modelmix 5.1.12 → 5.1.16

package/test/abort.test.js

@@ -0,0 +1,517 @@
+const { expect } = require('chai');
+const { EventEmitter } = require('events');
+const sinon = require('sinon');
+const {
+    MixCustom,
+    MixModeration,
+    MixOpenAI,
+    ModelMix
+} = require('../index.js');
+const createOpenAIProviders = require('../lib/providers/openai');
+const { rejectsAnthropicSamplingParams } = require('../lib/providers/anthropic');
+const {
+    fetchBinaryResponse,
+    fetchJsonResponse,
+    fetchStreamResponse
+} = require('../http-client');
+
+function deferred() {
+    let resolve;
+    let reject;
+    const promise = new Promise((resolvePromise, rejectPromise) => {
+        resolve = resolvePromise;
+        reject = rejectPromise;
+    });
+    return { promise, reject, resolve };
+}
+
+async function rejection(promise) {
+    try {
+        await promise;
+    } catch (error) {
+        return error;
+    }
+    throw new Error('Expected promise to reject.');
+}
+
+function createProvider(handler) {
+    const provider = new MixCustom();
+    provider.create = handler;
+    return provider;
+}
+
+function createModel(handler, config = {}) {
+    return ModelMix.new({
+        config: {
+            bottleneck: { maxConcurrent: 8, minTime: 0 },
+            ...config
+        }
+    }).attach('custom', createProvider(handler)).addText('test');
+}
+
+describe('AbortSignal execution contract', () => {
+    if (global.setupTestHooks) global.setupTestHooks();
+
+    it('passes the direct signal through every terminal method without storing it', async () => {
+        const controller = new AbortController();
+        const cases = [
+            ['message', model => model.message(controller.signal), 'plain'],
+            ['raw', model => model.raw(controller.signal), 'plain'],
+            ['stream', model => model.stream(() => {}, controller.signal), 'plain'],
+            ['block', model => model.block({}, controller.signal), '```plain```'],
+            ['json', model => model.json(null, {}, {}, controller.signal), '{"ok":true}'],
+            ['execute', model => model.execute({ signal: controller.signal }), 'plain']
+        ];
+
+        for (const [, invoke, message] of cases) {
+            let providerRequest;
+            const model = createModel(async request => {
+                providerRequest = request;
+                return { message, toolCalls: [] };
+            });
+
+            await invoke(model);
+
+            expect(providerRequest.signal).to.equal(controller.signal);
+            expect(providerRequest.config).to.not.have.property('signal');
+            expect(providerRequest.options).to.not.have.property('signal');
+            expect(model).to.not.have.property('signal');
+            expect(model.config).to.not.have.property('signal');
+            expect(model.options).to.not.have.property('signal');
+        }
+    });
+
+    it('rejects invalid and misplaced signals before provider work starts', async () => {
+        const signal = new AbortController().signal;
+        expect(() => ModelMix.new({ config: { signal } })).to.throw(TypeError, 'config.signal');
+        expect(() => ModelMix.new({ options: { signal } })).to.throw(TypeError, 'options.signal');
+        expect(() => new MixCustom({ config: { signal } })).to.throw(TypeError, 'config.signal');
+        expect(() => new MixCustom({ options: { signal } })).to.throw(TypeError, 'options.signal');
+
+        let providerCalls = 0;
+        const model = createModel(async () => {
+            providerCalls += 1;
+            return { message: 'unexpected', toolCalls: [] };
+        });
+
+        expect(await rejection(model.message({}))).to.be.instanceOf(TypeError);
+        expect(await rejection(model.execute({ config: { signal } }))).to.be.instanceOf(TypeError);
+        expect(await rejection(model.execute({ options: { signal } }))).to.be.instanceOf(TypeError);
+        expect(providerCalls).to.equal(0);
+
+        const pluginModel = createModel(async () => {
+            providerCalls += 1;
+            return { message: 'unexpected', toolCalls: [] };
+        }).use({
+            name: 'misplaced-signal',
+            execute(context, next) {
+                context.request.options.signal = signal;
+                return next();
+            }
+        });
+        const pluginError = await rejection(pluginModel.message());
+        expect(pluginError).to.be.instanceOf(TypeError);
+        expect(pluginError.message).to.include('options.signal');
+        expect(providerCalls).to.equal(0);
+    });
+
+    it('preserves a pre-aborted custom reason and leaves execution state untouched', async () => {
+        const reason = new Error('cancel before start');
+        const controller = new AbortController();
+        controller.abort(reason);
+        let providerCalls = 0;
+        const model = createModel(async () => {
+            providerCalls += 1;
+            return { message: 'unexpected', toolCalls: [] };
+        });
+        const originalMessages = model.messages.map(message => ({ ...message }));
+
+        expect(await rejection(model.message(controller.signal))).to.equal(reason);
+        expect(providerCalls).to.equal(0);
+        expect(model.lastRaw).to.equal(null);
