chore: release main

[github-actions]

🤖 I have created a release beep boop

launchdarkly-ai-server: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• client: once the handlers move onto this layer, prompt and completion content will be absent from spans unless the caller passes capture_content=True. Anyone reading gen_ai.prompt.0.content today will need to opt in.

Features

• add evaluations module scaffold, credentials, LD API client, result types (80fd4e4)
• AIC-3230: emit evaluation context identity (#65) (ca30e83)
• client: cache-aware token accounting and a reusable span lifecycle (f54ef59)
• client: end spans a cancelled run would otherwise strand (dff27f6)
• client: put conversation content on spans behind an opt-in flag (2bb8310)
• client: tell a cancelled stream apart from an abandoned one (14ed066)
• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• emit conversation id and judge evals (1b0f5bd)
• emit evaluation context identity on feature_flag spans (97528e3)
• emit gen_ai.conversation.id (#42) (c372e56)
• evaluations: configure run link UI base (4cbcbd6)
• evaluations: emit generation events (311b012)
• evaluations: event-based offline evaluations runner (#39) (4239739)
• evaluations: make summary poll interval and timeout configurable per run (cd84352)
• evaluations: print event emission timestamps (0079289)
• evaluations: require an SDK key and fail fast when it is missing (a1ba71e)
• gate evaluation generation result ingest (0ea442f)
• gate the judge explanation on capture_content (8a66365)
• record judge scores as gen_ai.evaluation.result (2604816)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)
• run client-side evaluations from the SDK (50fee37)

Bug Fixes

• AIC-3230: align user context canonical identity (68a5c2b)
• align evaluations with staging dataset and run APIs (28f8eb1)
• bind conversation id at stream() call time (bda8393)
• client: a tool that returned nothing does not return null (eb61eb2)
• client: carry the cache breakdown into the public UsageDict (c60fa7c)
• client: guard the fourth content writer, and assert the family rule (d637b95)
• client: keep the plain text a LangChain message holds in a list (c11adca)
• client: make end_span_once a no-op on a None span (a1772c1)
• client: read a LangChain ChatMessage's own speaker (d56103a)
• client: read a LangChain message's role from the field, not only the method (641f8e1)
• client: stop the content carriers disagreeing, and stop them crashing (b32d3fc)
• delay only LaunchDarkly invoke_agent spans during judge evals (af5a92c)
• evaluations: accept a BYOC client without an SDK key and reject NaN poll values (ab5a4e5)
• evaluations: avoid replaying non-idempotent POSTs and blocking the event loop (485dc85)
• evaluations: derive result from summary (a7acc68)
• evaluations: emit direct generation output (b98b643)
• evaluations: flatten generation event payload (a0254d0)
• evaluations: include evaluation run event ID (8cae0e4)
• evaluations: nest generation usage tokens (c88e9ae)
• evaluations: poll run summary to terminal state (75d3072)
• evaluations: return pending summaries promptly (ec3e290)
• evaluations: use API run source (433273a)
• return JudgeResult objects from inline judges (c9e20a2)
• scope the processor, keep streaming parenting, accept a nullish id (9794d36)
• stop exporting judge reasoning, validate the score, freeze the end time (8fa4f6f)
• use tool keys in tool call telemetry (841ba77)
• use tool keys in tool call telemetry (#56) (ae3130f)

Documentation

• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-python: 0.1.4

0.1.4 (2026-09-03)

Features

• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• evaluations: configure run link UI base (4cbcbd6)
• evaluations: emit generation events (311b012)
• evaluations: event-based offline evaluations runner (#39) (4239739)
• run client-side evaluations from the SDK (50fee37)

Bug Fixes

• evaluations: accept a BYOC client without an SDK key and reject NaN poll values (ab5a4e5)

launchdarkly-ai-claude-agents: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• claude-agents: stop an empty tool list switching off the built-ins
• claude-agents: the span this handler emits is renamed from claude.query and claude.query.stream to invoke_agent. Queries selecting on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True.

