@askalf/dario 4.8.70 → 4.8.72

package/README.md

@@ -16,7 +16,9 @@
 
 ---
 
-> ✨ **New — Claude Fable 5 fully supported.** `claude-fable-5`, its `[1m]` long-context variant, and the `fable` / `fable1m` shortcuts route on your subscription like any other model. dario tracked CC's new flagship the day it shipped, including the wire quirks Anthropic never documented: the `fallback-credit` beta fable *requires* (without it subscription fable traffic is silently soft-refused), its refusal of high effort levels, and the **per-model beta tailoring** matched to CC v2.1.170. `npm install -g @askalf/dario@latest`
+> 🗓️ **2026-06-15 — Anthropic splits Claude billing.** Agent-SDK and `claude -p` (headless) traffic stops counting against your subscription pool and moves to a small separate monthly credit ($20 / $100 / $200 by plan), then metered per-token API rates. Most proxies forward your requests in exactly the shape that gets reclassified into that bucket. dario rewrites every request into interactive Claude Code wire-shape before it leaves your machine, so your traffic stays in the subscription pool you already pay for — same install, no config change for the cliff. **[What changes, and how to verify it on your own machine →](#the-deadline-2026-06-15)**
+
+> ⛔ **Claude Fable 5 / Mythos 5 — temporarily suspended for all Anthropic customers.** On 2026-06-12 a US-government legal directive disabled Fable 5 and Mythos 5 for **every** Anthropic plan and tier — `api.anthropic.com` now returns `not_found` for them, and no account or proxy can route around it ([details](https://www.anthropic.com/news/fable-mythos-access)). As of **v4.8.71**, dario filters both families out of `/v1/models` and rejects any spelling (`fable`, `fable1m`, `claude-fable-5`, `claude-fable-5[1m]`, `claude:fable`) up front with a clean `404` pointing at `claude-opus-4-8` / `claude-sonnet-4-6` — instead of forwarding into a confusing upstream error. Reversible without a code change once access is restored: set `DARIO_SUSPENDED_MODELS=` (empty) on the instance. `npm install -g @askalf/dario@latest`
 >
 > ⚠️ Still on a version **before 4.8.39**? Upgrade now — those could silently corrupt code/structured content routed through the proxy (the identifier scrub stripped tokens like the JS `continue` keyword). **[Details →](https://github.com/askalf/dario/issues/457)**
 
@@ -73,10 +75,9 @@ Type `dario` with no args (in another terminal) to open a full-screen control pa
 │  Tokens out:      38,200               Subscription %:  98%         │
 │                                                                     │
 │  Per-model:                                                         │
-│   fable-5       ██████████░░░░░░░░░░░  50%  (124 req)               │
-│   opus-4-8      ██████░░░░░░░░░░░░░░░  30%  ( 74 req)               │
-│   sonnet-4-6    ███░░░░░░░░░░░░░░░░░░  14%  ( 34 req)               │
-│   haiku-4-5     █░░░░░░░░░░░░░░░░░░░░   6%  ( 15 req)               │
+│   opus-4-8      ████████████░░░░░░░░  60%  (148 req)               │
+│   sonnet-4-6    █████░░░░░░░░░░░░░░░  26%  ( 64 req)               │
+│   haiku-4-5     ███░░░░░░░░░░░░░░░░░  14%  ( 35 req)               │
 │                                                                     │
 │  Rate-limit:                                                        │
 │   5h  ████░░░░░░░░░░░░░░░░░░░░░░░░  18%                             │
@@ -86,18 +87,6 @@ Type `dario` with no args (in another terminal) to open a full-screen control pa
 