+        expect(model.messages).to.deep.equal(originalMessages);
+    });
+
+    it('aborts remote image preparation without mutating messages or invoking a provider', async () => {
+        const reason = new Error('stop image download');
+        const controller = new AbortController();
+        const fetchStarted = deferred();
+        let fetchedSignal;
+        sinon.stub(global, 'fetch').callsFake((_url, { signal }) => {
+            fetchedSignal = signal;
+            fetchStarted.resolve();
+            return new Promise((_, rejectPromise) => {
+                signal.addEventListener('abort', () => rejectPromise(signal.reason), { once: true });
+            });
+        });
+        let providerCalls = 0;
+        const model = ModelMix.new({ config: { bottleneck: { maxConcurrent: 1, minTime: 0 } } })
+            .attach('custom', createProvider(async () => {
+                providerCalls += 1;
+                return { message: 'unexpected', toolCalls: [] };
+            }))
+            .addImageFromUrl('https://example.test/image.png')
+            .addText('describe');
+        const originalMessages = JSON.parse(JSON.stringify(model.messages));
+
+        const execution = model.raw(controller.signal);
+        await fetchStarted.promise;
+        controller.abort(reason);
+
+        expect(await rejection(execution)).to.equal(reason);
+        expect(fetchedSignal).to.equal(controller.signal);
+        expect(providerCalls).to.equal(0);
+        expect(model.messages).to.deep.equal(originalMessages);
+    });
+
+    it('aborts an in-flight custom provider without falling back or recording its late result', async () => {
+        const pending = deferred();
+        const primaryStarted = deferred();
+        const reason = new Error('stop provider');
+        const controller = new AbortController();
+        let primaryCalls = 0;
+        let fallbackCalls = 0;
+        const model = ModelMix.new({ config: { bottleneck: { maxConcurrent: 2, minTime: 0 } } })
+            .attach('primary', createProvider(async () => {
+                primaryCalls += 1;
+                primaryStarted.resolve();
+                return pending.promise;
+            }))
+            .attach('fallback', createProvider(async () => {
+                fallbackCalls += 1;
+                return { message: 'fallback', toolCalls: [] };
+            }))
+            .addText('test');
+
+        const execution = model.raw(controller.signal);
+        await primaryStarted.promise;
+        controller.abort(reason);
+
+        expect(await rejection(execution)).to.equal(reason);
+        pending.resolve({ message: 'late', toolCalls: [] });
+        await new Promise(resolve => setImmediate(resolve));
+        expect(primaryCalls).to.equal(1);
+        expect(fallbackCalls).to.equal(0);
+        expect(model.lastRaw).to.equal(null);
+    });
+
+    it('rejects while queued and prevents the queued provider invocation', async () => {
+        const first = deferred();
+        const firstStarted = deferred();
+        const reason = new Error('leave queue');
+        const controller = new AbortController();
+        let providerCalls = 0;
+        const model = createModel(async () => {
+            providerCalls += 1;
+            if (providerCalls === 1) {
+                firstStarted.resolve();
+                return first.promise;
+            }
+            return { message: 'unexpected', toolCalls: [] };
+        }, {
+            max_history: -1,
+            bottleneck: { maxConcurrent: 1, minTime: 0 }
+        });
+
+        const running = model.raw();
+        await firstStarted.promise;
+        const queued = model.raw(controller.signal);
+        controller.abort(reason);
+
+        expect(await rejection(queued)).to.equal(reason);
+        first.resolve({ message: 'first', toolCalls: [] });
+        await running;
+        await new Promise(resolve => setImmediate(resolve));
+        expect(providerCalls).to.equal(1);
+    });
+
+    it('aborts retry backoff without retrying or falling back', async () => {
+        const reason = new Error('stop retry');
+        const controller = new AbortController();
+        let primaryCalls = 0;
+        let fallbackCalls = 0;
+        const model = ModelMix.new({
+            config: {
+                bottleneck: { maxConcurrent: 1, minTime: 0 },
+                retry: {
+                    enabled: true,
+                    retries: 2,
+                    baseDelayMs: 200,
+                    maxDelayMs: 200,
+                    retryableStatusCodes: [429]
+                }
+            }
+        })
+            .attach('primary', createProvider(async () => {
+                primaryCalls += 1;
+                throw { message: 'rate limited', statusCode: 429 };
+            }))
+            .attach('fallback', createProvider(async () => {
+                fallbackCalls += 1;
+                return { message: 'fallback', toolCalls: [] };
+            }))
+            .addText('test');
+
+        const execution = model.raw(controller.signal);
+        await new Promise(resolve => setTimeout(resolve, 20));
+        controller.abort(reason);
+
+        expect(await rejection(execution)).to.equal(reason);
+        expect(primaryCalls).to.equal(1);
+        expect(fallbackCalls).to.equal(0);
