Merge branch 'aklatzke/AIC-2263/sdk-dx-improvements' of github.com:launchdarkly/python-server-sdk-ai into aklatzke/AIC-2263/sdk-dx-improvements

Author: andrewklatzke

.github/workflows/lint-pr-title.yml

@@ -7,6 +7,9 @@ on:
       - edited
       - synchronize
 
+permissions:
+  pull-requests: read
+
 jobs:
   lint-pr-title:
     uses: launchdarkly/gh-actions/.github/workflows/lint-pr-title.yml@main

.github/workflows/release-please.yml

@@ -50,7 +50,7 @@ jobs:
       package-server-ai-optimization-released: ${{ steps.release.outputs['packages/optimization--release_created'] }}
       package-server-ai-optimization-tag-name: ${{ steps.release.outputs['packages/optimization--tag_name'] }}
     steps:
-      - uses: googleapis/release-please-action@16a9c90856f42705d54a6fda1823352bdc62cf38 # v4
+      - uses: googleapis/release-please-action@45996ed1f6d02564a971a2fa1b5860e934307cf7 # v5.0.0
         id: release
 
   release-server-ai:

.release-please-manifest.json

@@ -1,6 +1,6 @@
 {
-  "packages/sdk/server-ai": "0.18.0",
-  "packages/ai-providers/server-ai-langchain": "0.5.0",
-  "packages/ai-providers/server-ai-openai": "0.4.0",
+  "packages/sdk/server-ai": "1.0.1",
+  "packages/ai-providers/server-ai-langchain": "0.8.0",
+  "packages/ai-providers/server-ai-openai": "0.7.0",
   "packages/optimization": "0.1.0"
 }

packages/ai-providers/server-ai-langchain/CHANGELOG.md

@@ -2,6 +2,59 @@
 
 All notable changes to the LaunchDarkly Python AI LangChain provider package will be documented in this file. This project adheres to [Semantic Versioning](http://semver.org).
 
