@askalf/dario 4.8.98 → 4.8.99

package/README.md

@@ -12,7 +12,7 @@
   <a href="https://x.com/ask_alf"><img src="https://img.shields.io/badge/follow-@ask_alf-1da1f2?style=flat-square" alt="Follow on X"></a>
 </p>
 
-<p align="center"><em>Zero runtime dependencies · <a href="https://www.npmjs.com/package/@askalf/dario">SLSA-attested</a> every release · nothing phones home · ~19.2k lines you can read in a weekend · independent, unofficial, third-party (<a href="DISCLAIMER.md">DISCLAIMER.md</a>)</em></p>
+<p align="center"><em>Zero runtime dependencies · <a href="https://www.npmjs.com/package/@askalf/dario">SLSA-attested</a> every release · nothing phones home · ~20.5k lines you can read in a weekend · independent, unofficial, third-party (<a href="DISCLAIMER.md">DISCLAIMER.md</a>)</em></p>
 
 ---
 
@@ -228,6 +228,7 @@ Tune via `~/.dario/config.json` → `overageGuard`, or CLI flags: `--overage-beh
 - **Runs any non-Claude-Code agent.** A 64-entry schema-verified `TOOL_MAP` pre-maps Cline, Roo, Kilo, Cursor, Windsurf, Continue, Copilot, OpenHands, OpenClaw, Hermes, [hands](https://github.com/askalf/hands) tool names to CC's native set. No flag, no validator errors. → [`docs/integrations/agent-compat.md`](./docs/integrations/agent-compat.md)
 - **Recover output capability.** `dario proxy --system-prompt=partial` strips CC's tone/verbosity/no-comments constraints for 1.2–2.8× more output on open-ended work — empirically without flipping billing (the classifier doesn't read that slot). [Discussion #183](https://github.com/askalf/dario/discussions/183) has the per-prompt receipts. → [`docs/system-prompt.md`](./docs/system-prompt.md)
 - **Honor client thinking (`--honor-client-thinking`).** By default dario rebuilds the outbound request with CC's interactive thinking shape regardless of what the client sent. Pass this flag (or `DARIO_HONOR_CLIENT_THINKING=1`) to pass a non-CC client's own `thinking` block through unchanged. Off by default; the rebuild-to-CC path is what keeps the subscription pool routing.
+- **Preserve structured output (`--preserve-output-format`).** By default dario rebuilds `output_config` from CC's template (effort only), dropping a client's `output_config.format` JSON schema, so structured-output clients get unconstrained prose their strict parser rejects. Pass this flag (or `DARIO_PRESERVE_OUTPUT_FORMAT=1`) to carry the client's schema through unchanged — SDK clients like the Vercel AI SDK's `generateObject` then get schema-constrained output. Off by default; empirically without flipping billing (verified `five_hour` on `claude-sonnet-4-6`).
 - **Reachable from inside CC / any MCP client.** `dario subagent install` registers a CC sub-agent for in-session diagnostics; `dario mcp` exposes dario as a read-only MCP server. → [`docs/sub-agent.md`](./docs/sub-agent.md) · [`docs/mcp-server.md`](./docs/mcp-server.md)
 - **Shim mode** *(deprecated v4.2, removal scheduled v5.x)*. The original "no HTTP hop" path empirically matched only 3 of the 8 classifier axes and fell back to passthrough for the 1-block system prompts `claude -p` and the Agent SDK both send. Use **proxy mode** for any non-CC client — it's the only mode that rebuilds every request to CC's full canonical shape.
 
@@ -237,11 +238,11 @@ Tune via `~/.dario/config.json` → `overageGuard`, or CLI flags: `--overage-beh
 
