analyst agent is clean

This commit is contained in:
dal 2025-08-05 00:54:23 -06:00
parent 0450b8c350
commit 9214075cad
No known key found for this signature in database
GPG Key ID: 16F4B0E1E9F61122
3 changed files with 155 additions and 245 deletions

View File

@ -1,136 +1,32 @@
import { RuntimeContext } from '@mastra/core/runtime-context';
import type { CoreMessage } from 'ai';
import { initLogger, wrapTraced } from 'braintrust';
import { afterAll, beforeAll, describe, expect, test } from 'vitest';
import type { AnalystRuntimeContext } from '../../workflows/analyst-workflow';
import { analystAgent } from './analyst-agent';
import type { ModelMessage } from "ai";
import { describe, expect, test } from "vitest";
import { createAnalystAgent } from "./analyst-agent";
describe('Analyst Agent Integration Tests', () => {
beforeAll(async () => {
initLogger({
apiKey: process.env.BRAINTRUST_KEY,
projectName: 'ANALYST-AGENT',
});
});
describe("Analyst Agent Integration Tests", () => {
test("should generate response for data analysis query with conversation history", async () => {
const messages: ModelMessage[] = [];
afterAll(async () => {
// Cleanup if needed
// Wait 500ms before finishing
await new Promise((resolve) => setTimeout(resolve, 500));
});
try {
const analystAgent = createAnalystAgent({
sql_dialect_guidance: "postgresql",
});
test('should generate response for data analysis query with conversation history', async () => {
// Stubbed conversation history - to be filled in later
const conversationHistory: CoreMessage[] = [
// TODO: Add stubbed conversation history here
];
const streamResult = await analystAgent.stream({
messages,
});
const tracedAgentWorkflow = wrapTraced(
async (messages: CoreMessage[]) => {
// Step 1: Generate response with analyst agent using conversation history
try {
const chatId = 'da05b6fb-01b2-4c1c-bc7f-7e55029a5c75';
const resourceId = 'c2dd64cd-f7f3-4884-bc91-d46ae431901e';
let response = "";
for await (const chunk of streamResult.fullStream) {
if (chunk.type === "text-delta") {
response += chunk.text;
}
}
// Create runtime context with required properties
const runtimeContext = new RuntimeContext<AnalystRuntimeContext>([
['userId', resourceId],
['chatId', chatId],
['dataSourceId', 'cc3ef3bc-44ec-4a43-8dc4-681cae5c996a'],
['dataSourceSyntax', 'postgres'],
['organizationId', 'bf58d19a-8bb9-4f1d-a257-2d2105e7f1ce'],
// Note: No messageId provided for testing scenario
]);
// Use stream with conversation history instead of single prompt
const stream = await analystAgent.stream(messages, {
maxSteps: 15,
runtimeContext,
onStepFinish: async (step) => {
console.log('\n=== onStepFinish callback (with history) ===');
console.log('Step structure:', JSON.stringify(step, null, 2));
console.log('Tool calls:', step.toolCalls);
console.log('Response messages:', step.response.messages);
// Response text is not directly available on step.response
console.log('===========================\n');
},
});
let response = '';
for await (const chunk of stream.fullStream) {
if (chunk.type === 'text-delta') {
response += chunk.textDelta;
}
}
return response;
} catch (error) {
console.error('Error during agent execution:', error);
throw error;
}
},
{ name: 'analyst-agent-with-history' }
);
// Test with conversation history (stubbed for now)
const result = await tracedAgentWorkflow(
conversationHistory.length > 0
? (conversationHistory as CoreMessage[])
: [{ role: 'user', content: 'What are the top 5 customers by revenue?' }]
);
expect(result).toBeDefined();
expect(typeof result).toBe('string');
expect(result.length).toBeGreaterThan(0);
// Should have generated some analysis response
expect(result).not.toBe('');
console.log('Final result:', result);
}, 300000);
test('should generate response for analysis query', async () => {
const tracedAgentWorkflow = wrapTraced(
async (input: string) => {
// Step 1: Generate response with analyst agent
try {
const chatId = 'da05b6fb-01b2-4c1c-bc7f-7e55029a5c75';
const resourceId = 'c2dd64cd-f7f3-4884-bc91-d46ae431901e';
// Create runtime context with required properties
const runtimeContext = new RuntimeContext<AnalystRuntimeContext>([
['userId', resourceId],
['chatId', chatId],
['dataSourceId', 'cc3ef3bc-44ec-4a43-8dc4-681cae5c996a'],
['dataSourceSyntax', 'postgresql'],
['organizationId', 'bf58d19a-8bb9-4f1d-a257-2d2105e7f1ce'],
]);
const stream = await analystAgent.stream(input, {
maxSteps: 15,
runtimeContext,
});
let responseText = '';
for await (const chunk of stream.fullStream) {
if (chunk.type === 'text-delta') {
responseText += chunk.textDelta;
}
}
return responseText;
} catch (error) {
console.error(error);
throw error;
}
},
{ name: 'AnalystAgentWorkflow' }
);
// Execute the workflow
const response = await tracedAgentWorkflow('please continue with the analysis');
// Verify response structure
expect(response).toBeDefined();
expect(typeof response).toBe('string');
expect(response.length).toBeGreaterThan(0);
}, 300000);
expect(response).toBeDefined();
expect(typeof response).toBe("string");
} catch (error) {
console.error("Error during agent execution:", error);
throw error;
}
}, 300000);
});

