mirror of
https://github.com/andy-stack/vaultkeeper-ai.git
synced 2026-07-22 06:42:03 +00:00
d52b1c3c74
Introduce a new setting to control AI access to the web viewer tool and refactor prompt building to inject capability directives (memories, web search, web viewer) into all agent contexts. Update tool definitions to conditionally include web viewer and memory tools based on settings. Consolidate memories injection and active directives into a unified prompt builder pattern.
883 lines
21 KiB
TypeScript
883 lines
21 KiB
TypeScript
import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest';
|
|
import { PlanningAgent } from '../../Services/AIServices/PlanningAgent';
|
|
import { RegisterSingleton, DeregisterAllServices } from '../../Services/DependencyService';
|
|
import { Services } from '../../Services/Services';
|
|
import { Conversation } from '../../Conversations/Conversation';
|
|
import { ConversationContent } from '../../Conversations/ConversationContent';
|
|
import { Role } from '../../Enums/Role';
|
|
import { AITool } from '../../Enums/AITool';
|
|
import { AIToolCall } from '../../AIClasses/AIToolCall';
|
|
import { AIToolResponse } from '../../AIClasses/ToolDefinitions/AIToolResponse';
|
|
import { AIToolResponsePayload } from '../../AIClasses/ToolDefinitions/AIToolResponsePayload';
|
|
|
|
/**
|
|
* UNIT TESTS - PlanningAgent
|
|
*
|
|
* Tests the planning agent that:
|
|
* - Creates execution plans via SubmitPlan
|
|
* - Asks user questions during planning
|
|
* - Enforces tool restrictions (only planning tools allowed)
|
|
* - Handles replan vs initial plan scenarios
|
|
* - Respects max depth limits
|
|
*/
|
|
|
|
describe('PlanningAgent - Unit Tests', () => {
|
|
let service: PlanningAgent;
|
|
let mockAI: any;
|
|
let mockPrompt: any;
|
|
let mockAIToolService: any;
|
|
|
|
const createMockCallbacks = () => ({
|
|
onSubmit: vi.fn(),
|
|
onStreamingUpdate: vi.fn(),
|
|
onThoughtUpdate: vi.fn(),
|
|
onToolCallStarted: vi.fn(),
|
|
onPlanningStarted: vi.fn(),
|
|
onPlanningFinished: vi.fn(),
|
|
onUserQuestion: vi.fn().mockResolvedValue('User answer'),
|
|
onPlanUpdate: vi.fn(),
|
|
onPlanStepUpdate: vi.fn(),
|
|
onPlanReset: vi.fn(),
|
|
onComplete: vi.fn(),
|
|
onCancel: vi.fn()
|
|
});
|
|
|
|
beforeEach(() => {
|
|
// Mock IPrompt
|
|
mockPrompt = {
|
|
systemInstruction: vi.fn().mockReturnValue('System instruction'),
|
|
userInstruction: vi.fn().mockResolvedValue('User instruction'),
|
|
planningInstruction: vi.fn().mockResolvedValue('Planning instruction')
|
|
};
|
|
RegisterSingleton(Services.IPrompt, mockPrompt);
|
|
|
|
// Mock AIToolService
|
|
mockAIToolService = {
|
|
performAITool: vi.fn().mockResolvedValue(
|
|
new AIToolResponse(AITool.SearchVaultFiles, new AIToolResponsePayload({ results: [] }), 'test-tool-id')
|
|
)
|
|
};
|
|
RegisterSingleton(Services.AIToolService, mockAIToolService);
|
|
|
|
// Mock IAIClass
|
|
mockAI = {
|
|
systemPrompt: '',
|
|
userInstruction: '',
|
|
toolDefinitions: [],
|
|
agentType: undefined,
|
|
streamRequest: vi.fn()
|
|
};
|
|
RegisterSingleton(Services.IAIClass, mockAI);
|
|
|
|
// Mock SettingsService
|
|
RegisterSingleton(Services.SettingsService, {
|
|
settings: { enableMemories: false, allowUpdatingMemories: false }
|
|
});
|
|
|
|
// Create service
|
|
service = new PlanningAgent();
|
|
});
|
|
|
|
afterEach(() => {
|
|
DeregisterAllServices();
|
|
vi.restoreAllMocks();
|
|
});
|
|
|
|
describe('SubmitPlan - Valid plan submission', () => {
|
|
it('should submit a valid plan with multiple steps', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create a plan to organize notes'
|
|
}));
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
yield {
|
|
content: 'Creating plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Search for notes',
|
|
instruction: 'Search the vault for all notes'
|
|
},
|
|
{
|
|
description: 'Categorize notes',
|
|
instruction: 'Organize notes by topic'
|
|
},
|
|
{
|
|
description: 'Create index',
|
|
instruction: 'Create an index file'
|
|
}
|
|
],
|
|
user_message: 'Plan created'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(result?.executionSteps).toHaveLength(3);
|
|
expect(result?.executionSteps[0].description).toBe('Search for notes');
|
|
expect(result?.executionSteps[1].description).toBe('Categorize notes');
|
|
expect(result?.executionSteps[2].description).toBe('Create index');
|
|
});
|
|
|
|
it('should submit a plan with single step', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Simple task'
|
|
}));
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Complete task',
|
|
instruction: 'Do the single task'
|
|
}
|
|
]
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(result?.executionSteps).toHaveLength(1);
|
|
});
|
|
});
|
|
|
|
describe('SubmitPlan validation', () => {
|
|
it('should reject invalid plan structure', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// Invalid: missing required fields