Features

• claude-agents: emit invoke_agent, chat and execute_tool spans (7831c81)
• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• emit conversation id and judge evals (1b0f5bd)
• emit gen_ai.conversation.id (#42) (c372e56)
• gate the judge explanation on capture_content (8a66365)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• bind conversation id at stream() call time (bda8393)
• claude-agents: an abandoned stream is not an error for its tool spans (10b8158)
• claude-agents: an empty run must not claim it cost nothing (85c73ff)
• claude-agents: keep a tool span trackable when its content will not serialise (0738606)
• claude-agents: keep the spans of a cancelled run (b7dddfa)
• claude-agents: put the input content write inside the guard too (f0160c2)
• claude-agents: report a cancelled stream as cancelled, not abandoned (d4f1b6f)
• claude-agents: stop an empty tool list switching off the built-ins (50e82fd)
• scope the processor, keep streaming parenting, accept a nullish id (9794d36)
• use tool keys in tool call telemetry (841ba77)
• use tool keys in tool call telemetry (#56) (ae3130f)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-claude-messages: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• claude-messages: the span this handler emits is renamed from claude.messages and claude.messages.stream to invoke_agent. Queries that select on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True.

Features

• claude-messages: emit invoke_agent, chat and execute_tool spans (d4b5b6e)
• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• gate the judge explanation on capture_content (8a66365)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• claude-messages: close the tool span a cancelled tool leaves open (4f23cf4)
• claude-messages: do not invent the cache figures Anthropic did not send (db417b1)
• claude-messages: fail the streaming chat span instead of abandoning it (2f0e565)
• claude-messages: forward capture_content from the convenience wrapper (b7531d3)
• claude-messages: guard the streaming chat span's input write (6e358dd)
• claude-messages: keep every blocking chat-span write inside the guard (21fb591)
• claude-messages: keep the spans of a cancelled run (bb2e52f)
• claude-messages: keep the streaming turn's billed tokens on a content failure (1933374)
• claude-messages: put the input content write inside the guard too (4aaa497)
• claude-messages: record a tool result inside the guard that ends its span (19d932e)
• claude-messages: report a cancelled stream as cancelled, not abandoned (2d3379a)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-openai-agents: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• openai-agents: the span this handler emits is renamed from openai.agent.run and openai.agent.run.stream to invoke_agent. Queries selecting on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True.

Features

• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• gate the judge explanation on capture_content (8a66365)
• openai-agents: emit invoke_agent, chat and execute_tool spans (c5331eb)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• openai-agents: cancel the Runner even when there are no spans (d9caaac)
• openai-agents: close the tool span when the SDK reports no call id (1d66254)
• openai-agents: end the chat span the hook already stopped tracking (9aecf05)
• openai-agents: forward capture_content from the convenience wrapper (a2440e4)
• openai-agents: keep a tool span trackable when its content will not serialise (32eb830)
• openai-agents: keep the spans of a cancelled run (063a392)
• openai-agents: put the input content write inside the guard too (ead8ed5)
• openai-agents: report a cancelled stream as cancelled, not abandoned (2b246d0)
• openai-agents: report tool arguments as an object, not a JSON string (c646b6a)
• openai-agents: stop double-counting a failed run's token spend (8991dd5)
• openai-agents: stop tying token accounting to span creation (0aa7241)
• openai-agents: take the run total from the Runner, not from the hooks (64cd458)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-openai-messages: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• openai-messages: the span this handler emits is renamed from openai.response and openai.response.stream to invoke_agent. Queries selecting on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True.

Features

• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• gate the judge explanation on capture_content (8a66365)
• openai-messages: emit invoke_agent, chat and execute_tool spans (17ad375)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• openai-messages: close the tool span a cancelled tool leaves open (35a254d)
• openai-messages: drop output items that produced no content (66bacd0)
• openai-messages: fail the streaming chat span, and keep its tokens (83f8684)
• openai-messages: forward capture_content from the convenience wrapper (495391f)
• openai-messages: keep every chat-span write inside the guard that ends it (a4fed91)
• openai-messages: keep the spans of a cancelled run (c93a902)
• openai-messages: keep the tokens a turn already cost when its content fails (8501379)
• openai-messages: record a tool result inside the guard that ends its span (6b9813b)
• openai-messages: report a cancelled stream as cancelled, not abandoned (76c4cef)
• openai-messages: report tool arguments as an object, not a JSON string (a6b9da9)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-langchain-agents: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• langchain-agents: the span this handler emits is renamed from langchain.agent and langchain.agent.stream to invoke_agent. Queries selecting on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True. gen_ai.system changes value; see below.

