feat!: remove deprecated client method aliases and type aliases

[jsonbailey]

Summary

feat!: remove LDAIClient.config — use LDAIClient.completion_config instead
feat!: remove LDAIClient.create_chat — use LDAIClient.create_model instead
feat!: remove LDAIClient.agent — use LDAIClient.agent_config instead
feat!: remove LDAIClient.agents — use LDAIClient.agent_configs instead
feat!: remove LDAIConfigTracker.track_openai_metrics — use tracker.track_metrics_of(get_ai_metrics_from_response, fn) from launchdarkly-server-sdk-ai-openai instead
feat!: remove LDAIMetricSummary.duration — use LDAIMetricSummary.duration_ms instead
feat!: remove Chat class and the ldai.chat module — use the managed-runner / provider model runners instead
feat!: remove LDAIAgent alias — use AIAgentConfig instead
feat!: remove LDAIAgentConfig alias — use AIAgentConfig instead
feat!: remove LDAIAgentDefaults alias — use AIAgentConfigDefault instead
feat!: remove public AIConfig re-export — internal-only; use AICompletionConfig / AIAgentConfig / AIJudgeConfig instead
feat!: remove evaluation_metric_keys (plural) field from AIJudgeConfigDefault and AIJudgeConfig — use evaluation_metric_key (singular) instead; the read path still accepts the plural in flag variations for backward compatibility

Removes the following deprecated symbols from the launchdarkly-server-sdk-ai package ahead of the next release. Each replacement has existed and been recommended for some time; this PR drops the back-compat shims.

Removed	Replacement
LDAIClient.config()	LDAIClient.completion_config()
LDAIClient.create_chat()	LDAIClient.create_model()
LDAIClient.agent()	LDAIClient.agent_config()
LDAIClient.agents()	LDAIClient.agent_configs()
LDAIConfigTracker.track_openai_metrics()	LDAIConfigTracker.track_metrics_of with get_ai_metrics_from_response from ldai_openai
LDAIMetricSummary.duration	LDAIMetricSummary.duration_ms
Chat class + ldai/chat/ shim package	managed-runner / provider model runners
LDAIAgent type alias	AIAgentConfig
LDAIAgentConfig type alias	AIAgentConfig
LDAIAgentDefaults type alias	AIAgentConfigDefault
Public re-export of AIConfig from ldai	AIConfig is now internal-only (still defined as the base class in ldai.models)
evaluation_metric_keys (plural) field on AIJudgeConfig / AIJudgeConfigDefault	evaluation_metric_key (singular)

The corresponding tests have been updated to call the replacement APIs. The two dedicated track_openai_metrics tests were removed (their behavior is covered by track_metrics_of tests).

BREAKING CHANGE

This is a breaking release. Callers using any of the removed names must migrate to the replacements listed above. The ldai.chat import path is gone entirely — from ldai.chat import Chat will now fail.

Note on evaluation_metric_keys

The plural-form evaluation_metric_keys field is removed from both AIJudgeConfig and AIJudgeConfigDefault. to_dict() now only emits the singular evaluationMetricKey. To preserve backward compatibility with older LD-side judge flag variations that still set the plural, the read path in client.py (_extract_evaluation_metric_key, lines 219–232) falls back to evaluationMetricKeys[0] when only the plural is present. New configs cannot set the plural in code, but existing flag values continue to resolve correctly to the singular evaluation_metric_key.

Test plan

• make test — full unit test suite (server-ai, langchain, openai) passes (328 tests)
• make lint — mypy, isort, pycodestyle pass for all three packages

---

Note

Medium Risk
This is a breaking change that removes previously deprecated public APIs and re-exports, so downstream consumers will fail at import/runtime until migrated. Behavior is largely unchanged, but the surface-area reduction could impact integrations that relied on the deprecated names.

Overview
Removes multiple deprecated, backward-compatibility shims from launchdarkly-server-sdk-ai, tightening the public API surface.

Public exports and aliases were dropped, including ldai.chat.Chat, LDAIClient.config/create_chat/agent/agents, LDAIConfigTracker.track_openai_metrics, LDAIMetricSummary.duration, and legacy type aliases like LDAIAgent* plus the AIConfig re-export.

Unit tests were updated accordingly to use completion_config, create_model, agent_config/agent_configs, and duration_ms, and the dedicated track_openai_metrics tests were removed.

^{Reviewed by Cursor Bugbot for commit e50041e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default mode and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit e50041e. Configure here.}