+    });
+
+    it('isolates signals across concurrent executions on one instance', async () => {
+        const first = deferred();
+        const second = deferred();
+        const started = deferred();
+        const controllerA = new AbortController();
+        const controllerB = new AbortController();
+        const reason = new Error('only A');
+        const requests = [];
+        const model = createModel(async request => {
+            requests.push(request);
+            if (requests.length === 2) started.resolve();
+            return requests.length === 1 ? first.promise : second.promise;
+        }, {
+            max_history: -1,
+            bottleneck: { maxConcurrent: 2, minTime: 0 }
+        });
+
+        const executionA = model.raw(controllerA.signal);
+        const executionB = model.raw(controllerB.signal);
+        await started.promise;
+        controllerA.abort(reason);
+        second.resolve({ message: 'B', toolCalls: [] });
+
+        expect(await rejection(executionA)).to.equal(reason);
+        expect((await executionB).message).to.equal('B');
+        first.resolve({ message: 'late A', toolCalls: [] });
+        await new Promise(resolve => setImmediate(resolve));
+        expect(requests.map(request => request.signal)).to.deep.equal([
+            controllerA.signal,
+            controllerB.signal
+        ]);
+        expect(model.lastRaw.message).to.equal('B');
+    });
+
+    it('exposes the signal to plugins and inherits it in child invocations', async () => {
+        const controller = new AbortController();
+        const contexts = [];
+        let providerSignal;
+        const provider = createProvider(async request => {
+            providerSignal = request.signal;
+            return { message: 'child', toolCalls: [] };
+        });
+        const model = ModelMix.new({ config: { bottleneck: { maxConcurrent: 2, minTime: 0 } } })
+            .attach('custom', provider)
+            .use({
+                name: 'child',
+                execute(context, next) {
+                    contexts.push(context);
+                    if (context.execution.depth > 0) return next();
+                    return context.invoke({
+                        messages: [{ role: 'user', content: 'child prompt' }],
+                        history: false
+                    });
+                }
+            })
+            .addText('parent prompt');
+
+        expect((await model.raw(controller.signal)).message).to.equal('child');
+        expect(contexts).to.have.length(2);
+        expect(contexts.every(context => context.signal === controller.signal)).to.equal(true);
+        expect(contexts.every(context => !Object.hasOwn(context.request.config, 'signal'))).to.equal(true);
+        expect(contexts.every(context => !Object.hasOwn(context.request.options, 'signal'))).to.equal(true);
+        expect(providerSignal).to.equal(controller.signal);
+    });
+
+    it('passes the signal to local and MCP tools and preserves abort through tool handling', async () => {
+        const controller = new AbortController();
+        let localSignal;
+        const model = ModelMix.new().addTool({
+            name: 'local',
+            description: 'local tool',
+            inputSchema: { type: 'object' }
+        }, async (_args, signal) => {
+            localSignal = signal;
+            return 'local result';
+        });
+
+        const localResult = await model.processToolCalls([
+            { id: 'local-1', name: 'local', input: {} }
+        ], controller.signal);
+        expect(localSignal).to.equal(controller.signal);
+        expect(localResult[0].content).to.equal('local result');
+
+        let mcpArguments;
+        model.toolClient.remote = {
+            async callTool(...args) {
+                mcpArguments = args;
+                return { content: [{ type: 'text', text: 'remote result' }] };
+            }
+        };
+        const remoteResult = await model.processToolCalls([
+            { id: 'remote-1', name: 'remote', input: { value: 1 } }
+        ], controller.signal);
+        expect(remoteResult[0].content).to.equal('remote result');
+        expect(mcpArguments[2]).to.deep.equal({ signal: controller.signal });
+
+        const reason = new Error('stop local tool');
+        const abortController = new AbortController();
+        const toolStarted = deferred();
+        let providerCalls = 0;
+        const toolModel = createModel(async () => {
+            providerCalls += 1;
+            return {
+                message: '',
+                toolCalls: [{ id: 'slow-1', name: 'slow', input: {} }]
+            };
+        }, { max_history: -1 }).addTool({
+            name: 'slow',
+            description: 'slow tool',
+            inputSchema: { type: 'object' }
+        }, async (_args, signal) => {
+            toolStarted.resolve(signal);
+            return new Promise((_, rejectPromise) => {
+                signal.addEventListener('abort', () => rejectPromise(signal.reason), { once: true });
+            });
+        });
+        const originalMessages = JSON.parse(JSON.stringify(toolModel.messages));
+        const execution = toolModel.raw(abortController.signal);
+        expect(await toolStarted.promise).to.equal(abortController.signal);
