jeo-code 0.8.8 → 0.8.9

package/CHANGELOG.md

@@ -6,6 +6,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 The README mirrors the latest 5 entries — regenerate with `bun run changelog:sync`.
 
+## [0.8.9] - 2026-07-08
+_"프롬프트 라우팅 속도 문제가 있는거같은데 근본원인 알려줘" — prompt routing incurred significant latency due to a design oversight in LLM-based escalation: when heuristic confidence falls below threshold (standard conceptual questions with no code blocks/file paths conflict-trigger 0.35 confidence), it makes a blocking, synchronous LLM classifier call. If the user's global `thinkingLevel` is enabled (medium/high/xhigh), the cheap classifier model (Haiku 4.5 / GPT-4o-mini) also ran with reasoning enabled, wasting 500ms–1500ms on internal thoughts for a simple 1-word JSON response. The same reasoning-latency leak existed on all other background/internal LLM calls (compaction summarizer, goal verifier, memory distiller), blocking user turns on compaction boundaries._
+
+### Fixed
+- **Exposed explicit reasoning disable (`reasoningEffort: "none"`)** (`src/ai/types.ts`, `src/ai/model-manager.ts`) — added `"none"` option to `CallOptions["reasoningEffort"]`. `resolveCall` maps `"none"` to a wire-level value of `undefined` (disabling thinking) while bypassing the global config's `thinkingLevel` fallback.
+- **Optimized prompt routing classifier latency** (`src/agent/prompt-router.ts`) — passed `reasoningEffort: "none"` to `escalateToLlm`'s `callLlm` options, cutting the classifier's thinking tokens entirely.
+- **Optimized background LLM call latency** (`src/agent/compaction.ts`, `src/agent/goal-verifier.ts`, `src/agent/memory.ts`) — passed `reasoningEffort: "none"` to the compaction summarizer, goal verifier, and memory distiller calls, preventing the model from thinking about summaries or logs and unblocking turn boundaries.
+
+### Added
+- **Reasoning-disable unit tests** (`test/model-manager.test.ts`) — added tests verifying that `resolveCall` with `reasoningEffort: "none"` resolves to `undefined` (thinking disabled) even when the global default is enabled, while ordinary levels and default values resolve correctly.
+
 ## [0.8.8] - 2026-07-08
 _"실동작검증을 통해서 유효성 평가하고 개선해줘" — live-verified the v0.8.5 fallback mechanism using active Claude Code OAuth credentials on this machine (confirmed 200 OK for both Fable 5 and Opus 4.8 via the proxy). Identified two critical improvement vectors during verification: (1) server-side fallback was previously scoped strictly to API-key credentials, leaving OAuth users (who also hit `reasoning_extraction` refusals) with today's reactive-only recovery; (2) the sequential `postAnthropic` retry ladder could throw immediately on a combination of different error types (e.g. temperature error first, then fallback error, then success)._
 

