CHANGELOG.md•5.26 kB
# @arizeai/phoenix-client
## 5.2.0
### Minor Changes
- f9d8b06: switch licensing to apache 2
## 5.1.1
### Patch Changes
- Updated dependencies [8bbff3a]
- @arizeai/phoenix-otel@0.2.1
## 5.1.0
### Minor Changes
- de6f111: refactor to use phoenix-otel across the client
### Patch Changes
- Updated dependencies [de6f111]
- @arizeai/phoenix-otel@0.2.0
## 5.0.0
### Major Changes
- 950fda5: feat: Add support for dataset splits
This release introduces support for dataset splits, enabling you to segment and query specific portions of your dataset examples. The `DatasetSelector` interface has been enhanced to support filtering by splits, allowing for more granular dataset management and experimentation.
## New Features
- **Dataset Splits Support**: Query dataset examples by split using the enhanced `DatasetSelector` interface
- **Split-based Experimentation**: Run experiments on specific dataset splits for targeted evaluation
- **Enhanced Dataset Types**: Updated type definitions to support split-based dataset operations
## Breaking Changes
- **`runExperiment` API Changes**:
- The `datasetVersionId` parameter has been removed from `runExperiment`
- Version selection is now handled through the `DatasetSelector` interface
- Pass `versionId` and `splits` as properties of the `DatasetSelector` argument instead
## Migration Guide
**Before:**
```typescript
runExperiment({
dataset: { datasetId: "my-dataset" },
datasetVersionId: "version-123",
// ... other params
});
```
**After:**
```typescript
runExperiment({
dataset: {
datasetId: "my-dataset",
versionId: "version-123",
splits: ["train", "test"],
},
// ... other params
});
```
## 4.2.0
### Minor Changes
- 85430fa: feat: Add configurable DiagLogLevel to runExperiment
- c7cc7d9: feat: Add createOrGetDataset helper function to phoenix-client
Additionally clean up build artifacts and type-checking amongst example scripts.
## 4.1.0
### Minor Changes
- 2981780: add session annotation functions
## 4.0.3
### Patch Changes
- e3a8ce2: pass through the tracer provider to experiments so that there is no need to configure it twice
- c85780b: Add support for generics across evals and experiments
## 4.0.2
### Patch Changes
- 1b71c66: make sure repetion numbers are greater than 0
## 4.0.1
### Patch Changes
- e72a9ad: don't swallow errors, allow for incomplete datasets (e.g. just imputs)
## 4.0.0
### Major Changes
- 7732f99: Breaking change for AI SDK users. Support for messages conversion for the AI SDK 5
## 3.2.0
### Minor Changes
- 4f43901: add support for logging document annotations
## 3.1.0
### Minor Changes
- ee0c829: switch to batch span processor by default and make it configurable
## 3.0.0
### Major Changes
- 3e80a50: delete span method
### Minor Changes
- 8711bde: update major version of openai to ^5
## 2.4.0
### Minor Changes
- fe55fc5: get dataset with versionId
## 2.3.5
### Patch Changes
- 83748e6: add type exports and better documentation across packages
## 2.3.4
### Patch Changes
- 20db91d: Add tracing to evals, add tracing controls
## 2.3.3
### Patch Changes
- 2609fcd: bump target JS to es2017 for native async
## 2.3.2
### Patch Changes
- 6ef8e47: fix dataset pull by name
## 2.3.1
### Patch Changes
- b3e30db: simplify types for task output in experiments client api
## 2.3.0
### Minor Changes
- 3c97cc7: Add the ability to get a dataset by name
## 2.2.0
### Minor Changes
- 1906611: add getSpan method
### Patch Changes
- 4c52db4: ollama provider added
## 2.1.1
### Patch Changes
- 5dd53be: add in xai to playground as provider
## 2.1.0
### Minor Changes
- b162720: add support for deepseek
## 2.0.1
### Patch Changes
- da7800a: feat(phoenix-client): Log the experiement/dataset link when calling runExperiment
## 2.0.0
### Major Changes
- 118e881: feat: add support for dataset creation and the ability to re-run experiments
## 1.3.0
### Minor Changes
- 536258e: feat(phoenix-client): Export traces from experiments to Phoenix
## 1.2.0
### Minor Changes
- f7fae3b: feat(phoenix-client): Record experiment results to Phoenix server
- 9273417: feat: Enqueue experiment runs
- 4dd23c8: support for annotation logging on spans
## 1.1.0
### Minor Changes
- fff5511: feat: Update openapi schema with new endpoints
## 1.0.2
### Patch Changes
- c99ee6f: Update type definitions to include max_completion_tokens openai parameter
## 1.0.1
### Patch Changes
- 2ffeb64: fix: Remove runtime dependency on `ai` package
## 1.0.0
### Major Changes
- 3f9e392: feat: Add support for Phoenix Prompts
Phoenix can now manage Prompts, and the `@arizeai/phoenix-client` package has been updated to support this.
In this initial release, we support the following:
- Fully typed Prompt REST endpoints
- Prompt Pulling
- Converting a Prompt to invocation parameters for the following LLM SDKs:
- OpenAI
- Anthropic
- Vercel AI SDK
- You can use any of the Vercel AI SDK Providers with your prompt
### Patch Changes
- 95bfc7c: Add the ability to push prompts via the typescript client sdk
## 0.0.1
### Patch Changes
- 76a9cdf: pre-release of phoenix-client