|
|
yield {
|
|
content: 'Invalid plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
// Missing instruction
|
|
description: 'Incomplete step'
|
|
}
|
|
]
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Valid plan
|
|
yield {
|
|
content: 'Valid plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Complete step',
|
|
instruction: 'Full instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(attemptCount).toBeGreaterThan(1);
|
|
});
|
|
|
|
it('should retry when no plan is submitted', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// No plan submitted
|
|
yield {
|
|
content: 'Thinking about the plan...',
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Submit plan on retry
|
|
yield {
|
|
content: 'Plan ready',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Step 1',
|
|
instruction: 'Instruction 1'
|
|
}
|
|
]
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(attemptCount).toBe(2);
|
|
});
|
|
});
|
|
|
|
describe('AskUserQuestionPlanning', () => {
|
|
it('should ask user question during planning', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
callbacks.onUserQuestion.mockResolvedValue('Notes about AI');
|
|
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Search for relevant notes'
|
|
}));
|
|
|
|
let toolCallCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
toolCallCount++;
|
|
if (toolCallCount === 1) {
|
|
// Ask user question
|
|
yield {
|
|
content: 'Question',
|
|
toolCall: new AIToolCall(
|
|
AITool.AskUserQuestionPlanning,
|
|
{
|
|
question: 'What topic should I search for?',
|
|
user_message: 'Asking user for clarification'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Submit plan with user's answer incorporated
|
|
yield {
|
|
content: 'Plan with answer',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Search for AI notes',
|
|
instruction: 'Search for notes about AI'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(callbacks.onUserQuestion).toHaveBeenCalledWith('What topic should I search for?');
|
|
expect(result).toBeDefined();
|
|
expect(toolCallCount).toBe(2);
|
|
});
|
|
|
|
it('should handle multiple user questions in planning', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
callbacks.onUserQuestion
|
|
.mockResolvedValueOnce('markdown')
|
|
.mockResolvedValueOnce('daily notes');
|
|
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Organize files'
|
|
}));
|
|
|
|
let toolCallCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
toolCallCount++;
|
|
if (toolCallCount === 1) {
|
|
yield {
|
|
content: 'Q1',
|
|
toolCall: new AIToolCall(
|
|
AITool.AskUserQuestionPlanning,
|
|
{
|
|
question: 'What file type?',
|
|
user_message: 'Asking about file type'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else if (toolCallCount === 2) {
|
|
yield {
|
|
content: 'Q2',
|
|
toolCall: new AIToolCall(
|
|
AITool.AskUserQuestionPlanning,
|
|
{
|
|
question: 'Which folder?',
|
|
user_message: 'Asking about folder'
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Organize markdown files in daily notes',
|
|
instruction: 'Process files'
|
|
}
|
|
]
|
|
},
|
|
'tool-3'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(callbacks.onUserQuestion).toHaveBeenCalledTimes(2);
|
|
expect(result).toBeDefined();
|
|
});
|
|
|
|
it('should reject invalid AskUserQuestionPlanning arguments', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// Invalid: missing question
|
|
yield {
|
|
content: 'Invalid',
|
|
toolCall: new AIToolCall(
|
|
AITool.AskUserQuestionPlanning,
|
|
{
|
|
// Missing question field
|
|
user_message: 'Message'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Valid plan
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Step',
|
|
instruction: 'Instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(attemptCount).toBeGreaterThan(1);
|
|
});
|
|
});
|
|
|
|
describe('Tool restrictions', () => {
|
|
it('should allow planning tools (SearchVaultFiles)', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Plan to find files'
|
|
}));
|
|
|
|
let searchCalled = false;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
if (!searchCalled) {
|
|
searchCalled = true;
|
|
yield {
|
|
content: 'Searching',
|
|
toolCall: new AIToolCall(
|
|
AITool.SearchVaultFiles,
|
|
{
|
|
search_terms: ['test'],
|
|
user_message: 'Searching'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Process files',
|
|
instruction: 'Process found files'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(mockAIToolService.performAITool).toHaveBeenCalled();
|
|
expect(result).toBeDefined();
|
|
});
|
|
|
|
it('should allow planning tools (ReadVaultFiles)', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Plan based on file content'
|
|
}));
|
|
|
|
let readCalled = false;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
if (!readCalled) {
|
|
readCalled = true;
|
|
yield {
|
|
content: 'Reading',
|
|
toolCall: new AIToolCall(
|
|