+        abortController.abort(reason);
+
+        expect(await rejection(execution)).to.equal(reason);
+        expect(providerCalls).to.equal(1);
+        expect(toolModel.messages).to.deep.equal(originalMessages);
+    });
+
+    it('forwards the signal through every fetch helper without serializing it', async () => {
+        const controller = new AbortController();
+        const calls = [];
+        sinon.stub(global, 'fetch').callsFake(async (_url, init) => {
+            calls.push(init);
+            if (calls.length === 1) {
+                return {
+                    ok: true,
+                    status: 200,
+                    headers: new Headers(),
+                    text: async () => '{"ok":true}'
+                };
+            }
+            if (calls.length === 2) {
+                return {
+                    ok: true,
+                    status: 200,
+                    headers: new Headers(),
+                    arrayBuffer: async () => Uint8Array.from([1, 2, 3]).buffer
+                };
+            }
+            return {
+                ok: true,
+                status: 200,
+                headers: new Headers(),
+                body: new ReadableStream({
+                    start(streamController) {
+                        streamController.close();
+                    }
+                })
+            };
+        });
+
+        await fetchJsonResponse('https://example.test/json', { body: '{}', signal: controller.signal });
+        await fetchBinaryResponse('https://example.test/binary', { signal: controller.signal });
+        const stream = await fetchStreamResponse('https://example.test/stream', { body: '{}', signal: controller.signal });
+        stream.data.resume();
+
+        expect(calls).to.have.length(3);
+        expect(calls.every(call => call.signal === controller.signal)).to.equal(true);
+        expect(JSON.parse(calls[0].body)).to.not.have.property('signal');
+    });
+
+    it('keeps the signal outside a built-in provider payload', async () => {
+        const controller = new AbortController();
+        const reason = new Error('stop OpenAI request');
+        const fetchStarted = deferred();
+        let request;
+        sinon.stub(global, 'fetch').callsFake((_url, init) => {
+            request = init;
+            fetchStarted.resolve();
+            return new Promise((_, rejectPromise) => {
+                init.signal.addEventListener('abort', () => rejectPromise(init.signal.reason), { once: true });
+            });
+        });
+        const model = ModelMix.new({ config: { bottleneck: { maxConcurrent: 1, minTime: 0 } } })
+            .gpt51()
+            .addText('test');
+
+        const execution = model.raw(controller.signal);
+        await fetchStarted.promise;
+        expect(request.signal).to.equal(controller.signal);
+        expect(JSON.parse(request.body)).to.not.have.property('signal');
+        controller.abort(reason);
+        expect(await rejection(execution)).to.equal(reason);
+    });
+
+    it('aborts the realtime WebSocket and preserves the custom reason', async () => {
+        const created = deferred();
+        class FakeWebSocket extends EventEmitter {
+            constructor() {
+                super();
+                this.readyState = 0;
+                this.terminated = false;
+                created.resolve(this);
+                setImmediate(() => {
+                    if (this.terminated) return;
+                    this.readyState = 1;
+                    this.emit('open');
+                });
+            }
+
+            send() {}
+
+            close() {
+                this.terminate();
+            }
+
+            terminate() {
+                if (this.terminated) return;
+                this.terminated = true;
+                this.readyState = 3;
+                this.emit('close');
+            }
+        }
+        const wsPath = require.resolve('ws');
+        const originalWebSocket = require.cache[wsPath].exports;
+        require.cache[wsPath].exports = FakeWebSocket;
+        try {
+            const { MixOpenAIWebSocket } = createOpenAIProviders({
+                ModelMix,
+                MixCustom,
+                MixOpenAI,
+                MixModeration,
+                rejectsAnthropicSamplingParams
+            });
+            const provider = new MixOpenAIWebSocket({
+                config: {
+                    apiKey: 'test-key',
+                    realtimeUrl: 'ws://example.test',
+                    websocketTimeoutMs: 2000
+                }
+            });
+            const model = ModelMix.new({ config: { bottleneck: { maxConcurrent: 1, minTime: 0 } } })
+                .attach('realtime-test', provider)
+                .addText('test');
+            const controller = new AbortController();
+            const reason = new Error('close realtime');
+            const execution = model.raw(controller.signal);
+            const socket = await created.promise;
+            controller.abort(reason);
+            expect(await rejection(execution)).to.equal(reason);
+            expect(socket.terminated).to.equal(true);
+        } finally {
+            require.cache[wsPath].exports = originalWebSocket;
+        }
+    });
+});