package/README.ja.md

@@ -262,11 +262,11 @@ CI は `.github/workflows/npm-publish.yml` で公開します — GitHub リリ
 ## 変更履歴 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.8.9]** (2026-07-08) — "프롬프트 라우팅 속도 문제가 있는거같은데 근본원인 알려줘" — prompt routing incurred significant latency due to a design oversight in LLM-based escalation: when heuristic confidence falls below threshold (standard conceptual questions with no code blocks/file paths conflict-trigger 0.35 confidence), it makes a blocking, synchronous LLM classifier call. If the user's global `thinkingLevel` is enabled (medium/high/xhigh), the cheap classifier model (Haiku 4.5 / GPT-4o-mini) also ran with reasoning enabled, wasting 500ms–1500ms on internal thoughts for a simple 1-word JSON response. The same reasoning-latency leak existed on all other background/internal LLM calls (compaction summarizer, goal verifier, memory distiller), blocking user turns on compaction boundaries.
 - **[0.8.8]** (2026-07-08) — "실동작검증을 통해서 유효성 평가하고 개선해줘" — live-verified the v0.8.5 fallback mechanism using active Claude Code OAuth credentials on this machine (confirmed 200 OK for both Fable 5 and Opus 4.8 via the proxy). Identified two critical improvement vectors during verification: (1) server-side fallback was previously scoped strictly to API-key credentials, leaving OAuth users (who also hit `reasoning_extraction` refusals) with today's reactive-only recovery; (2) the sequential `postAnthropic` retry ladder could throw immediately on a combination of different error types (e.g. temperature error first, then fallback error, then success).
 - **[0.8.7]** (2026-07-08) — Fix-forward for a CI-caught release failure in v0.8.6 (this same session): `test/launch-prompt-routing.test.ts` shipped with a dropped closing brace (a git 3-way merge misaligned on a duplicate `});` line during this session's own working-tree isolation, then a subsequent `git stash push --keep-index` silently reset the working tree to the pre-fix staged content — the fix was applied once, verified once, then invisibly reverted before the commit that actually shipped). The v0.8.6 GitHub Actions release workflow correctly caught the syntax error at `bun test` and failed BEFORE the npm publish step ran; `npm view jeo-code version` confirmed 0.8.5 remained latest throughout — nothing broken ever reached the registry. Per this repo's established pattern for a same-day CI-caught regression (v0.8.0 → v0.8.1), shipping forward rather than retargeting the already-publicly-failed v0.8.6 tag.
 - **[0.8.6]** (2026-07-08) — Explicit `/model` pin always winning over PromptRouter was correct but opaque: `/route status` showed "routing: on" even while a session pin blocked routing from ever evaluating a prompt, and there was no way back to routed mode short of restarting the session. Also closed a related escalation gap: `routing.enabled` with `roles.smol` unset made ambiguous-prompt LLM escalation silently skip every turn, even when a cheaper credentialed model existed to run the classifier.
 - **[0.8.5]** (2026-07-08) — "딥리서치 다음과같이 페이블5 모델이 응답하게 되는경우를 피하기 위해" — a claude-fable-5 turn hit `Error: Refusal (reasoning_extraction): This request was blocked as it seems to violate Anthropic's Terms of Service.` jeo already had a full reactive recovery ladder (context reset → thinking-artifact strip → guidance strip → fail-fast on a category-shaped refusal), but every rung still resends to the SAME model — a deterministic classifier trip re-refuses identically no matter how much context gets stripped. Deep-researched Anthropic's current docs and found the actual fix: a documented, dedicated **server-side fallback** beta released alongside Fable 5 that retries a safety-classifier decline against a different model inside the SAME request, before the caller ever sees an error.
-- **[0.8.4]** (2026-07-08) — Follow-up to the 0.8.3 `jeo --tmux` E2E verification, which surfaced two real bugs: (1) the fake `test-anthropic-key` fixture found sitting in the user's REAL `~/.jeo/config.json` turned out to be reproducible — jeo's own test suite could silently write to the real config with no guard, and a prior session's routing tests had done exactly that; (2) the E2E run's TUI showed `'antigravity/gemini-pro-agent' is not in the live antigravity catalog` for a model that WAS live — traced to the note firing on a FAILED provider discovery (timeout/expired OAuth) rather than a genuinely absent model.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.ko.md