 ---
 
-## The money
-
-| Setup | Monthly cost — heavy user |
-|---|---|
-| Cursor + Anthropic API direct | **$80–$300** |
-| Multi-tool heavy use (Cursor + Aider + Cline + Continue), per-token | **$200–$600+** |
-| **Any of the above + dario** | **$20–$200 flat** — your existing Pro/Max plan, nothing extra |
-
-Switching providers is a model-name change, not a reconfigure. Add a backend once and the same `localhost:3456` speaks OpenAI, Groq, OpenRouter, or a local Ollama too.
-
----
-
 ## The deadline: 2026-06-15
 
 On **2026-06-15**, Anthropic splits Claude billing in two. Agentic traffic — Agent SDK, `claude -p` headless — stops counting against your subscription pool and gets a separate small monthly credit. [Announced 2026-05-13](https://support.claude.com/en/articles/15036540-use-the-claude-agent-sdk-with-your-claude-plan) via Claude's Help Center and a [@ClaudeDevs X post](https://x.com/ClaudeDevs/status/2054610152817619388) — no anthropic.com blog post, no email to most subscribers, no mention in CC release notes.
@@ -123,6 +112,18 @@ Same install, same `localhost:3456`, no config change for the cliff. Verify on y
 
 ---
 
+## The money
+
+| Setup | Monthly cost — heavy user |
+|---|---|
+| Cursor + Anthropic API direct | **$80–$300** |
+| Multi-tool heavy use (Cursor + Aider + Cline + Continue), per-token | **$200–$600+** |
+| **Any of the above + dario** | **$20–$200 flat** — your existing Pro/Max plan, nothing extra |
+
+Switching providers is a model-name change, not a reconfigure. Add a backend once and the same `localhost:3456` speaks OpenAI, Groq, OpenRouter, or a local Ollama too.
+
+---
+
 ## The principle dario operates on
 
 Two layers, separated:
@@ -166,6 +167,7 @@ dario doesn't *guess* Claude Code's request shape — it captures it live from y
 | `TodoWrite`/`TodoRead` replaced by the `Task*` family, no migration note | Clients hardcoding `todo_*` send unrecognized tools | [v3.38.6](https://github.com/askalf/dario/pull/274) — 13 min later |
 | **Claude Fable 5** ships as CC's flagship (v2.1.170) with undocumented wire quirks: a required `fallback-credit-2026-06-01` beta, soft-refusal of `max`/`xhigh` effort, and a `[1m]`-only `context-1m` beta | Subscription fable traffic 200-refused (empty body) without the flag; wrong effort silently returns nothing | [v4.8.46–52](https://github.com/askalf/dario/pulls?q=fable) — 2026-06-09 |
 | CC tailors the `anthropic-beta` set **per model** — opus 9, sonnet 8 (no `mid-conversation-system`), haiku 6 (also no `effort`) | Proxies sending one beta set diverge from CC for non-opus models | [v4.8.53](https://github.com/askalf/dario/pull/478) — 2026-06-09 |
+| **Fable 5 / Mythos 5 globally suspended** by a US-government directive (2026-06-12, every customer + tier) — `api.anthropic.com` now `not_found`s them | dario kept advertising `claude-fable-5[1m]` in `/v1/models` and forwarded fable traffic into a confusing upstream 404 | [v4.8.71](https://github.com/askalf/dario/pull/518) — 2026-06-13 (filtered out + clean local 404; reversible via `DARIO_SUSPENDED_MODELS`) |
 
 And it gets subtler: v4.2.1 (2026-05-17) shipped receipts for **same CC binary, different wire output 24 hours apart** — Anthropic ships changes through CC's *remote configuration*, not just npm releases. So dario runs **three classes of drift detection**, all auto-detecting and auto-PR'ing:
 
