feat: Add OpenAI provider support for AI assistance by vinzee · Pull Request #1960 · hyperdxio/hyperdx

vinzee · 2026-03-21T18:46:01Z

Summary

Add OpenAI-compatible provider support for AI-powered chart assistance, enabling use of any OpenAI Chat Completions-compatible endpoint (Azure OpenAI, OpenRouter, LiteLLM proxies) as an alternative to Anthropic.

Key changes:

Implement getOpenAIModel() in controllers/ai.ts using @ai-sdk/openai
Support custom request headers via AI_CLIENT_ID / AI_USERNAME env vars
Support arbitrary request body fields via AI_EXTRA_BODY env var (JSON),
useful for proxy/gateway compatibility (e.g. LiteLLM passthrough mode)
Add detailed error logging for AI API call failures (status code, response
body, URL) to aid debugging provider issues
Adapt to sql-formatter's breaking rename of reservedPhrases to reservedKeywordPhrases in the
TokenizerOptions interface

Screenshots or video

Before	After

How to test locally or on Vercel

Set env vars: AI_PROVIDER=openai, AI_API_KEY=<key>,
AI_BASE_URL=<endpoint>, AI_MODEL_NAME=<model>
Optionally set AI_CLIENT_ID, AI_USERNAME for custom headers
Optionally set AI_EXTRA_BODY='{"passthrough":true}' for extra body fields
Open Chart Explorer, select a source, use the AI assistant to generate a chart
Verify the assistant returns a valid chart config

References

Linear Issue:
Related PRs:

vercel · 2026-03-21T18:46:05Z

@vinzee is attempting to deploy a commit to the HyperDX Team on Vercel.

A member of the Team first needs to authorize it.

changeset-bot · 2026-03-21T18:46:08Z

🦋 Changeset detected

Latest commit: cea33b8

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
@hyperdx/api	Patch
@hyperdx/app	Patch
@hyperdx/otel-collector	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

github-actions · 2026-03-21T18:47:51Z

PR Review

⚠️ Changeset marks a new feature as patch → Should be minor per semver (new openai provider is new functionality, not a bug fix)
⚠️ AI_REQUEST_HEADERS only wired to OpenAI provider, not Anthropic → Either document this is OpenAI-only (e.g. rename to OPENAI_REQUEST_HEADERS) or also wire it up in getAnthropicModel(); the generic env var name implies it works for any provider
ℹ️ parseJSON<Record<string, string>> is a cast, not a runtime validation → Non-string header values (e.g. {"key": 123}) will silently pass type-check; acceptable for server-side env vars but worth noting

Otherwise the implementation is clean, tests are comprehensive, error messages are helpful, and the pattern follows existing conventions in the codebase.

Enables using OpenAI-compatible endpoints (OpenAI, Azure OpenAI, OpenRouter, LiteLLM proxies) as an alternative to Anthropic for AI-powered query assistance. Adds support for custom headers (X-Client-Id, X-Username) and request body customization via AI_EXTRA_BODY for proxy/gateway compatibility. Enhances error logging with detailed API call diagnostics. Also upgrades sql-formatter to 15.7.0 to address breaking API changes.

- Require `AI_MODEL_NAME` explicitly for the OpenAI provider instead of falling back to an Anthropic model ID (`claude-sonnet-4-5-20250929`) which would fail against actual OpenAI endpoints - Add error handling around `JSON.parse(AI_EXTRA_BODY)` so invalid JSON throws a descriptive config error instead of an unhandled exception - Move `AI_CLIENT_ID`, `AI_USERNAME`, and `AI_EXTRA_BODY` into the config module for consistency with other AI env vars - Truncate `responseBody` in error logs to 500 chars to avoid leaking sensitive data from provider error responses

- Remove accidentally committed package-lock.json (37k lines) and add it to .gitignore -- this is a Yarn 4.5.1 workspace project - Log a warning when AI_EXTRA_BODY custom fetch cannot parse the request body as JSON, instead of silently forwarding it unmodified

This commit introduces a new test file to thoroughly validate the `getAIModel` function. The tests cover various scenarios including provider routing, API key handling, base URL configuration, model name selection, custom headers for OpenAI, and the `AI_EXTRA_BODY` functionality. This ensures robust and predictable behavior for AI model instantiation based on environment variables.

packages/app/tests/e2e/core/navigation.spec.ts

packages/common-utils/src/sqlFormatter.ts

packages/api/src/controllers/ai.ts

.changeset/nine-zoos-buy.md

.gitignore

packages/api/src/routers/api/ai.ts

packages/api/src/controllers/__tests__/ai.test.ts

Replaced `AI_CLIENT_ID` and `AI_USERNAME` environment variables with a single `AI_REQUEST_HEADERS` variable. This new variable accepts a JSON string, allowing users to define arbitrary custom headers for AI provider requests. This change enhances the extensibility of AI provider integrations. Additionally, improved error handling for AI API call failures by providing more concise error messages and removing excessive logging of truncated response bodies.

brandon-pereira · 2026-03-23T22:07:52Z

@vinzee PR LGTM! Nice work! There is some linting issues, otherwise this is good to go :)

vinzee force-pushed the va/enable_ai_assistance branch from c9115f5 to 6506b05 Compare March 21, 2026 18:48

brandon-pereira self-requested a review March 23, 2026 17:36

vinzee added 6 commits March 23, 2026 10:51

Fix linter errors

4f49f7f

fix: add ChangeSet

3e05a19