@@ -263,11 +263,11 @@ CI는 `.github/workflows/npm-publish.yml`로 배포합니다 — GitHub 릴리
 ## 변경 이력 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.8.9]** (2026-07-08) — "프롬프트 라우팅 속도 문제가 있는거같은데 근본원인 알려줘" — prompt routing incurred significant latency due to a design oversight in LLM-based escalation: when heuristic confidence falls below threshold (standard conceptual questions with no code blocks/file paths conflict-trigger 0.35 confidence), it makes a blocking, synchronous LLM classifier call. If the user's global `thinkingLevel` is enabled (medium/high/xhigh), the cheap classifier model (Haiku 4.5 / GPT-4o-mini) also ran with reasoning enabled, wasting 500ms–1500ms on internal thoughts for a simple 1-word JSON response. The same reasoning-latency leak existed on all other background/internal LLM calls (compaction summarizer, goal verifier, memory distiller), blocking user turns on compaction boundaries.
 - **[0.8.8]** (2026-07-08) — "실동작검증을 통해서 유효성 평가하고 개선해줘" — live-verified the v0.8.5 fallback mechanism using active Claude Code OAuth credentials on this machine (confirmed 200 OK for both Fable 5 and Opus 4.8 via the proxy). Identified two critical improvement vectors during verification: (1) server-side fallback was previously scoped strictly to API-key credentials, leaving OAuth users (who also hit `reasoning_extraction` refusals) with today's reactive-only recovery; (2) the sequential `postAnthropic` retry ladder could throw immediately on a combination of different error types (e.g. temperature error first, then fallback error, then success).
 - **[0.8.7]** (2026-07-08) — Fix-forward for a CI-caught release failure in v0.8.6 (this same session): `test/launch-prompt-routing.test.ts` shipped with a dropped closing brace (a git 3-way merge misaligned on a duplicate `});` line during this session's own working-tree isolation, then a subsequent `git stash push --keep-index` silently reset the working tree to the pre-fix staged content — the fix was applied once, verified once, then invisibly reverted before the commit that actually shipped). The v0.8.6 GitHub Actions release workflow correctly caught the syntax error at `bun test` and failed BEFORE the npm publish step ran; `npm view jeo-code version` confirmed 0.8.5 remained latest throughout — nothing broken ever reached the registry. Per this repo's established pattern for a same-day CI-caught regression (v0.8.0 → v0.8.1), shipping forward rather than retargeting the already-publicly-failed v0.8.6 tag.
 - **[0.8.6]** (2026-07-08) — Explicit `/model` pin always winning over PromptRouter was correct but opaque: `/route status` showed "routing: on" even while a session pin blocked routing from ever evaluating a prompt, and there was no way back to routed mode short of restarting the session. Also closed a related escalation gap: `routing.enabled` with `roles.smol` unset made ambiguous-prompt LLM escalation silently skip every turn, even when a cheaper credentialed model existed to run the classifier.
 - **[0.8.5]** (2026-07-08) — "딥리서치 다음과같이 페이블5 모델이 응답하게 되는경우를 피하기 위해" — a claude-fable-5 turn hit `Error: Refusal (reasoning_extraction): This request was blocked as it seems to violate Anthropic's Terms of Service.` jeo already had a full reactive recovery ladder (context reset → thinking-artifact strip → guidance strip → fail-fast on a category-shaped refusal), but every rung still resends to the SAME model — a deterministic classifier trip re-refuses identically no matter how much context gets stripped. Deep-researched Anthropic's current docs and found the actual fix: a documented, dedicated **server-side fallback** beta released alongside Fable 5 that retries a safety-classifier decline against a different model inside the SAME request, before the caller ever sees an error.
-- **[0.8.4]** (2026-07-08) — Follow-up to the 0.8.3 `jeo --tmux` E2E verification, which surfaced two real bugs: (1) the fake `test-anthropic-key` fixture found sitting in the user's REAL `~/.jeo/config.json` turned out to be reproducible — jeo's own test suite could silently write to the real config with no guard, and a prior session's routing tests had done exactly that; (2) the E2E run's TUI showed `'antigravity/gemini-pro-agent' is not in the live antigravity catalog` for a model that WAS live — traced to the note firing on a FAILED provider discovery (timeout/expired OAuth) rather than a genuinely absent model.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.md

