@sunerpy/opencode-kiro-auth 0.11.0 → 0.13.0

package/README.md

@@ -8,8 +8,9 @@
 [![license](https://img.shields.io/npm/l/@sunerpy/opencode-kiro-auth)](https://www.npmjs.com/package/@sunerpy/opencode-kiro-auth)
 
 > OpenCode plugin that lets OpenCode use AWS Kiro (CodeWhisperer) as a model
-> provider — Claude Sonnet, Opus, and Haiku, plus the open-weight models Kiro
-> proxies (DeepSeek, GLM, MiniMax, Qwen3) — with substantial trial quotas.
+> provider — Claude Sonnet, Opus, and Haiku, plus the other models Kiro proxies
+> (OpenAI GPT 5.6, DeepSeek, GLM, MiniMax, Qwen3) — with substantial trial
+> quotas.
 
 ## Table of Contents
 

package/dist/constants.js

@@ -81,6 +81,11 @@ export const MODEL_MAPPING = {
     'minimax-m2.5': 'minimax-m2.5',
     'minimax-m2.1': 'minimax-m2.1',
     'qwen3-coder-next': 'qwen3-coder-next',
+    // OpenAI GPT 5.6 — probe-confirmed (200, us-east-1); other naming guesses 400.
+    // Effort variants (-low..-max) are parsed by resolveModelVariant, not mapped here.
+    'gpt-5.6-sol': 'gpt-5.6-sol',
+    'gpt-5.6-terra': 'gpt-5.6-terra',
+    'gpt-5.6-luna': 'gpt-5.6-luna',
     // Legacy / internal mappings kept for backwards compatibility
     'claude-3-7-sonnet': 'CLAUDE_3_7_SONNET_20250219_V1_0',
     'nova-swe': 'AGI_NOVA_SWE_V1_5',

package/dist/plugin/effort.d.ts

@@ -11,6 +11,15 @@ export declare function supportsEffort(kiroModel: string): boolean;
  * Check if a model supports xhigh effort level.
  */
 export declare function supportsXHighEffort(kiroModel: string): boolean;
+/**
+ * Build the additionalModelRequestFields payload carrying the effort level.
+ *
+ * GPT and Claude take effort through different, mutually-exclusive wire fields
+ * (each rejects the other's with HTTP 400, probe-confirmed):
+ * - GPT (Mantle):  `reasoning.effort`
+ * - Claude:        `output_config.effort`
+ */
+export declare function buildEffortRequestFields(kiroModel: string, effort: Effort): Record<string, unknown>;
 /**
  * Resolve effort level for a given model.
  * - Returns undefined if model doesn't support effort

package/dist/plugin/effort.js

@@ -2,11 +2,23 @@
  * Effort levels ordered from lowest to highest reasoning depth.
  */
 export const EFFORT_LEVELS = ['low', 'medium', 'high', 'xhigh', 'max'];
+/**
+ * OpenAI GPT models (Kiro proxies these via Mantle). They accept the effort enum
+ * through a DIFFERENT wire field than Claude — see buildEffortRequestFields.
+ * All five levels (incl. xhigh/max) are probe-confirmed: credit usage scales
+ * monotonically low<medium<high<xhigh<max (.omo/evidence/task-gpt56-effort-probe.txt).
+ */
+const GPT_REASONING_MODELS = new Set(['gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna']);
 /**
  * Models that support the 5-value effort enum (including xhigh).
  * These models support up to 128k thinking tokens with max effort.
  */
-const XHIGH_CAPABLE_MODELS = new Set(['claude-opus-4.7', 'claude-opus-4.8', 'claude-sonnet-5']);
+const XHIGH_CAPABLE_MODELS = new Set([
+    'claude-opus-4.7',
+    'claude-opus-4.8',
+    'claude-sonnet-5',
+    ...GPT_REASONING_MODELS
+]);
 /**
  * Models that support the 4-value effort enum (no xhigh).
  * xhigh requests on these models are clamped to max.
@@ -33,6 +45,20 @@ export function supportsEffort(kiroModel) {
 export function supportsXHighEffort(kiroModel) {
     return XHIGH_CAPABLE_MODELS.has(kiroModel);
 }
+/**
+ * Build the additionalModelRequestFields payload carrying the effort level.
+ *
+ * GPT and Claude take effort through different, mutually-exclusive wire fields
+ * (each rejects the other's with HTTP 400, probe-confirmed):
+ * - GPT (Mantle):  `reasoning.effort`
+ * - Claude:        `output_config.effort`
+ */
+export function buildEffortRequestFields(kiroModel, effort) {
+    if (GPT_REASONING_MODELS.has(kiroModel)) {
+        return { reasoning: { effort } };
+    }
+    return { output_config: { effort } };
+}
 /**
  * Resolve effort level for a given model.
  * - Returns undefined if model doesn't support effort