View File

@ -1,68 +1,82 @@
import { type ModelMessage, hasToolCall, stepCountIs, streamText } from 'ai';
import { wrapTraced } from 'braintrust';
import z from 'zod';
import { hasToolCall, type ModelMessage, stepCountIs, streamText } from "ai";
import { wrapTraced } from "braintrust";
import z from "zod";
import {
createDashboards,
createMetrics,
doneTool,
modifyDashboards,
modifyMetrics,
} from '../../tools';
import { Sonnet4 } from '../../utils/models/sonnet-4';
import { getAnalystAgentSystemPrompt } from './get-analyst-agent-system-prompt';
createDashboards,
createMetrics,
doneTool,
modifyDashboards,
modifyMetrics,
} from "../../tools";
import { Sonnet4 } from "../../utils/models/sonnet-4";
import { getAnalystAgentSystemPrompt } from "./get-analyst-agent-system-prompt";
const DEFAULT_CACHE_OPTIONS = {
anthropic: { cacheControl: { type: 'ephemeral', ttl: '1h' } },
anthropic: { cacheControl: { type: "ephemeral", ttl: "1h" } },
};
const STOP_CONDITIONS = [stepCountIs(18), hasToolCall('doneTool')];
const STOP_CONDITIONS = [stepCountIs(18), hasToolCall("doneTool")];
const AnalystAgentSchema = z.object({
sql_dialect_guidance: z.string().describe('The SQL dialect guidance for the analyst agent.'),
const AnalystAgentOptionsSchema = z.object({
sql_dialect_guidance: z
.string()
.describe("The SQL dialect guidance for the analyst agent."),
});
const AnalystStreamSchema = z.object({
messages: z
.array(z.custom<ModelMessage>())
.describe('The messages to send to the analyst agent.'),
const AnalystStreamOptionsSchema = z.object({
messages: z
.array(z.custom<ModelMessage>())
.describe("The messages to send to the analyst agent."),
});
export type AnalystAgentSchema = z.infer<typeof AnalystAgentSchema>;
export type AnalystStreamSchema = z.infer<typeof AnalystStreamSchema>;
export type AnalystAgentOptionsSchema = z.infer<
typeof AnalystAgentOptionsSchema
>;
export type AnalystStreamOptions = z.infer<typeof AnalystStreamOptionsSchema>;
export function createAnalystAgent(analystAgentSchema: AnalystAgentSchema) {
const steps: never[] = [];
export function createAnalystAgent(
analystAgentSchema: AnalystAgentOptionsSchema,
) {
const steps: never[] = [];
const systemMessage = {
role: 'system',
content: getAnalystAgentSystemPrompt(analystAgentSchema.sql_dialect_guidance),
providerOptions: DEFAULT_CACHE_OPTIONS,
} as ModelMessage;
const systemMessage = {
role: "system",
content: getAnalystAgentSystemPrompt(
analystAgentSchema.sql_dialect_guidance,
),
providerOptions: DEFAULT_CACHE_OPTIONS,
} as ModelMessage;
async function stream({ messages }: AnalystStreamSchema) {
return wrapTraced(
async () =>
streamText({
model: Sonnet4,
tools: { createMetrics, modifyMetrics, createDashboards, modifyDashboards, doneTool },
messages: [systemMessage, ...messages],
stopWhen: STOP_CONDITIONS,
toolChoice: 'required',
maxOutputTokens: 10000,
temperature: 0,
}),
{
name: 'Analyst Agent',
}
);
}
async function stream({ messages }: AnalystStreamOptions) {
return wrapTraced(
() =>
streamText({
model: Sonnet4,
tools: {
createMetrics,
modifyMetrics,
createDashboards,
modifyDashboards,
doneTool,
},
messages: [systemMessage, ...messages],
stopWhen: STOP_CONDITIONS,
toolChoice: "required",
maxOutputTokens: 10000,
temperature: 0,
}),
{
name: "Analyst Agent",
},
)();
}
async function getSteps() {
return steps;
}
async function getSteps() {
return steps;
}
return {
stream,
getSteps,
};
return {
stream,
getSteps,
};
}

