Richard/simple agent

[rguan72]

Description

[Provide a brief description of the changes in this PR]

How Has This Been Tested?

[Describe the tests you ran to verify your changes]

Backporting (check the box to trigger backport action)

Note: You have to check that the action passes, otherwise resolve the conflicts manually and tag the patches.

• This PR should be backported (make sure to check that the backport attempt succeeds)
• [Optional] Override Linear Check

---

Summary by cubic

Adds a simple agent-based chat streaming path with web and internal search tools, replacing the old packet pipeline. Also upgrades evals with a Braintrust quality classifier and a no-send-logs option, and strengthens tracing by masking large payloads.

• New Features

  • Introduced agent-driven chat flow (Litellm + OpenAI Agents) with tools: web_search, web_fetch, and internal_search.
  • Streams SSE/WebSocket-friendly delta events; integrated into process_message with a clean system prompt and inline tag-based citations.
  • Internal search uses existing SearchTool with tenant-aware DB session and safe overrides.

• Evals and Tracing

  • Braintrust evals now include a “quality” LLMClassifier score comparing Onyx vs ChatGPT.
  • Added --no-send-logs flag (and plumbing) to prevent sending eval logs.
  • Tracing masks data over 20k chars with safer truncation; reduced prompt logging noise.
  • Dataset CSV import no longer injects extra instruction text into questions.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
internal-search	Ready	Preview	Comment	Sep 25, 2025 7:21pm

[blacksmith-sh]

10 Jobs Failed:

Run Integration Tests v2 / prepare-build failed on "Generate OpenAPI schema"

[...]
  PERM_SYNC_SHAREPOINT_CERTIFICATE_PASSWORD: ***
  PERM_SYNC_SHAREPOINT_DIRECTORY_ID: ***
  GITHUB_REPO_NAME: ***/onyx
  pythonLocation: /opt/hostedtoolcache/Python/3.11.13/arm64
  PKG_CONFIG_PATH: /opt/hostedtoolcache/Python/3.11.13/arm64/lib/pkgconfig
  Python_ROOT_DIR: /opt/hostedtoolcache/Python/3.11.13/arm64
  Python2_ROOT_DIR: /opt/hostedtoolcache/Python/3.11.13/arm64
  Python3_ROOT_DIR: /opt/hostedtoolcache/Python/3.11.13/arm64
  LD_LIBRARY_PATH: /opt/hostedtoolcache/Python/3.11.13/arm64/lib
  PYTHONPATH: .
Traceback (most recent call last):
  File "/home/runner/_work/onyx/onyx/backend/scripts/onyx_openapi_schema.py", line 11, in <module>
    from onyx.main import app as app_fn
  File "/home/runner/_work/onyx/onyx/backend/onyx/main.py", line 55, in <module>
    from onyx.evals.tracing import setup_braintrust
  File "/home/runner/_work/onyx/onyx/backend/onyx/evals/tracing.py", line 5, in <module>
    from agents import set_trace_processors
ModuleNotFoundError: No module named 'agents'
Error: Process completed with exit code 1.

Run Integration Tests v2 / required failed on "Run actions/github-script@v7"

[...]
  retry-exempt-status-codes: 400,401,403,404,422
env:
  PRIVATE_REGISTRY: experimental-registry.blacksmith.sh:5000
  PRIVATE_REGISTRY_USERNAME: ***
  PRIVATE_REGISTRY_PASSWORD: ***
  OPENAI_API_KEY: ***
  SLACK_BOT_TOKEN: ***
  CONFLUENCE_TEST_SPACE_URL: ***
  CONFLUENCE_USER_NAME: ***
  CONFLUENCE_ACCESS_TOKEN: ***
  JIRA_BASE_URL: ***
  JIRA_USER_EMAIL: ***
  JIRA_API_TOKEN: ***
  PERM_SYNC_SHAREPOINT_CLIENT_ID: ***
  PERM_SYNC_SHAREPOINT_PRIVATE_KEY: ***
  PERM_SYNC_SHAREPOINT_CERTIFICATE_PASSWORD: ***
  PERM_SYNC_SHAREPOINT_DIRECTORY_ID: ***
  GITHUB_REPO_NAME: ***/onyx
Error: One or more upstream jobs failed or were cancelled.

8 jobs failed running on non-Blacksmith runners.

---

Summary: 4 successful workflows, 5 failed workflows

• ❌ Connector Tests (1 job failed)
• ❌ External Dependency Unit Tests (1 job succeeded, 5 jobs failed)
• ✅ Helm - Lint and Test Charts (1 job succeeded)
• ✅ Check Lazy Imports (1 job succeeded)
• ✅ Quality Checks PR (1 job succeeded)
• ✅ Python Unit Tests (1 job succeeded)
• ❌ Run Integration Tests v2 (3 jobs succeeded, 2 jobs failed)
• ❌ Ensure PR references Linear (1 job failed)
• ❌ Python Checks (1 job failed)

Last updated: 2025-09-25 19:31:12 UTC