@@ -184,7 +186,7 @@ You point every tool at one URL. dario reads each request, decides which backend
 
 | Client speaks | Model | Routes to | What happens |
 |---|---|---|---|
-| Anthropic Messages | `claude-*` / `fable` / `opus` / `sonnet` / `haiku` | Claude backend | OAuth swap + CC template replay → `api.anthropic.com` |
+| Anthropic Messages | `claude-*` / `opus` / `sonnet` / `haiku` | Claude backend | OAuth swap + CC template replay → `api.anthropic.com` |
 | Anthropic Messages | `gpt-*`, `llama-*`, … | OpenAI-compat backend | Anthropic→OpenAI translation, forwarded |
 | OpenAI Chat | `gpt-*` / `o1-*` / `o3-*` | OpenAI-compat backend | Auth swap, body forwarded byte-for-byte |
 | OpenAI Chat | `claude-*` | Claude backend | OpenAI→Anthropic translation, then Claude path |
@@ -192,7 +194,7 @@ You point every tool at one URL. dario reads each request, decides which backend
 
 The tool doesn't know. The backend doesn't know. dario is the seam.
 
-**The full Claude lineup, autodetected.** Claude Fable 5 (CC's flagship), Opus 4.8, Sonnet 4.6, and Haiku 4.5 — plus `[1m]` long-context variants, generated by one rule for every family — by full id (`claude-fable-5`, `claude-opus-4-8`) or shortcut (`fable` / `opus` / `sonnet` / `haiku`, append `1m` for the long-context form). `GET /v1/models` asks Anthropic's live catalog (TTL-cached, baked fallback when offline), and the family shortcuts track it — a new model shows up and resolves the day it lands, no dario release needed; the model-specific wire shape (effort level, beta set, thinking config) is applied automatically.
+**The full Claude lineup, autodetected.** Opus 4.8, Sonnet 4.6, and Haiku 4.5 — plus `[1m]` long-context variants, generated by one rule for every family — by full id (`claude-opus-4-8`) or shortcut (`opus` / `sonnet` / `haiku`, append `1m` for the long-context form). `GET /v1/models` asks Anthropic's live catalog (TTL-cached, baked fallback when offline), and the family shortcuts track it — a new model shows up and resolves the day it lands, no dario release needed; the model-specific wire shape (effort level, beta set, thinking config) is applied automatically. **Globally-suspended families are filtered out** of both the live catalog and the baked fallback, so `/v1/models` never advertises a model that 404s upstream and suspended ids are rejected locally with an actionable error — currently Claude Fable 5 and Mythos 5 (US-government directive, 2026-06-12); governed by `DARIO_SUSPENDED_MODELS` (default `fable`), set it empty to re-enable when access is restored.
 
 ---
 
@@ -299,6 +301,9 @@ No. `five_hour` and `seven_day` are both subscription billing — different acco
 **Will the 2026-06-15 split break my setup? / What if Anthropic ships another silent change?**
 No, and it's caught automatically — see [The deadline](#the-deadline-2026-06-15) and [How it stays working](#how-it-works-and-how-it-stays-working). dario rewrites every request to interactive-CC shape before it reaches `api.anthropic.com`, and the three-class drift watcher picks up new changes (npm-release hourly, remote-config every 30 min, classifier-rule daily). v3.38.5 + v3.38.6 — 13 minutes apart, same day as v2.1.142's silent drops — are the prior art.
 
+**I'm getting `404 not_found` for `claude-fable-5` / `fable`. Did dario break?**
+No — Claude Fable 5 and Mythos 5 were disabled for **all** Anthropic customers by a US-government legal directive on 2026-06-12 (every plan and tier; [details](https://www.anthropic.com/news/fable-mythos-access)). `api.anthropic.com` returns `not_found` for them, so no subscription or proxy can route them. Since v4.8.71 dario drops both families from `/v1/models` and rejects them locally with a clear 404 pointing at `claude-opus-4-8` / `claude-sonnet-4-6`, instead of forwarding into a confusing upstream error. When access is restored, set `DARIO_SUSPENDED_MODELS=` (empty) on the instance to re-enable — no upgrade needed.
+
 Full FAQ: [`docs/faq.md`](./docs/faq.md)
 
