|
2 | 2 |
|
3 | 3 | All notable changes to the LaunchDarkly Python AI package will be documented in this file. This project adheres to [Semantic Versioning](http://semver.org). |
4 | 4 |
|
| 5 | +## [0.18.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-0.17.0...launchdarkly-server-sdk-ai-0.18.0) (2026-04-21) |
| 6 | + |
| 7 | + |
| 8 | +### ⚠ BREAKING CHANGES |
| 9 | + |
| 10 | +* Add per-execution runId, at-most-once tracking, and cross-process tracker resumption ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) |
| 11 | +* `AICompletionConfig`, `AIAgentConfig`, and `AIJudgeConfig`: `tracker` field replaced with `create_tracker` factory callable ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) |
| 12 | +* `ManagedModel`, `ManagedAgent`, `Judge`, and `ManagedAgentGraph` constructors no longer accept a tracker parameter; tracker is now created on-demand via `create_tracker` ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) |
| 13 | +* `ManagedAgentGraph.get_tracker()` method removed ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) |
| 14 | +* rename track_latency to track_duration on AIGraphTracker ([#138](https://github.com/launchdarkly/python-server-sdk-ai/issues/138)) |
| 15 | +* `graph_key` keyword argument removed from all `LDAIConfigTracker.track_*()` methods; trackers retrieved via a graph definition are automatically configured with the correct graph key ([#134](https://github.com/launchdarkly/python-server-sdk-ai/issues/134)) |
| 16 | +* Flatten JudgeResponse and EvalScore into new JudgeResult ([#132](https://github.com/launchdarkly/python-server-sdk-ai/issues/132)) |
| 17 | +* `LDAIConfigTracker.track_eval_scores()` and `track_judge_response()` removed and consolidated into `track_judge_result()` ([#132](https://github.com/launchdarkly/python-server-sdk-ai/issues/132)) |
| 18 | +* `AIGraphTracker.track_judge_response()` method removed ([#132](https://github.com/launchdarkly/python-server-sdk-ai/issues/132)) |
| 19 | +* `Judge.evaluate()` and `evaluate_messages()` now always return a `JudgeResult` object; use `result.sampled` to check if evaluation was actually executed ([#132](https://github.com/launchdarkly/python-server-sdk-ai/issues/132)) |
| 20 | + |
| 21 | +### Features |
| 22 | + |
| 23 | +* Add per-execution runId, at-most-once tracking, and cross-process tracker resumption ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) ([68685cd](https://github.com/launchdarkly/python-server-sdk-ai/commit/68685cd9623105b0b01dd57942538c047615f4f2)) |
| 24 | +* Flatten JudgeResponse and EvalScore into new JudgeResult ([#132](https://github.com/launchdarkly/python-server-sdk-ai/issues/132)) ([af4e463](https://github.com/launchdarkly/python-server-sdk-ai/commit/af4e46332c5d9a668a7172012c89871d0fb90b56)) |
| 25 | +* Move graph_key to AIConfigTracker instantiation ([#134](https://github.com/launchdarkly/python-server-sdk-ai/issues/134)) ([20fff24](https://github.com/launchdarkly/python-server-sdk-ai/commit/20fff24fcd02aa101d7f9a6c21dc6a25e7916a1c)) |
| 26 | +* rename track_latency to track_duration on AIGraphTracker ([#138](https://github.com/launchdarkly/python-server-sdk-ai/issues/138)) ([05758a7](https://github.com/launchdarkly/python-server-sdk-ai/commit/05758a735db0c2defc4e08d02282a46f559220e5)) |
| 27 | +* Add `LDAIConfigTracker.resumption_token()` and `from_resumption_token()` methods to continue tracking a run from a separate process ([#133](https://github.com/launchdarkly/python-server-sdk-ai/issues/133)) ([68685cd](https://github.com/launchdarkly/python-server-sdk-ai/commit/68685cd9623105b0b01dd57942538c047615f4f2)) |
| 28 | + |
5 | 29 | ## [0.17.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-0.16.1...launchdarkly-server-sdk-ai-0.17.0) (2026-04-02) |
6 | 30 |
|
7 | 31 |
|
|
0 commit comments