1
0

full-loop.e2e.ts 2.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748
  1. import { mkdtemp, rm } from 'node:fs/promises'
  2. import { tmpdir } from 'node:os'
  3. import { join } from 'node:path'
  4. import { afterEach, describe, expect, it } from 'vitest'
  5. import type { Context } from 'cordis'
  6. import { AgentId } from '@deepseek-ai/dsh-agent'
  7. import { codingHarness, finalText, SYSTEM_PROMPT, waitForIdle } from './harness.ts'
  8. /**
  9. * The first place a REAL model meets the REAL bash tool: the cheap canary
  10. * before the coding-task e2e. Key-gated (see vitest.e2e.config.ts).
  11. */
  12. let ctx: Context | undefined
  13. let workdir: string | undefined
  14. afterEach(async () => {
  15. // Always dispose the harness, even on failure/retry/timeout: agent-loop
  16. // teardown stops the loop and LocalBashExecutor teardown kills any
  17. // process the model left behind.
  18. await ctx?.fiber.dispose()
  19. ctx = undefined
  20. if (workdir !== undefined) await rm(workdir, { recursive: true, force: true })
  21. workdir = undefined
  22. })
  23. describe.skipIf(!process.env.DEEPSEEK_API_KEY)('full loop: real model + real bash tool', () => {
  24. it('runs a bash command on request and reports its output', async () => {
  25. workdir = await mkdtemp(join(tmpdir(), 'dsh-full-loop-e2e-'))
  26. ctx = await codingHarness(workdir, { persona: SYSTEM_PROMPT })
  27. const agent = ctx.agentLoop.create(AgentId('e2e-loop'), { model: 'deepseek-v4-flash' })
  28. agent.send([{ type: 'text', text: 'Run `echo e2e-ok` with the bash tool and tell me its exact output.' }])
  29. await waitForIdle(ctx, agent)
  30. const events = [...agent.session.events]
  31. const calls = events.filter(event => event.type === 'tool/call')
  32. expect(calls.length).toBeGreaterThan(0)
  33. expect(calls.some(event => event.data.name === 'bash')).toBe(true)
  34. const results = events.filter(event => event.type === 'tool/result')
  35. const resultTexts = results.flatMap(event =>
  36. event.data.content.filter(block => block.type === 'text').map(block => block.text))
  37. expect(resultTexts.some(text => text.includes('e2e-ok'))).toBe(true)
  38. expect(finalText(events)).toContain('e2e-ok')
  39. }, 120_000)
  40. })