Features

• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• gate the judge explanation on capture_content (8a66365)
• langchain-agents: emit invoke_agent, chat and execute_tool spans (7683c9d)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• langchain-agents: a callback that raises must not leave a span unreachable (2169a05)
• langchain-agents: a reported zero token count is not a missing one (0d5e38c)
• langchain-agents: an abandoned stream marks its open spans, it does not fail them (d087d73)
• langchain-agents: forward capture_content from the convenience wrapper (06b210d)
• langchain-agents: keep a turn's billed tokens when its content write fails (485db67)
• langchain-agents: keep the spans of a cancelled run (ff53cc6)
• langchain-agents: make a successful run's total agree with its own spans (6333808)
• langchain-agents: put the input content write inside the guard too (0fa4ed1)
• langchain-agents: record the root completion the way the chat span does (1f4242b)
• langchain-agents: report a cancelled stream as cancelled, not abandoned (010a9ca)
• langchain-agents: total the streaming root from the callbacks when the messages report nothing (6ff5a94)
• langchain-agents: track the chat span before writing content that can raise (883da7b)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

launchdarkly-ai-langchain-messages: 0.2.0

0.2.0 (2026-09-03)

⚠ BREAKING CHANGES

• langchain-messages: the span this handler emits is renamed from langchain.invoke and langchain.stream to invoke_agent. Queries selecting on the old names will not match. Prompt and completion content is no longer on spans unless the caller passes capture_content=True. gen_ai.system changes value; see below.

Features

• emit $ld:ai:sdk:info event per AI package (#62) (65136a3)
• gate the judge explanation on capture_content (8a66365)
• langchain-messages: emit invoke_agent, chat and execute_tool spans (51032b5)
• record judge scores as gen_ai.evaluation.result (#43) (7c8b0c9)

Bug Fixes

• langchain-messages: close the tool span a cancelled tool leaves open (53c7f9a)
• langchain-messages: do not serialise the output nobody asked for (0c9a5dd)
• langchain-messages: fail the streaming chat span, and keep its tokens (b084536)
• langchain-messages: forward capture_content from the convenience wrapper (79d8030)
• langchain-messages: guard the per-turn chat span's input write too (2b901b4)
• langchain-messages: guard the streaming root's output write (178891a)
• langchain-messages: keep every structured-turn write inside the guard (de76f8b)
• langchain-messages: keep the spans of a cancelled run (9c30fb3)
• langchain-messages: one structured strategy, and don't report zeros as spend (d7129d4)
• langchain-messages: put the input content write inside the guard too (a68fdff)
• langchain-messages: record a tool result inside the guard that ends its span (72c7481)
• langchain-messages: report a cancelled stream as cancelled, not abandoned (9b8e10d)
• langchain-messages: two more ways a run misreported itself (ad90d70)
• langchain-messages: two ways a run could vanish from its own trace (b8bdfaf)

Documentation

• describe the span tree the SDK now emits (bc7815a)
• describe the span tree the SDK now emits (#37) (6a82ef5)

---

This PR was generated with Release Please. See documentation.

---

Note

Overview
This is a Release Please cut that bumps published package versions and appends 0.2.0 / 0.1.4 changelog sections—no application logic changes in the diff itself.

launchdarkly-ai-server (packages/client) goes to 0.2.0, documenting client-side evaluations (offline runner, generation events, LD API wiring), richer telemetry (gen_ai.conversation.id, judge scores, $ld:ai:sdk:info), and shared span lifecycle/token accounting. Breaking: prompt/completion on spans now requires capture_content=True (e.g. gen_ai.prompt.0.content is opt-in).

Integration packages (claude-*, openai-*, langchain-*) bump to 0.2.0 with aligned changelog notes: root spans rename from provider-specific names to invoke_agent, nested chat / execute_tool spans, and the same content opt-in. LangChain handlers also note gen_ai.system value changes.

launchdarkly-ai-python (packages/ai) bumps to 0.1.4 (barrel re-export of server changes). .release-please-manifest.json and each package’s pyproject.toml / __version__ are updated to match.

^{Reviewed by Cursor Bugbot for commit 1108cfe. Bugbot is set up for automated code reviews on this repo. Configure here.}