full-loop.e2e.ts 2.2 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849
  1. import { createUserMessage } from '@deepseek-ai/dsh-llm'
  2. import { mkdtemp, rm } from 'node:fs/promises'
  3. import { tmpdir } from 'node:os'
  4. import { join } from 'node:path'
  5. import { afterEach, describe, expect, it } from 'vitest'
  6. import type { Context } from 'cordis'
  7. import { codingHarness, finalText, SYSTEM_PROMPT, waitForIdle } from './harness.ts'
  8. import { SessionId } from '@deepseek-ai/dsh-session'
  9. /**
  10. * The first place a REAL model meets the REAL bash tool: the cheap canary
  11. * before the coding-task e2e. Key-gated (see vitest.e2e.config.ts).
  12. */
  13. let ctx: Context | undefined
  14. let workdir: string | undefined
  15. afterEach(async () => {
  16. // Always dispose the harness, even on failure/retry/timeout: agent-loop
  17. // teardown stops the loop and LocalBashExecutor teardown kills any
  18. // process the model left behind.
  19. await ctx?.fiber.dispose()
  20. ctx = undefined
  21. if (workdir !== undefined) await rm(workdir, { recursive: true, force: true })
  22. workdir = undefined
  23. })
  24. describe.skipIf(!process.env.DEEPSEEK_API_KEY)('full loop: real model + real bash tool', () => {
  25. it('runs a bash command on request and reports its output', async () => {
  26. workdir = await mkdtemp(join(tmpdir(), 'dsh-full-loop-e2e-'))
  27. ctx = await codingHarness(workdir, { persona: SYSTEM_PROMPT })
  28. const agent = ctx.agentLoop.create(SessionId('e2e-loop'), { provider: 'deepseek', model: 'deepseek-v4-flash' })
  29. agent.followup(createUserMessage({ content: [{ type: 'text', text: 'Run `echo e2e-ok` with the bash tool and tell me its exact output.' }], source: { kind: 'user' } }))
  30. await waitForIdle(ctx, agent)
  31. const events = [...agent.session.events]
  32. const calls = events.filter(event => event.type === 'tool/call')
  33. expect(calls.length).toBeGreaterThan(0)
  34. expect(calls.some(event => event.data.name === 'bash')).toBe(true)
  35. const results = events.filter(event => event.type === 'tool/result')
  36. const resultTexts = results.flatMap(event =>
  37. event.data.message.content[0].content.filter(block => block.type === 'text').map(block => block.text))
  38. expect(resultTexts.some(text => text.includes('e2e-ok'))).toBe(true)
  39. expect(finalText(events)).toContain('e2e-ok')
  40. }, 120_000)
  41. })