 | Signal | Status |
 |---|---|
-| Source | **~19.2k** lines of TypeScript across **44** files — auditable in a weekend |
+| Source | **~20.5k** lines of TypeScript across **47** files — auditable in a weekend |
 | Dependencies | **0 runtime.** Verify: `npm ls --production` |
 | Provenance | Every release [SLSA-attested](https://www.npmjs.com/package/@askalf/dario) via GitHub Actions + Sigstore |
 | Scanning | [CodeQL](https://github.com/askalf/dario/actions/workflows/codeql.yml) on every push and weekly |
-| Tests | **92 test files**, **86 in the default `npm test` suite** (`test/all.test.mjs`) — green on every release |
+| Tests | **99 test files**, **92 in the default `npm test` suite** (`test/all.test.mjs`) — green on every release |
 | Drift response | hourly [`cc-drift-watch.yml`](./.github/workflows/cc-drift-watch.yml) + auto-publish on merge — median CC-release → dario-release under one hour |
 | Credentials | Never logged, redacted from errors, `0600` on disk in `0700` dirs; MCP server redacts at the tool boundary |
 | Network | Binds `127.0.0.1` by default; upstream only to configured backends over HTTPS; hardcoded SSRF allowlist |
@@ -328,7 +329,7 @@ PRs welcome. Small TypeScript codebase, zero runtime deps. Architecture + file-b
 git clone https://github.com/askalf/dario && cd dario
 npm install
 npm run dev    # tsx, no build step
-npm test       # 86 test files via test/all.test.mjs, green on every release
+npm test       # 92 test files via test/all.test.mjs, green on every release
 npm run e2e    # live proxy + OAuth (needs a working Claude backend)
 ```
 

package/dist/cc-template.d.ts

@@ -427,6 +427,7 @@ export declare function buildCCRequest(clientBody: Record<string, unknown>, bill
     systemPrompt?: string;
     skipFields?: ReadonlySet<string>;
     honorClientThinking?: boolean;
+    preserveOutputFormat?: boolean;
 }): {
     body: Record<string, unknown>;
     toolMap: Map<string, ToolMapping>;

package/dist/cc-template.js

@@ -1614,6 +1614,22 @@ export function buildCCRequest(clientBody, billingTag, cacheControl, identity, o
             ccRequest.output_config = { effort: resolveEffort(opts.effort, clientBody, model) };
         }
     }
+    // --preserve-output-format: carry the client's `output_config.format`
+    // (Anthropic's native structured-output JSON schema) through to upstream.
+    // dario rebuilds output_config from the CC template (effort only), so a
+    // structured-output client's schema is otherwise dropped and the model
+    // free-runs into prose its strict parser rejects. Unlike the injected
+    // thinking/effort above (gated on !isHaiku because haiku rejects them),
+    // this is the caller's OWN directive — it rides on whatever model the
+    // caller chose, and is independent of skipFields, which opts out dario's
+    // injected fields, not the caller's schema constraint.
+    if (opts.preserveOutputFormat) {
+        const clientOutputConfig = clientBody.output_config;
+        if (clientOutputConfig?.format !== undefined) {
+            const existing = ccRequest.output_config ?? {};
+            ccRequest.output_config = { ...existing, format: clientOutputConfig.format };
+        }
+    }
     ccRequest.stream = stream;
     // Replay the captured top-level key order. The hardcoded build order above
     // matches CC v2.1.104 and is kept as a deterministic fallback; when a live

package/dist/cli.js

@@ -523,6 +523,11 @@ async function proxy() {
     // ProxyOptions.honorClientThinking for rationale.
     const honorClientThinking = args.includes('--honor-client-thinking')
         || ['1', 'true', 'yes', 'on'].includes((process.env['DARIO_HONOR_CLIENT_THINKING'] ?? '').toLowerCase());
+    // --preserve-output-format — carry the client body's `output_config.format`
+    // (structured-output JSON schema) through to upstream instead of dropping it
+    // during the CC rebuild. See ProxyOptions.preserveOutputFormat for rationale.
+    const preserveOutputFormat = args.includes('--preserve-output-format')
+        || ['1', 'true', 'yes', 'on'].includes((process.env['DARIO_PRESERVE_OUTPUT_FORMAT'] ?? '').toLowerCase());
     // Non-loopback bind without DARIO_API_KEY turns dario into an open
     // OAuth-subscription relay for anyone on the reachable network. Refuse
     // to start rather than rely on the operator to read the startup banner.
@@ -542,7 +547,7 @@ async function proxy() {
         console.error(`[dario] Override (not recommended): pass --unsafe-no-auth if you have out-of-band network controls and accept the risk.`);