@@ -265,11 +265,11 @@ Huge thanks to [gajae-code](https://github.com/Yeachan-Heo/gajae-code) for the i
 ## Changelog
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.8.9]** (2026-07-08) — "프롬프트 라우팅 속도 문제가 있는거같은데 근본원인 알려줘" — prompt routing incurred significant latency due to a design oversight in LLM-based escalation: when heuristic confidence falls below threshold (standard conceptual questions with no code blocks/file paths conflict-trigger 0.35 confidence), it makes a blocking, synchronous LLM classifier call. If the user's global `thinkingLevel` is enabled (medium/high/xhigh), the cheap classifier model (Haiku 4.5 / GPT-4o-mini) also ran with reasoning enabled, wasting 500ms–1500ms on internal thoughts for a simple 1-word JSON response. The same reasoning-latency leak existed on all other background/internal LLM calls (compaction summarizer, goal verifier, memory distiller), blocking user turns on compaction boundaries.
 - **[0.8.8]** (2026-07-08) — "실동작검증을 통해서 유효성 평가하고 개선해줘" — live-verified the v0.8.5 fallback mechanism using active Claude Code OAuth credentials on this machine (confirmed 200 OK for both Fable 5 and Opus 4.8 via the proxy). Identified two critical improvement vectors during verification: (1) server-side fallback was previously scoped strictly to API-key credentials, leaving OAuth users (who also hit `reasoning_extraction` refusals) with today's reactive-only recovery; (2) the sequential `postAnthropic` retry ladder could throw immediately on a combination of different error types (e.g. temperature error first, then fallback error, then success).
 - **[0.8.7]** (2026-07-08) — Fix-forward for a CI-caught release failure in v0.8.6 (this same session): `test/launch-prompt-routing.test.ts` shipped with a dropped closing brace (a git 3-way merge misaligned on a duplicate `});` line during this session's own working-tree isolation, then a subsequent `git stash push --keep-index` silently reset the working tree to the pre-fix staged content — the fix was applied once, verified once, then invisibly reverted before the commit that actually shipped). The v0.8.6 GitHub Actions release workflow correctly caught the syntax error at `bun test` and failed BEFORE the npm publish step ran; `npm view jeo-code version` confirmed 0.8.5 remained latest throughout — nothing broken ever reached the registry. Per this repo's established pattern for a same-day CI-caught regression (v0.8.0 → v0.8.1), shipping forward rather than retargeting the already-publicly-failed v0.8.6 tag.
 - **[0.8.6]** (2026-07-08) — Explicit `/model` pin always winning over PromptRouter was correct but opaque: `/route status` showed "routing: on" even while a session pin blocked routing from ever evaluating a prompt, and there was no way back to routed mode short of restarting the session. Also closed a related escalation gap: `routing.enabled` with `roles.smol` unset made ambiguous-prompt LLM escalation silently skip every turn, even when a cheaper credentialed model existed to run the classifier.
 - **[0.8.5]** (2026-07-08) — "딥리서치 다음과같이 페이블5 모델이 응답하게 되는경우를 피하기 위해" — a claude-fable-5 turn hit `Error: Refusal (reasoning_extraction): This request was blocked as it seems to violate Anthropic's Terms of Service.` jeo already had a full reactive recovery ladder (context reset → thinking-artifact strip → guidance strip → fail-fast on a category-shaped refusal), but every rung still resends to the SAME model — a deterministic classifier trip re-refuses identically no matter how much context gets stripped. Deep-researched Anthropic's current docs and found the actual fix: a documented, dedicated **server-side fallback** beta released alongside Fable 5 that retries a safety-classifier decline against a different model inside the SAME request, before the caller ever sees an error.
-- **[0.8.4]** (2026-07-08) — Follow-up to the 0.8.3 `jeo --tmux` E2E verification, which surfaced two real bugs: (1) the fake `test-anthropic-key` fixture found sitting in the user's REAL `~/.jeo/config.json` turned out to be reproducible — jeo's own test suite could silently write to the real config with no guard, and a prior session's routing tests had done exactly that; (2) the E2E run's TUI showed `'antigravity/gemini-pro-agent' is not in the live antigravity catalog` for a model that WAS live — traced to the note firing on a FAILED provider discovery (timeout/expired OAuth) rather than a genuinely absent model.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.zh.md

