 @@
       "image"
     ],
     "cost": {
-      "input": 0.094,
-      "output": 0.33,
-      "cacheRead": 0.047,
+      "input": 0.0893,
+      "output": 0.321,
+      "cacheRead": 0.04465,
       "cacheWrite": 0
     },
     "contextWindow": 262142,
@@ -611,9 +611,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.168,
-      "output": 0.63,
-      "cacheRead": 0.084,
+      "input": 0.178,
+      "output": 0.68,
+      "cacheRead": 0.089,
       "cacheWrite": 0.04
     },
     "contextWindow": 131072,
@@ -677,9 +677,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.1175,
-      "output": 1.136,
-      "cacheRead": 0.05875,
+      "input": 0.109375,
+      "output": 1.045,
+      "cacheRead": 0.054688,
       "cacheWrite": 0.12
     },
     "contextWindow": 262144,
@@ -698,9 +698,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.445,
-      "output": 2.145,
-      "cacheRead": 0.2225,
+      "input": 0.46125,
+      "output": 2.29375,
+      "cacheRead": 0.230625,
       "cacheWrite": 0.44
     },
     "contextWindow": 106000,
@@ -762,9 +762,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.057,
-      "output": 0.196,
-      "cacheRead": 0.0285,
+      "input": 0.055,
+      "output": 0.194,
+      "cacheRead": 0.0275,
       "cacheWrite": 0.03
     },
     "contextWindow": 64000,
@@ -784,9 +784,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.4446,
-      "output": 0.9846,
-      "cacheRead": 0.2223,
+      "input": 0.5466,
+      "output": 1.0226,
+      "cacheRead": 0.2733,
       "cacheWrite": 0.2
     },
     "contextWindow": 128000,
@@ -850,9 +850,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.05975,
-      "output": 0.185,
-      "cacheRead": 0.029875,
+      "input": 0.0578,
+      "output": 0.188,
+      "cacheRead": 0.0289,
       "cacheWrite": 0
     },
     "contextWindow": 262144,

package/package.json

@@ -1,9 +1,18 @@
 {
   "name": "pi-io-provider",
-  "version": "1.1.57",
+  "version": "1.1.59",
   "description": "IO Intelligence provider extension for pi - Access DeepSeek, Kimi, GLM, Llama, Qwen, Mistral, and more through the IO Intelligence API",
   "type": "module",
   "main": "index.ts",
+  "scripts": {
+    "clean": "echo 'nothing to clean'",
+    "build": "echo 'nothing to build'",
+    "check": "tsc --noEmit && vitest run && node scripts/test-pi-compat.mjs",
+    "test": "vitest run",
+    "update-models": "node scripts/update-models.js",
+    "test:pi": "node scripts/test-pi-compat.mjs",
+    "typecheck": "tsc --noEmit"
+  },
   "keywords": [
     "pi",
     "extension",
@@ -32,17 +41,10 @@
   "devDependencies": {
     "@earendil-works/pi-coding-agent": "1.1.0",
     "typescript": "6.0.3",
-    "@types/node": "^25.9.1"
+    "@types/node": "^25.9.1",
+    "vitest": "4.1.7"
   },
   "peerDependencies": {
     "@earendil-works/pi-coding-agent": "*"
-  },
-  "scripts": {
-    "clean": "echo 'nothing to clean'",
-    "build": "echo 'nothing to build'",
-    "check": "tsc --noEmit && node scripts/test-pi-compat.mjs",
-    "update-models": "node scripts/update-models.js",
-    "test:pi": "node scripts/test-pi-compat.mjs",
-    "typecheck": "tsc --noEmit"
   }
-}
+}

package/scripts/update-models.js

@@ -339,6 +339,7 @@ A [pi](https://github.com/badlogic/pi-mono) extension that adds [IO Intelligence
 - **Prompt caching** — Cache read/write support on most models
 - **Confidential inference** — Verifiable TEE inference with attestation (via /private/ endpoints)
 - **Streaming** — Real-time token streaming
+- **Quiet live model sync** — Models refresh from the IO Intelligence API in the background, at most hourly and by one pi process at a time ([details](#model-sync))
 
 ## Available Models
 
@@ -412,6 +413,7 @@ Get your API key from [io.net](https://io.net).
 | Variable | Required | Description |
 |----------|----------|-------------|
 | \`IOINTELLIGENCE_API_KEY\` | No | Your IO Intelligence API key (fallback if not in auth.json) |
+| \`PI_PROVIDER_SYNC_TTL_MS\` | No | Model sync freshness window in ms (default \`3600000\`; \`0\` = revalidate every session start) |
 
 ## Configuration
 
@@ -435,6 +437,20 @@ Once loaded, select a model with:
 
 Or use \`/models\` to browse all available IO Intelligence models.
 
+## Model sync
+
+Startup never waits on the network. pi starts with the cached catalog (\`~/.pi/agent/cache/io-intelligence-models.json\`,
+falling back to the bundled \`models.json\`) and revalidates against \`/api/v1/models\` in the background — only when:
+
+- an IO Intelligence API key is configured (no key, no request),
+- the cache is older than 1 hour, missing, or was fetched with a different API key, and
+- no other pi process is already refreshing it (an \`io-intelligence-models.lock\` file next to the cache).
+
+So \`/new\`, \`/reload\`, \`pi -p\` and fan-out child agents share one cache instead of each calling the API; a
+session that skips the fetch still picks up a cache another process refreshed. A failed refresh is retried after
+5 minutes at the earliest. \`PI_PROVIDER_SYNC_TTL_MS\` changes the window (\`0\` revalidates on every session start).
+Only a hash fingerprint of the key is stored (\`io-intelligence-models.sync.json\`), never the key itself.
+
 ## API Documentation
 
 - IO Intelligence Docs: https://io.net/docs/guides/confidential-inference/quick-start