         process.exit(1);
     }
-    await startProxy({ port, host, verbose, verboseBodies, model, passthrough, preserveTools, hybridTools, mergeTools, noAutoDetect, strictTls, pacingMinMs, pacingJitterMs, thinkTimeBaseMs, thinkTimePerTokenMs, thinkTimeJitterMs, thinkTimeMaxMs, sessionStartMinMs, sessionStartJitterMs, stealth, drainOnClose, sessionIdleRotateMs, sessionRotateJitterMs, sessionMaxAgeMs, sessionPerClient, preserveOrchestrationTags, noLiveCapture, strictTemplate, maxConcurrent, maxQueued, queueTimeoutMs, effort, maxTokens, logFile, passthroughBetas, skipFields, systemPrompt, overageGuardEnabled, overageGuardBehavior, overageGuardCooldownMs, overageGuardNotifyOs, honorClientThinking });
+    await startProxy({ port, host, verbose, verboseBodies, model, passthrough, preserveTools, hybridTools, mergeTools, noAutoDetect, strictTls, pacingMinMs, pacingJitterMs, thinkTimeBaseMs, thinkTimePerTokenMs, thinkTimeJitterMs, thinkTimeMaxMs, sessionStartMinMs, sessionStartJitterMs, stealth, drainOnClose, sessionIdleRotateMs, sessionRotateJitterMs, sessionMaxAgeMs, sessionPerClient, preserveOrchestrationTags, noLiveCapture, strictTemplate, maxConcurrent, maxQueued, queueTimeoutMs, effort, maxTokens, logFile, passthroughBetas, skipFields, systemPrompt, overageGuardEnabled, overageGuardBehavior, overageGuardCooldownMs, overageGuardNotifyOs, honorClientThinking, preserveOutputFormat });
 }
 /**
  * Parse `--system-prompt=<verbatim|partial|aggressive|filepath>` (or the
@@ -1413,6 +1418,17 @@ async function help() {
                              \`thinking\`. Env:
                              DARIO_HONOR_CLIENT_THINKING.
 
+    --preserve-output-format
+                             Pass the client body's
+                             \`output_config.format\` (structured-output
+                             JSON schema) through to upstream instead of
+                             dropping it during the CC rebuild. Lets SDK
+                             clients (e.g. the Vercel AI SDK's
+                             \`generateObject\`) get schema-constrained
+                             output through dario. No effect when the
+                             client omits \`output_config.format\`. Env:
+                             DARIO_PRESERVE_OUTPUT_FORMAT.
+
     --upstream-proxy=URL / --via=URL
                              Route all of dario's outbound fetch
                              calls (api.anthropic.com, OpenAI-compat

package/dist/proxy.d.ts

@@ -317,6 +317,24 @@ interface ProxyOptions {
      * Env: DARIO_HONOR_CLIENT_THINKING=1.
      */
     honorClientThinking?: boolean;
+    /**
+     * When set, the client body's `output_config.format` (Anthropic's native
+     * structured-output JSON schema) is carried through to the upstream instead
+     * of being dropped during the CC rebuild. dario rebuilds `output_config`
+     * from the CC template (effort only), so structured-output clients — e.g.
+     * the Vercel AI SDK's `generateObject` — otherwise get unconstrained prose
+     * that fails their strict schema parse.
+     *
+     * Independent of skipFields: that opts out dario's INJECTED fields, while
+     * this preserves the caller's own schema constraint. The constraint is
+     * enforced upstream during decoding, so it holds regardless of the injected
+     * CC identity or thinking shape. Verified on claude-sonnet-4-6 that
+     * subscription (five_hour) routing is unchanged — headers, beta flags,
+     * metadata, and OAuth identity are untouched.
+     *
+     * Env: DARIO_PRESERVE_OUTPUT_FORMAT=1.
+     */
+    preserveOutputFormat?: boolean;
     /**
      * System-prompt mode for the Claude backend. Empirically validated as
      * unfingerprinted by the billing classifier in docs/research/system-prompt-classifier-study.md.

package/dist/proxy.js

@@ -1726,6 +1726,7 @@ export async function startProxy(opts = {}) {
                             systemPrompt: opts.systemPrompt,
                             skipFields,
                             honorClientThinking: opts.honorClientThinking ?? false,
+                            preserveOutputFormat: opts.preserveOutputFormat ?? false,
                         });
                         // Prompt-cache the tools + conversation prefix (the system prompt
                         // is already cached in ccBody's system blocks). Mirrors CC's cache

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@askalf/dario",
-  "version": "4.8.98",
+  "version": "4.8.99",
   "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": {