@@ -262,11 +262,11 @@ CI 通过 `.github/workflows/npm-publish.yml` 发布 — GitHub 发布 release
 ## 更新日志 (Changelog)
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
+- **[0.8.9]** (2026-07-08) — "프롬프트 라우팅 속도 문제가 있는거같은데 근본원인 알려줘" — prompt routing incurred significant latency due to a design oversight in LLM-based escalation: when heuristic confidence falls below threshold (standard conceptual questions with no code blocks/file paths conflict-trigger 0.35 confidence), it makes a blocking, synchronous LLM classifier call. If the user's global `thinkingLevel` is enabled (medium/high/xhigh), the cheap classifier model (Haiku 4.5 / GPT-4o-mini) also ran with reasoning enabled, wasting 500ms–1500ms on internal thoughts for a simple 1-word JSON response. The same reasoning-latency leak existed on all other background/internal LLM calls (compaction summarizer, goal verifier, memory distiller), blocking user turns on compaction boundaries.
 - **[0.8.8]** (2026-07-08) — "실동작검증을 통해서 유효성 평가하고 개선해줘" — live-verified the v0.8.5 fallback mechanism using active Claude Code OAuth credentials on this machine (confirmed 200 OK for both Fable 5 and Opus 4.8 via the proxy). Identified two critical improvement vectors during verification: (1) server-side fallback was previously scoped strictly to API-key credentials, leaving OAuth users (who also hit `reasoning_extraction` refusals) with today's reactive-only recovery; (2) the sequential `postAnthropic` retry ladder could throw immediately on a combination of different error types (e.g. temperature error first, then fallback error, then success).
 - **[0.8.7]** (2026-07-08) — Fix-forward for a CI-caught release failure in v0.8.6 (this same session): `test/launch-prompt-routing.test.ts` shipped with a dropped closing brace (a git 3-way merge misaligned on a duplicate `});` line during this session's own working-tree isolation, then a subsequent `git stash push --keep-index` silently reset the working tree to the pre-fix staged content — the fix was applied once, verified once, then invisibly reverted before the commit that actually shipped). The v0.8.6 GitHub Actions release workflow correctly caught the syntax error at `bun test` and failed BEFORE the npm publish step ran; `npm view jeo-code version` confirmed 0.8.5 remained latest throughout — nothing broken ever reached the registry. Per this repo's established pattern for a same-day CI-caught regression (v0.8.0 → v0.8.1), shipping forward rather than retargeting the already-publicly-failed v0.8.6 tag.
 - **[0.8.6]** (2026-07-08) — Explicit `/model` pin always winning over PromptRouter was correct but opaque: `/route status` showed "routing: on" even while a session pin blocked routing from ever evaluating a prompt, and there was no way back to routed mode short of restarting the session. Also closed a related escalation gap: `routing.enabled` with `roles.smol` unset made ambiguous-prompt LLM escalation silently skip every turn, even when a cheaper credentialed model existed to run the classifier.
 - **[0.8.5]** (2026-07-08) — "딥리서치 다음과같이 페이블5 모델이 응답하게 되는경우를 피하기 위해" — a claude-fable-5 turn hit `Error: Refusal (reasoning_extraction): This request was blocked as it seems to violate Anthropic's Terms of Service.` jeo already had a full reactive recovery ladder (context reset → thinking-artifact strip → guidance strip → fail-fast on a category-shaped refusal), but every rung still resends to the SAME model — a deterministic classifier trip re-refuses identically no matter how much context gets stripped. Deep-researched Anthropic's current docs and found the actual fix: a documented, dedicated **server-side fallback** beta released alongside Fable 5 that retries a safety-classifier decline against a different model inside the SAME request, before the caller ever sees an error.
-- **[0.8.4]** (2026-07-08) — Follow-up to the 0.8.3 `jeo --tmux` E2E verification, which surfaced two real bugs: (1) the fake `test-anthropic-key` fixture found sitting in the user's REAL `~/.jeo/config.json` turned out to be reproducible — jeo's own test suite could silently write to the real config with no guard, and a prior session's routing tests had done exactly that; (2) the E2E run's TUI showed `'antigravity/gemini-pro-agent' is not in the live antigravity catalog` for a model that WAS live — traced to the note firing on a FAILED provider discovery (timeout/expired OAuth) rather than a genuinely absent model.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jeo-code",
-  "version": "0.8.8",
+  "version": "0.8.9",
   "description": "Clean, highly optimized AI coding agent using spec-first loop",
   "type": "module",
   "main": "src/cli.ts",

package/src/agent/compaction.ts

@@ -403,6 +403,7 @@ export async function maybeCompact(
         {
           model: opts.model,
           systemPrompt,
+          reasoningEffort: "none",
           signal: opts.signal,
         }
       );