 ---

package/dist/cc-template.d.ts

@@ -39,6 +39,10 @@ export declare const CC_TOOL_DEFINITIONS: {
     description: string;
     input_schema: Record<string, unknown>;
 }[];
+/** Lowercased CC tool name → canonical name, for identity-matching a CC client's
+ *  own tools (so newer built-ins not yet in TOOL_MAP map to themselves rather
+ *  than being treated as foreign and round-robined). Tracks the live bundle. */
+export declare const CC_NATIVE_LOWER: Map<string, string>;
 /** CC's static system prompt (~25KB). */
 export declare const CC_SYSTEM_PROMPT: string;
 /** CC's agent identity string. */

package/dist/cc-template.js

@@ -45,6 +45,10 @@ export function filterToolsForPlatform(tools, platform) {
 }
 /** CC's exact tool definitions for the current platform — filtered from the bundled union. */
 export const CC_TOOL_DEFINITIONS = filterToolsForPlatform(TEMPLATE.tools, process.platform);
+/** Lowercased CC tool name → canonical name, for identity-matching a CC client's
+ *  own tools (so newer built-ins not yet in TOOL_MAP map to themselves rather
+ *  than being treated as foreign and round-robined). Tracks the live bundle. */
+export const CC_NATIVE_LOWER = new Map(CC_TOOL_DEFINITIONS.map((t) => [String(t.name).toLowerCase(), String(t.name)]));
 /** CC's static system prompt (~25KB). */
 export const CC_SYSTEM_PROMPT = TEMPLATE.system_prompt;
 /** CC's agent identity string. */
@@ -1230,7 +1234,19 @@ export function buildCCRequest(clientBody, billingTag, cacheControl, identity, o
         const claimedCC = new Set();
         for (const tool of clientTools) {
             const name = (tool.name || '').toLowerCase();
-            const mapping = TOOL_MAP[name];
+            // A CC client's OWN tools map to THEMSELVES (identity). TOOL_MAP only
+            // carries the legacy core surface + cross-client aliases, so it lags CC's
+            // newer built-ins (Agent, AskUserQuestion, Cron*, Task*, NotebookEdit,
+            // Enter/ExitPlanMode, Workflow, …). Without this, those land "unmapped",
+            // get round-robined onto Read/Bash/etc., and the collision corrupts the
+            // calls of the tools that DID map (the dock saw Read's `file_path` arrive
+            // as `path`/`filePath`). Identity-matching against the live CC tool set
+            // (CC_NATIVE_LOWER, refreshed by every capture-and-bake) fixes it for
+            // every current-and-future CC client and keeps the canonical CC fingerprint.
+            const mapping = TOOL_MAP[name]
+                ?? (CC_NATIVE_LOWER.get(name)
+                    ? { ccTool: CC_NATIVE_LOWER.get(name), translateArgs: (a) => a, translateBack: (a) => a }
+                    : undefined);
             if (mapping) {
                 // In hybrid mode, clone the shared mapping and attach the
                 // client-declared top-level field names from input_schema.
@@ -1266,8 +1282,8 @@ export function buildCCRequest(clientBody, billingTag, cacheControl, identity, o
         const CC_FALLBACK_TOOLS = ['Bash', 'Read', 'Grep', 'Glob', 'WebSearch', 'WebFetch'];
         for (const tool of clientTools) {
             const name = (tool.name || '').toLowerCase();
-            if (TOOL_MAP[name])
-                continue;
+            if (TOOL_MAP[name] || CC_NATIVE_LOWER.has(name))
+                continue; // mapped, or a CC-native tool (identity in pass 1)
             unmappedTools.push(tool.name);
             if (opts.hybridTools)
                 continue; // dropped — see comment above