package/test/effort.test.js

@@ -242,6 +242,19 @@ describe('Unified effort scale', () => {
             expect(mapEffort('google', -1, 'gemini-3.7-flash')).to.equal(null);
         });
 
+        it('clamps Gemini 3.8 Flash to low, medium, and high', () => {
+            expect(mapEffort('google', 0, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'low' }
+            });
+            expect(mapEffort('google', 50, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'medium' }
+            });
+            expect(mapEffort('google', 100, 'gemini-3.8-flash')).to.deep.equal({
+                thinkingConfig: { thinkingLevel: 'high' }
+            });
+            expect(mapEffort('google', -1, 'gemini-3.8-flash')).to.equal(null);
+        });
+
         it('clamps Gemini levels for models with fewer steps', () => {
             expect(mapEffort('google', 10, 'gemini-3-pro-preview')).to.deep.equal({
                 thinkingConfig: { thinkingLevel: 'low' }

package/test/fallback.test.js

@@ -34,12 +34,12 @@ describe('Provider Fallback Chain Tests', () => {
         });
 
         it('should attach chain shortcuts from arguments with optional per-model effort', () => {
-            model.chain('sonnet5', 'gpt56luna@20', 'gemini37flash@-1');
+            model.chain('sonnet5', 'gpt56luna@20', 'gemini38flash@-1');
 
             expect(model.models.map(({ key }) => key)).to.deep.equal([
                 'claude-sonnet-5',
                 'gpt-5.6-luna',
-                'gemini-3.7-flash'
+                'gemini-3.8-flash'
             ]);
             expect(model.models[0].provider.config).to.not.have.property('effort');
             expect(model.models[1].provider.config.effort).to.equal(20);

package/test/live.mcp.js

@@ -181,8 +181,8 @@ describe('Live MCP Integration Tests', function () {
             }
         });
 