+## [0.8.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-langchain-0.7.1...launchdarkly-server-sdk-ai-langchain-0.8.0) (2026-05-19)
+
+
+### ⚠ BREAKING CHANGES
+
+* Remove async from create_model/agent/agent_graph methods ([#187](https://github.com/launchdarkly/python-server-sdk-ai/issues/187))
+
+### Features
+
+* Remove async from create_model/agent/agent_graph methods ([#187](https://github.com/launchdarkly/python-server-sdk-ai/issues/187)) ([dddc00a](https://github.com/launchdarkly/python-server-sdk-ai/commit/dddc00ab3d8b57cd58b71b57296680b8e95560fd))
+
+## [0.7.1](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-langchain-0.7.0...launchdarkly-server-sdk-ai-langchain-0.7.1) (2026-05-14)
+
+
+### Bug Fixes
+
+* Make judge runners non-multi-turn ([#185](https://github.com/launchdarkly/python-server-sdk-ai/issues/185)) ([5c21bd0](https://github.com/launchdarkly/python-server-sdk-ai/commit/5c21bd08bebfbdc5672afa93ad99099202955c92))
+
+## [0.7.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-langchain-0.6.0...launchdarkly-server-sdk-ai-langchain-0.7.0) (2026-05-13)
+
+
+### ⚠ BREAKING CHANGES
+
+* Narrow AgentGraphRunner.run input from Any to str ([#177](https://github.com/launchdarkly/python-server-sdk-ai/issues/177))
+* Rename LDAIMetrics.usage and AIGraphMetrics.usage to .tokens ([#175](https://github.com/launchdarkly/python-server-sdk-ai/issues/175))
+* rename GraphMetrics/GraphMetricSummary to AIGraphMetrics/AIGraphMetricSummary ([#173](https://github.com/launchdarkly/python-server-sdk-ai/issues/173))
+
+### Features
+
+* Narrow AgentGraphRunner.run input from Any to str ([#177](https://github.com/launchdarkly/python-server-sdk-ai/issues/177)) ([cc7a0fe](https://github.com/launchdarkly/python-server-sdk-ai/commit/cc7a0fe64549337c71036cd67973bf0af91c7a42))
+* rename GraphMetrics/GraphMetricSummary to AIGraphMetrics/AIGraphMetricSummary ([#173](https://github.com/launchdarkly/python-server-sdk-ai/issues/173)) ([583939d](https://github.com/launchdarkly/python-server-sdk-ai/commit/583939dbe5c11c3c11ed960ee0b46ec377f1bb70))
+* Rename LDAIMetrics.usage and AIGraphMetrics.usage to .tokens ([#175](https://github.com/launchdarkly/python-server-sdk-ai/issues/175)) ([d8c4a70](https://github.com/launchdarkly/python-server-sdk-ai/commit/d8c4a702d4b68f146de7ca520bc35f2aa0b155f6))
+
+## [0.6.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-langchain-0.5.0...launchdarkly-server-sdk-ai-langchain-0.6.0) (2026-05-05)
+
+
+### ⚠ BREAKING CHANGES
+
+* `LangChainModelRunner.invoke_model()` and `invoke_structured_model()` have been removed. Use the unified `run(input, output_type=...)` method instead, which returns `RunnerResult` in place of `ModelResponse` / `StructuredResponse`.
+
+
+### Features
+
+* Add judge evaluation support to agent graphs ([#142](https://github.com/launchdarkly/python-server-sdk-ai/issues/142)) ([3d5a6a9](https://github.com/launchdarkly/python-server-sdk-ai/commit/3d5a6a91a87c7475a83a7e440cd4b71337cfd56f))
+* Migrate LangGraph runner to AgentGraphRunnerResult; clean up legacy shape detection ([#156](https://github.com/launchdarkly/python-server-sdk-ai/issues/156)) ([efa8e00](https://github.com/launchdarkly/python-server-sdk-ai/commit/efa8e00103d3870d379167769ae38f438b019ec4))
+* Support conversation history directly in AI Provider model runners ([#166](https://github.com/launchdarkly/python-server-sdk-ai/issues/166)) ([4bb3e78](https://github.com/launchdarkly/python-server-sdk-ai/commit/4bb3e7813f7c087302ba8446dea6a4a41f012c2e))
+* Update LangChain runners to implement Runner protocol returning RunnerResult ([#150](https://github.com/launchdarkly/python-server-sdk-ai/issues/150)) ([62a8e25](https://github.com/launchdarkly/python-server-sdk-ai/commit/62a8e252f4389884fa2f6a90e325db4a8f79376a))
+
+
+### Bug Fixes
+
+* build judge input as string; strip legacy judge config messages ([#165](https://github.com/launchdarkly/python-server-sdk-ai/issues/165)) ([e6942a6](https://github.com/launchdarkly/python-server-sdk-ai/commit/e6942a6e2d4db17ae1fa6191521f8ac4fb48f30d))
+
 ## [0.5.0](https://github.com/launchdarkly/python-server-sdk-ai/compare/launchdarkly-server-sdk-ai-langchain-0.4.1...launchdarkly-server-sdk-ai-langchain-0.5.0) (2026-04-21)
 
 