package/dist/model-catalog.d.ts

@@ -32,6 +32,12 @@
  * — the same ordering normalizeUpstreamIds() produces for live data.
  */
 export declare const BAKED_BASE_MODELS: readonly string[];
+/** The set of suspended families, resolved from env at call time. */
+export declare function suspendedFamilies(): Set<string>;
+/** True when `model` (any spelling) belongs to a suspended family. */
+export declare function isSuspendedModel(model: string): boolean;
+/** Drop suspended families from an advertised base list. */
+export declare function dropSuspendedModels(bases: readonly string[]): string[];
 /**
  * THE long-context rule — applied identically to every family. A base id
  * takes a `[1m]` variant unless it's the haiku family (CC's picker never

package/dist/model-catalog.js

@@ -43,6 +43,45 @@ export const BAKED_BASE_MODELS = [
     'claude-sonnet-4-6',
     'claude-haiku-4-5',
 ];
+/**
+ * Globally-suspended model families — dario neither advertises nor proxies
+ * them, even when upstream still lists the id, because api.anthropic.com
+ * returns `not_found` for every request to them.
+ *
+ * Claude Fable 5 AND Mythos 5 were disabled for ALL Anthropic customers by a
+ * US-government legal directive on 2026-06-12 (all plans/tiers; other models
+ * unaffected) — https://www.anthropic.com/news/fable-mythos-access. dario's
+ * baked fallback still carries `claude-fable-5`, and upstream may still list
+ * it, so without this filter dario keeps advertising a model that 404s.
+ *
+ * TEMP — reversible without a code change: `DARIO_SUSPENDED_MODELS` overrides
+ * the default (comma-separated families or model ids; each is normalized to
+ * its family). Set `DARIO_SUSPENDED_MODELS=` (empty) to re-enable everything
+ * once access is restored. Matching is by FAMILY, so every spelling is caught:
+ * `fable`, `fable1m`, `claude-fable-5`, `claude-fable-5[1m]`, `claude:fable`.
+ */
+const DEFAULT_SUSPENDED_MODELS = 'fable';
+/** The set of suspended families, resolved from env at call time. */
+export function suspendedFamilies() {
+    const raw = process.env.DARIO_SUSPENDED_MODELS ?? DEFAULT_SUSPENDED_MODELS;
+    return new Set(raw
+        .split(',')
+        .map((s) => s.trim())
+        .filter((s) => s.length > 0)
+        .map((s) => modelFamily(s) ?? s.toLowerCase()));
+}
+/** True when `model` (any spelling) belongs to a suspended family. */
+export function isSuspendedModel(model) {
+    const fams = suspendedFamilies();
+    if (fams.size === 0)
+        return false;
+    const fam = modelFamily(model);
+    return fam !== null && fams.has(fam);
+}
+/** Drop suspended families from an advertised base list. */
+export function dropSuspendedModels(bases) {
+    return bases.filter((b) => !isSuspendedModel(b));
+}
 /**
  * THE long-context rule — applied identically to every family. A base id
  * takes a `[1m]` variant unless it's the haiku family (CC's picker never
@@ -201,7 +240,7 @@ async function fetchUpstreamBases(deps) {
 async function refresh(deps) {
     const now = deps.now ?? Date.now;
     lastAttempt = now();
-    const bases = await fetchUpstreamBases(deps);
+    const bases = dropSuspendedModels(await fetchUpstreamBases(deps));
     cache = { bases, source: 'upstream', fetchedAt: now() };
     deps.log?.(`[dario] model catalog: autodetected ${bases.length} base models upstream`);
 }
@@ -242,7 +281,7 @@ export async function getModelCatalog(deps = {}) {
         }
     }
     if (cache === null)
-        cache = { bases: [...BAKED_BASE_MODELS], source: 'baked', fetchedAt: 0 };
+        cache = { bases: dropSuspendedModels(BAKED_BASE_MODELS), source: 'baked', fetchedAt: 0 };
     return cache;
 }
 /**
@@ -251,7 +290,7 @@ export async function getModelCatalog(deps = {}) {
  * Never blocks the hot path on the network.
  */
 export function getCachedBases() {
-    return cache?.bases ?? BAKED_BASE_MODELS;
+    return cache?.bases ?? dropSuspendedModels(BAKED_BASE_MODELS);
 }
 /** Fire-and-forget warmup so the first client /v1/models call is served warm. */
 export function prewarmModelCatalog(deps = {}) {