AITool.ReadVaultFiles,
|
|
{
|
|
file_paths: ['test.md'],
|
|
user_message: 'Reading'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Update file',
|
|
instruction: 'Update based on content'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(mockAIToolService.performAITool).toHaveBeenCalled();
|
|
expect(result).toBeDefined();
|
|
});
|
|
|
|
it('should deny destructive tools (WriteVaultFile)', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// Try to write file (not allowed in planning)
|
|
yield {
|
|
content: 'Writing',
|
|
toolCall: new AIToolCall(
|
|
AITool.WriteVaultFile,
|
|
{
|
|
file_path: 'test.md',
|
|
content: 'content'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Submit valid plan
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Write file',
|
|
instruction: 'Write file in execution'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
// WriteVaultFile should be denied, not executed
|
|
expect(mockAIToolService.performAITool).not.toHaveBeenCalledWith(
|
|
expect.objectContaining({ name: AITool.WriteVaultFile })
|
|
);
|
|
expect(result).toBeDefined();
|
|
expect(attemptCount).toBeGreaterThan(1);
|
|
});
|
|
|
|
it('should deny execution tools (CompleteTask)', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// Try to complete task (execution agent tool)
|
|
yield {
|
|
content: 'Completing',
|
|
toolCall: new AIToolCall(
|
|
AITool.CompleteTask,
|
|
{
|
|
success: true,
|
|
description: 'Done'
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Step',
|
|
instruction: 'Instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(attemptCount).toBeGreaterThan(1);
|
|
});
|
|
|
|
it('should deny orchestration tools (CompleteStep)', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// Try to complete step (orchestration agent tool)
|
|
yield {
|
|
content: 'Completing',
|
|
toolCall: new AIToolCall(
|
|
AITool.CompleteStep,
|
|
{
|
|
confirm_completion: true
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Step',
|
|
instruction: 'Instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-2'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(attemptCount).toBeGreaterThan(1);
|
|
});
|
|
});
|
|
|
|
describe('Max depth protection', () => {
|
|
it('should return undefined when max planning depth is exceeded', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
// Never submit a plan, forcing retries
|
|
yield {
|
|
content: 'Still thinking...',
|
|
isComplete: true
|
|
};
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeUndefined();
|
|
expect(attemptCount).toBe(3); // MAX_AGENT_DEPTH
|
|
});
|
|
|
|
it('should track depth across retries', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount < 3) {
|
|
// No plan on first two attempts
|
|
yield {
|
|
content: 'Working on it',
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Submit plan on third attempt (just before max depth)
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Final step',
|
|
instruction: 'Final instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(attemptCount).toBe(3);
|
|
});
|
|
});
|
|
|
|
describe('Empty plan handling', () => {
|
|
it('should accept empty steps array as valid plan', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Create plan'
|
|
}));
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
// Empty plan is valid according to schema
|
|
yield {
|
|
content: 'Empty plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: []
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
// Empty plan should be accepted
|
|
expect(result).toBeDefined();
|
|
expect(result?.executionSteps).toEqual([]);
|
|
});
|
|
});
|
|
|
|
describe('Conversation structure', () => {
|
|
it('should maintain conversation across retries', async () => {
|
|
const callbacks = createMockCallbacks();
|
|
const conversation = new Conversation();
|
|
conversation.contents.push(new ConversationContent({
|
|
role: Role.User,
|
|
content: 'Initial request'
|
|
}));
|
|
|
|
let attemptCount = 0;
|
|
|
|
mockAI.streamRequest.mockImplementation(async function* () {
|
|
attemptCount++;
|
|
if (attemptCount === 1) {
|
|
// No plan
|
|
yield {
|
|
content: 'Need more info',
|
|
isComplete: true
|
|
};
|
|
} else {
|
|
// Check that retry message was added
|
|
const lastUserMessage = conversation.contents
|
|
.filter(c => c.role === Role.User)
|
|
.pop();
|
|
const hasRetryMessage = lastUserMessage?.content?.includes('submit');
|
|
|
|
if (hasRetryMessage) {
|
|
yield {
|
|
content: 'Plan',
|
|
toolCall: new AIToolCall(
|
|
AITool.SubmitPlan,
|
|
{
|
|
steps: [
|
|
{
|
|
description: 'Step',
|
|
instruction: 'Instruction'
|
|
}
|
|
]
|
|
},
|
|
'tool-1'
|
|
),
|
|
isComplete: true
|
|
};
|
|
}
|
|
}
|
|
});
|
|
|
|
service.resolveAIProvider();
|
|
const result = await service.runPlanningAgent(conversation, callbacks);
|
|
|
|
expect(result).toBeDefined();
|
|
expect(conversation.contents.length).toBeGreaterThan(1);
|
|
});
|
|
});
|
|
});
|