packages/ai-providers/server-ai-langchain/README.md

@@ -37,80 +37,69 @@ pip install langchain-google-genai
 ```python
 import asyncio
 from ldclient import LDClient, Config, Context
-from ldai import init
-from ldai_langchain import LangChainProvider
+from ldai import LDAIClient
+from ldai.models import AICompletionConfigDefault, ModelConfig, ProviderConfig
 
 # Initialize LaunchDarkly client
 ld_client = LDClient(Config("your-sdk-key"))
-ai_client = init(ld_client)
-
-# Get AI configuration. Pass a default for improved resiliency when the flag is unavailable or
-# LaunchDarkly is unreachable; omit for a disabled default. Example:
-#   from ldai.models import AICompletionConfigDefault, LDMessage, ModelConfig, ProviderConfig
-#   default = AICompletionConfigDefault(
-#       enabled=True,
-#       model=ModelConfig("gpt-4"),
-#       provider=ProviderConfig("openai"),
-#       messages=[LDMessage(role="system", content="You are a helpful assistant.")]
-#   )
-#   config = ai_client.config("ai-config-key", context, default)
+ai_client = LDAIClient(ld_client)
+
 context = Context.builder("user-123").build()
-config = ai_client.config("ai-config-key", context)
 
 async def main():
-    # Create a LangChain provider from the AI configuration
-    provider = await LangChainProvider.create(config)
-
-    # Use the provider to invoke the model
-    from ldai.models import LDMessage
-    messages = [
-        LDMessage(role="system", content="You are a helpful assistant."),
-        LDMessage(role="user", content="Hello, how are you?"),
-    ]
-    
-    response = await provider.invoke_model(messages)
-    print(response.message.content)
+    # Create a ManagedModel backed by the LangChain provider
+    model = ai_client.create_model(
+        "ai-config-key",
+        context,
+        AICompletionConfigDefault(
+            enabled=True,
+            model=ModelConfig("gpt-4"),
+            provider=ProviderConfig("langchain"),
+        ),
+    )
+
+    if model:
+        result = await model.run("Hello, how are you?")
+        print(result.content)
 
 asyncio.run(main())
 ```
 
 ## Usage
 
-### Using LangChainProvider with the Create Factory
+### Using `create_model` (recommended)
 
-The simplest way to use the LangChain provider is with the static `create` factory method, which automatically creates the appropriate LangChain model based on your LaunchDarkly AI configuration:
+The recommended entry point is `LDAIClient.create_model`, which evaluates a
+LaunchDarkly AI config flag, selects the LangChain runner automatically, and
+returns a `ManagedModel` that wraps the runner:
 
 ```python
-from ldai_langchain import LangChainProvider
-
-# Create provider from AI configuration
-provider = await LangChainProvider.create(ai_config)
+model = ai_client.create_model("ai-config-key", context)
 
-# Invoke the model
-response = await provider.invoke_model(messages)
+if model:
+    result = await model.run("What is feature flagging?")
+    print(result.content)
 ```
 
-### Using an Existing LangChain Model
+### Using the runner directly
 
-If you already have a LangChain model configured, you can use it directly:
+If you need to construct a runner manually (e.g. for testing), you can use
+`LangChainModelRunner` from the `ldai_langchain` package:
 
 ```python
 from langchain_openai import ChatOpenAI
-from ldai_langchain import LangChainProvider
+from ldai_langchain import LangChainModelRunner
 
-# Create your own LangChain model
 llm = ChatOpenAI(model="gpt-4", temperature=0.7)
+runner = LangChainModelRunner(llm)
 
-# Wrap it with LangChainProvider
-provider = LangChainProvider(llm)
-
-# Use with LaunchDarkly tracking
-response = await provider.invoke_model(messages)
+result = await runner.run("Hello!")
+print(result.content)
 ```
 
 ### Structured Output
 
-The provider supports structured output using LangChain's `with_structured_output`:
+Pass a JSON schema dict as `output_type` to request structured output:
 
 ```python
 response_structure = {
@@ -122,92 +111,62 @@ response_structure = {
     "required": ["sentiment", "confidence"],
 }
 
-result = await provider.invoke_structured_model(messages, response_structure)
-print(result.data)  # {"sentiment": "positive", "confidence": 0.95}
+result = await runner.run(messages, output_type=response_structure)
+print(result.parsed)  # {"sentiment": "positive", "confidence": 0.95}
 ```
 
 ### Tracking Metrics
 
-Use the provider with LaunchDarkly's tracking capabilities:
+`ManagedModel.run()` automatically tracks metrics via the associated
+`LDAIConfigTracker`. For manual tracking, use the tracker directly:
 
 ```python
-# Get the AI config with tracker
-config = ai_client.config("ai-config-key", context)
-
-# Create provider
-provider = await LangChainProvider.create(config)
+model = ai_client.create_model("ai-config-key", context)
 
-# Track metrics automatically
-async def invoke():
-    return await provider.invoke_model(messages)
-
-response = await config.tracker.track_metrics_of_async(
-    invoke,
-    lambda r: r.metrics
-)
+if model:
+    result = await model.run("Explain feature flags.")
+    # Metrics are tracked automatically; access them via result.metrics
+    print(result.metrics.tokens)
 ```
 