package/src/agent/goal-verifier.ts

@@ -56,7 +56,8 @@ Do not include any other text, markdown formatting, or code blocks. Output raw J
     ], {
       model,
       jsonMode: true,
-      maxTokens: 1000
+      maxTokens: 1000,
+      reasoningEffort: "none",
     });
 
     const parsed = JSON.parse(response.trim());

package/src/agent/memory.ts

@@ -929,7 +929,7 @@ export async function distillSessionMemory(
 
     const timeoutMs = opts.timeoutMs ?? 20_000;
     const distilled = await Promise.race([
-      callLlm(prompt, { model: opts.model, jsonMode: true, maxTokens: 2_000 }),
+      callLlm(prompt, { model: opts.model, jsonMode: true, maxTokens: 2_000, reasoningEffort: "none" }),
       new Promise<never>((_, reject) => setTimeout(() => reject(new Error(`memory distill timed out after ${timeoutMs}ms`)), timeoutMs)),
     ]);
 

package/src/agent/prompt-router.ts

@@ -477,6 +477,7 @@ async function escalateToLlm(prompt: string, smolModel: string, signal: AbortSig
       model: smolModel,
       jsonMode: true,
       maxTokens: 200,
+      reasoningEffort: "none",
       signal,
     });
     const parsed = tryExtractJsonObject<{ tier?: unknown }>(raw);

package/src/ai/model-manager.ts

@@ -404,7 +404,7 @@ export function credentialForCall(
   return effective;
 }
 
-async function resolveCall(options: Partial<CallOptions>, kind: "request" | "stream" = "request"): Promise<Resolved> {
+export async function resolveCall(options: Partial<CallOptions>, kind: "request" | "stream" = "request"): Promise<Resolved> {
   const config = await readGlobalConfig();
   const aliases = { ...((config as { modelAliases?: Record<string, string> }).modelAliases ?? {}) };
   const model = expandAlias(options.model ?? config.defaultModel, { ...BUILTIN_ALIASES, ...aliases });
@@ -429,7 +429,9 @@ async function resolveCall(options: Partial<CallOptions>, kind: "request" | "str
     numCtx: options.numCtx ?? (provider === "ollama" ? (config as { ollamaNumCtx?: number }).ollamaNumCtx : undefined),
     onUsage: options.onUsage,
     signal: options.signal,
-    reasoningEffort: options.reasoningEffort ?? thinkingToReasoningEffort(config.thinkingLevel),
+    reasoningEffort: options.reasoningEffort === "none"
+      ? undefined
+      : (options.reasoningEffort ?? thinkingToReasoningEffort(config.thinkingLevel)),
     onReasoning: options.onReasoning,
     onReasoningArtifact: options.onReasoningArtifact,
     tools: options.tools,

package/src/ai/providers/anthropic.ts

@@ -189,6 +189,7 @@ function anthropicAdaptiveEffort(effort: NonNullable<CallOptions["reasoningEffor
     case "low": return "low";
     case "medium": return "medium";
     case "high": return "high";
+    default: return "medium";
   }
 }
 

package/src/ai/providers/gemini.ts

@@ -69,6 +69,7 @@ export function geminiThinkingLevel(effort: NonNullable<CallOptions["reasoningEf
     case "medium": return "MEDIUM";
     case "high":
     case "xhigh": return "HIGH";
+    default: return "MEDIUM";
   }
 }
 

package/src/ai/types.ts

@@ -107,7 +107,7 @@ export interface CallOptions {
   /** Abort in-flight provider requests (Ctrl-C / timeout / supersede). */
   signal?: AbortSignal;
   /** Reasoning effort for reasoning models (o-series / gpt-5), mapped from thinkingLevel. */
-  reasoningEffort?: "low" | "medium" | "high";
+  reasoningEffort?: "low" | "medium" | "high" | "none";
   /** How an OpenAI-compatible backend enables/streams native reasoning (gjc parity):
    *  "openai" → `reasoning_effort`; "openrouter" → `reasoning: {effort}`; "qwen" →
    *  `enable_thinking: true`; "zai" → `thinking: {type:"enabled"}`. Set per provider by