package/dist/proxy.js

@@ -17,7 +17,7 @@ import { loadAllAccounts, loadAccount, refreshAccountToken, resyncLoginFromCrede
 import { getOpenAIBackend, isOpenAIModel, forwardToOpenAI } from './openai-backend.js';
 import { RequestQueue, QueueFullError, QueueTimeoutError, DEFAULT_MAX_CONCURRENT, DEFAULT_MAX_QUEUED, DEFAULT_QUEUE_TIMEOUT_MS } from './request-queue.js';
 import { redactSecrets } from './redact.js';
-import { BAKED_BASE_MODELS, withLongContextVariants, buildOpenAIModelsList, getModelCatalog, getCachedBases, resolveAliasAgainst, prewarmModelCatalog } from './model-catalog.js';
+import { BAKED_BASE_MODELS, withLongContextVariants, buildOpenAIModelsList, getModelCatalog, getCachedBases, resolveAliasAgainst, prewarmModelCatalog, isSuspendedModel } from './model-catalog.js';
 const ANTHROPIC_API = 'https://api.anthropic.com';
 const DEFAULT_PORT = 3456;
 const MAX_BODY_BYTES = 10 * 1024 * 1024; // 10 MB — generous for large prompts, prevents abuse
@@ -1587,6 +1587,26 @@ export async function startProxy(opts = {}) {
                     const result = isOpenAI ? openaiToAnthropic(parsed, modelOverride) : (modelOverride ? { ...parsed, model: modelOverride } : parsed);
                     const r = result;
                     requestModel = (r.model || '').toLowerCase();
+                    // Suspended-model guard. Fable 5 / Mythos 5 are disabled for ALL
+                    // Anthropic customers (US-gov directive 2026-06-12); forwarding any
+                    // spelling of them 404s upstream with a confusing not_found. Reject
+                    // up front with an actionable error instead — runs before the
+                    // template build / account lease / forwarding, so nothing to unwind.
+                    // TEMP + reversible: governed by DARIO_SUSPENDED_MODELS (default
+                    // 'fable'); set it empty to re-enable when access is restored.
+                    if (requestModel && isSuspendedModel(requestModel)) {
+                        if (verbose)
+                            console.log(`[dario] suspended model rejected: ${requestModel} (${urlPath})`);
+                        res.writeHead(404, JSON_HEADERS);
+                        res.end(JSON.stringify({
+                            type: 'error',
+                            error: {
+                                type: 'not_found_error',
+                                message: `Model "${requestModel}" is suspended in this dario instance. Claude Fable 5 and Mythos 5 are disabled for all Anthropic customers (US government directive 2026-06-12 — https://www.anthropic.com/news/fable-mythos-access). Use claude-opus-4-8 or claude-sonnet-4-6 instead. To re-enable when access is restored, set DARIO_SUSPENDED_MODELS= (empty) on the dario instance.`,
+                            },
+                        }));
+                        return;
+                    }
                     // In passthrough mode, skip all Claude-specific injection — OAuth swap only.
                     // count_tokens also forwards thin (see resolveProxyTarget) — the endpoint
                     // counts the CLIENT's own prompt, so template injection would distort it.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@askalf/dario",
-  "version": "4.8.70",
+  "version": "4.8.72",
   "description": "Use your Claude Pro/Max subscription in any tool — Cursor, Cline, Aider, the Agent SDK, your scripts — at subscription pricing, not per-token API bills. One local Anthropic + OpenAI-compatible endpoint.",
   "type": "module",
   "bin": {