 ### Static Utility Methods
 
-The `LangChainProvider` class provides several utility methods:
+The `ldai_langchain` helper module provides several utility functions:
 
 #### Converting Messages
 
 ```python
 from ldai.models import LDMessage
-from ldai_langchain import LangChainProvider
+from ldai_langchain.langchain_helper import convert_messages_to_langchain
 
 messages = [
     LDMessage(role="system", content="You are helpful."),
     LDMessage(role="user", content="Hello!"),
 ]
 
-# Convert to LangChain messages
-langchain_messages = LangChainProvider.convert_messages_to_langchain(messages)
+langchain_messages = convert_messages_to_langchain(messages)
 ```
 
 #### Extracting Metrics
 
 ```python
-from ldai_langchain import LangChainProvider
+from ldai_langchain.langchain_helper import get_ai_metrics_from_response
 
 # After getting a response from LangChain
-metrics = LangChainProvider.get_ai_metrics_from_response(ai_message)
+metrics = get_ai_metrics_from_response(ai_message)
 print(f"Success: {metrics.success}")
-print(f"Tokens used: {metrics.usage.total if metrics.usage else 'N/A'}")
+print(f"Tokens used: {metrics.tokens.total if metrics.tokens else 'N/A'}")
 ```
 
 #### Provider Name Mapping
 
 ```python
-# Map LaunchDarkly provider names to LangChain provider names
-langchain_provider = LangChainProvider.map_provider("gemini")  # Returns "google-genai"
-```
-
-## API Reference
+from ldai_langchain.langchain_helper import map_provider_name
 
-### LangChainProvider
-
-#### Constructor
-
-```python
-LangChainProvider(llm: BaseChatModel, logger: Optional[Any] = None)
+# Map LaunchDarkly provider names to LangChain provider names
+langchain_provider = map_provider_name("gemini")  # Returns "google-genai"
 ```
 
-#### Static Methods
-
-- `create(ai_config: AIConfigKind, logger: Optional[Any] = None) -> LangChainProvider` - Factory method to create a provider from AI configuration
-- `convert_messages_to_langchain(messages: List[LDMessage]) -> List[BaseMessage]` - Convert LaunchDarkly messages to LangChain messages
-- `get_ai_metrics_from_response(response: AIMessage) -> LDAIMetrics` - Extract metrics from a LangChain response
-- `map_provider(ld_provider_name: str) -> str` - Map LaunchDarkly provider names to LangChain names
-- `create_langchain_model(ai_config: AIConfigKind) -> BaseChatModel` - Create a LangChain model from AI configuration
-
-#### Instance Methods
-
-- `invoke_model(messages: List[LDMessage]) -> ChatResponse` - Invoke the model with messages
-- `invoke_structured_model(messages: List[LDMessage], response_structure: Dict[str, Any]) -> StructuredResponse` - Invoke with structured output
-- `get_chat_model() -> BaseChatModel` - Get the underlying LangChain model
-
 ## Documentation
 
 For full documentation, please refer to the [LaunchDarkly AI SDK documentation](https://docs.launchdarkly.com/sdk/ai/python).

packages/ai-providers/server-ai-langchain/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "launchdarkly-server-sdk-ai-langchain"
-version = "0.5.0"
+version = "0.8.0"
 description = "LaunchDarkly AI SDK LangChain Provider"
 authors = [{name = "LaunchDarkly", email = "dev@launchdarkly.com"}]
 license = {text = "Apache-2.0"}
@@ -20,7 +20,7 @@ classifiers = [
     "Topic :: Software Development :: Libraries",
 ]
 dependencies = [
-    "launchdarkly-server-sdk-ai>=0.18.0",
+    "launchdarkly-server-sdk-ai>=1.0.1", # x-release-please-version
     "langchain-core>=1.0.0",
     "langchain>=1.0.0",
 ]