View File

@ -1,81 +1,81 @@
import * as fs from 'node:fs';
import * as path from 'node:path';
import { describe, expect, it } from 'vitest';
import { getAnalystAgentSystemPrompt } from './get-analyst-agent-system-prompt';
import * as fs from "node:fs";
import * as path from "node:path";
import { describe, expect, it } from "vitest";
import { getAnalystAgentSystemPrompt } from "./get-analyst-agent-system-prompt";
describe('Analyst Agent Instructions', () => {
it('should validate template file contains expected variables', () => {
const promptPath = path.join(__dirname, 'analyst-agent-prompt.txt');
const content = fs.readFileSync(promptPath, 'utf-8');
describe("Analyst Agent Instructions", () => {
it("should validate template file contains expected variables", () => {
const promptPath = path.join(__dirname, "analyst-agent-prompt.txt");
const content = fs.readFileSync(promptPath, "utf-8");
// Expected template variables
const expectedVariables = ['sql_dialect_guidance', 'date'];
// Expected template variables
const expectedVariables = ["sql_dialect_guidance", "date"];
// Find all template variables in the file
const templateVariablePattern = /\{\{([^}]+)\}\}/g;
const foundVariables = new Set<string>();
// Find all template variables in the file
const templateVariablePattern = /\{\{([^}]+)\}\}/g;
const foundVariables = new Set<string>();
const matches = Array.from(content.matchAll(templateVariablePattern));
for (const match of matches) {
if (match[1] && match[1] !== 'variable') {
foundVariables.add(match[1]);
}
}
const matches = Array.from(content.matchAll(templateVariablePattern));
for (const match of matches) {
if (match[1] && match[1] !== "variable") {
foundVariables.add(match[1]);
}
}
// Convert to arrays for easier comparison
const foundVariablesArray = Array.from(foundVariables).sort();
const expectedVariablesArray = expectedVariables.sort();
// Convert to arrays for easier comparison
const foundVariablesArray = Array.from(foundVariables).sort();
const expectedVariablesArray = expectedVariables.sort();
// Check that we have exactly the expected variables
expect(foundVariablesArray).toEqual(expectedVariablesArray);
// Check that we have exactly the expected variables
expect(foundVariablesArray).toEqual(expectedVariablesArray);
// Also verify each expected variable exists
for (const variable of expectedVariables) {
expect(content).toMatch(new RegExp(`\\{\\{${variable}\\}\\}`));
}
// Also verify each expected variable exists
for (const variable of expectedVariables) {
expect(content).toMatch(new RegExp(`\\{\\{${variable}\\}\\}`));
}
// Ensure no unexpected variables exist
expect(foundVariables.size).toBe(expectedVariables.length);
});
// Ensure no unexpected variables exist
expect(foundVariables.size).toBe(expectedVariables.length);
});
it('should load and process the prompt template correctly', () => {
const sqlDialectGuidance = 'Test SQL guidance for PostgreSQL';
const result = getAnalystInstructions(sqlDialectGuidance);
it("should load and process the prompt template correctly", () => {
const sqlDialectGuidance = "Test SQL guidance for PostgreSQL";
const result = getAnalystAgentSystemPrompt(sqlDialectGuidance);
expect(result).toBeDefined();
expect(typeof result).toBe('string');
expect(result.length).toBeGreaterThan(0);
expect(result).toBeDefined();
expect(typeof result).toBe("string");
expect(result.length).toBeGreaterThan(0);
// Should contain the SQL guidance we provided
expect(result).toContain(sqlDialectGuidance);
// Should contain the SQL guidance we provided
expect(result).toContain(sqlDialectGuidance);
// Should not contain any unreplaced template variables
expect(result).not.toMatch(/\{\{sql_dialect_guidance\}\}/);
expect(result).not.toMatch(/\{\{date\}\}/);
// Should not contain any unreplaced template variables
expect(result).not.toMatch(/\{\{sql_dialect_guidance\}\}/);
expect(result).not.toMatch(/\{\{date\}\}/);
// Should contain the current date in YYYY-MM-DD format
const currentDate = new Date().toISOString().split('T')[0];
expect(result).toContain(currentDate);
});
// Should contain the current date in YYYY-MM-DD format
const currentDate = new Date().toISOString().split("T")[0];
expect(result).toContain(currentDate);
});
it('should contain expected sections from the prompt template', () => {
const result = getAnalystInstructions('Test guidance');
it("should contain expected sections from the prompt template", () => {
const result = getAnalystAgentSystemPrompt("Test guidance");
// Check for key sections that should be in the prompt
expect(result).toContain('<intro>');
expect(result).toContain('<analysis_mode_capability>');
expect(result).toContain('<sql_best_practices>');
expect(result).toContain('<visualization_and_charting_guidelines>');
expect(result).toContain('You are a Buster');
});
// Check for key sections that should be in the prompt
expect(result).toContain("<intro>");
expect(result).toContain("<analysis_mode_capability>");
expect(result).toContain("<sql_best_practices>");
expect(result).toContain("<visualization_and_charting_guidelines>");
expect(result).toContain("You are a Buster");
});
it('should throw an error for empty SQL dialect guidance', () => {
expect(() => {
getAnalystInstructions('');
}).toThrow('SQL dialect guidance is required');
it("should throw an error for empty SQL dialect guidance", () => {
expect(() => {
getAnalystAgentSystemPrompt("");
}).toThrow("SQL dialect guidance is required");
expect(() => {
getAnalystInstructions(' '); // whitespace only
}).toThrow('SQL dialect guidance is required');
});
expect(() => {
getAnalystAgentSystemPrompt(" "); // whitespace only
}).toThrow("SQL dialect guidance is required");
});
});