-        it('should use custom MCP tools with Gemini 3.7 Flash', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+        it('should use custom MCP tools with Gemini 3.8 Flash', async function () {
+            const model = ModelMix.new(setup).gemini38flash();
 
             // Add password generator tool
             model.addTool({
@@ -220,7 +220,7 @@ describe('Live MCP Integration Tests', function () {
             model.addText('Generate a secure password of 16 characters with symbols.');
 
             const response = await model.message();
-            console.log(`Gemini 3.7 Flash with MCP tools: ${response}`);
+            console.log(`Gemini 3.8 Flash with MCP tools: ${response}`);
 
             expect(response).to.be.a('string');
             // Check password is mentioned and a generated password string is present
@@ -407,8 +407,8 @@ describe('Live MCP Integration Tests', function () {
             expect(result.factorial_result).to.equal(120);
         });
 
-        it('should use MCP tools with JSON output using Gemini 3.7 Flash', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+        it('should use MCP tools with JSON output using Gemini 3.8 Flash', async function () {
+            const model = ModelMix.new(setup).gemini38flash();
 
             // Add system info tool
             model.addTool({
@@ -447,7 +447,7 @@ describe('Live MCP Integration Tests', function () {
                 generated_at: ""
             });
 
-            console.log(`Gemini 3.7 Flash with MCP tools JSON result:`, result);
+            console.log(`Gemini 3.8 Flash with MCP tools JSON result:`, result);
 
             expect(result).to.be.an('object');
             expect(result.timestamp).to.be.a('number');

package/test/live.test.js

@@ -59,7 +59,7 @@ describe('Live Integration Tests', function () {
         });
 
         it('should process images with Google Gemini', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+            const model = ModelMix.new(setup).gemini38flash();
 
             model.addImageFromUrl(blueSquareBase64)
                 .addText('What color is this image? Answer in one word only.');
@@ -120,7 +120,7 @@ describe('Live Integration Tests', function () {
         });
 
         it('should return structured JSON with Google Gemini', async function () {
-            const model = ModelMix.new(setup).gemini37flash();
+            const model = ModelMix.new(setup).gemini38flash();
 
             model.addText('Generate information about a fictional city.');
 
@@ -181,7 +181,9 @@ describe('Live Integration Tests', function () {
     describe('Additional Model Tests', function () {
 
         it('should work with GPT-OSS model', async function () {
-            const model = ModelMix.new(setup).gptOss();
+            const model = ModelMix.new(setup).gptOss({
+                options: { max_tokens: 128 }
+            });
 
             model.addText('Say "gptoss test successful" and nothing else.');
 
@@ -241,8 +243,11 @@ describe('Live Integration Tests', function () {
 
     describe('JSON Structured Output for New Models', function () {
 
-        it('should return structured JSON with KimiK25 Thinking', async function () {
-            const model = ModelMix.new(setup).kimiK25();
+        it('should return structured JSON with Kimi K3', async function () {
+            const model = ModelMix.new(setup).kimiK3({
+                options: { max_tokens: 512 },
+                mix: { moonshot: false, together: true }
+            });
 
             model.addText('Generate information about a fictional vehicle.');
 
@@ -253,7 +258,7 @@ describe('Live Integration Tests', function () {
                 manufacturer: "Future Motors"
             });
 
-            console.log(`KimiK25 JSON result:`, result);
+            console.log(`Kimi K3 JSON result:`, result);
 
             expect(result).to.be.an('object');
             expect(result).to.have.property('name');