package/dist/plugin/models.js

@@ -10,7 +10,10 @@ const VARIANT_BASE_ALLOWLIST = new Set([
     'claude-opus-4-8',
     'claude-opus-4-7',
     'claude-sonnet-5',
-    'claude-sonnet-4-6'
+    'claude-sonnet-4-6',
+    'gpt-5.6-sol',
+    'gpt-5.6-terra',
+    'gpt-5.6-luna'
 ]);
 const EFFORT_SUFFIXES = ['low', 'medium', 'high', 'xhigh', 'max'];
 /**

package/dist/plugin/sdk-client.js

@@ -1,5 +1,6 @@
 import { CodeWhispererStreamingClient } from '@aws/codewhisperer-streaming-client';
 import { KIRO_CONSTANTS } from '../constants.js';
+import { buildEffortRequestFields } from './effort.js';
 const clientCache = new Map();
 const KIRO_CLI_MAX_ATTEMPTS = 3;
 export function createSdkClient(auth, region, effort) {
@@ -22,19 +23,15 @@ export function createSdkClient(auth, region, effort) {
         args.request.headers['x-amzn-kiro-agent-mode'] = 'vibe';
         return next(args);
     }, { step: 'build', name: 'addKiroHeaders' });
-    // Inject additionalModelRequestFields for effort-based thinking control
+    // Effort wire shape differs per family (GPT: reasoning.effort, Claude:
+    // output_config.effort; each 400s the other), so derive it from the body's modelId.
     if (effort) {
         client.middlewareStack.add((next) => async (args) => {
-            // The SDK serializes input to args.input, we need to modify the body
-            // before it's sent. The body is in args.request.body as a string.
             if (args.request?.body) {
                 try {
                     const body = JSON.parse(args.request.body);
-                    body.additionalModelRequestFields = {
-                        output_config: {
-                            effort
-                        }
-                    };
+                    const wireModel = body?.conversationState?.currentMessage?.userInputMessage?.modelId;
+                    body.additionalModelRequestFields = buildEffortRequestFields(wireModel, effort);
                     args.request.body = JSON.stringify(body);
                 }
                 catch {

package/dist/plugin.js

@@ -259,6 +259,96 @@ export const createKiroPlugin = (id) => async ({ client, directory }) => {
                         name: 'Qwen3 Coder Next (0.05x)',
                         limit: { context: 256000, output: 64000 },
                         modalities: { input: ['text'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol': {
+                        name: 'GPT 5.6 Sol (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol-low': {
+                        name: 'GPT 5.6 Sol (low) (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol-medium': {
+                        name: 'GPT 5.6 Sol (medium) (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol-high': {
+                        name: 'GPT 5.6 Sol (high) (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol-xhigh': {
+                        name: 'GPT 5.6 Sol (xhigh) (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-sol-max': {
+                        name: 'GPT 5.6 Sol (max) (2.4x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra': {
+                        name: 'GPT 5.6 Terra (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra-low': {
+                        name: 'GPT 5.6 Terra (low) (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra-medium': {
+                        name: 'GPT 5.6 Terra (medium) (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra-high': {
+                        name: 'GPT 5.6 Terra (high) (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra-xhigh': {
+                        name: 'GPT 5.6 Terra (xhigh) (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-terra-max': {
+                        name: 'GPT 5.6 Terra (max) (1.2x)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna': {
+                        name: 'GPT 5.6 Luna',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna-low': {
+                        name: 'GPT 5.6 Luna (low)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna-medium': {
+                        name: 'GPT 5.6 Luna (medium)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna-high': {
+                        name: 'GPT 5.6 Luna (high)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna-xhigh': {
+                        name: 'GPT 5.6 Luna (xhigh)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
+                    },
+                    'gpt-5.6-luna-max': {
+                        name: 'GPT 5.6 Luna (max)',
+                        limit: { context: 272000, output: 64000 },
+                        modalities: { input: ['text', 'image'], output: ['text'] }
                     }
                 };
             }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sunerpy/opencode-kiro-auth",
-  "version": "0.11.0",
+  "version": "0.13.0",
   "description": "OpenCode plugin for AWS Kiro (CodeWhisperer) providing access to Claude models",
   "type": "module",
   "main": "dist/index.js",