chore(deps): bump @posthog/ai from 8.7.1 to 8.10.1

[dependabot]

Bumps @posthog/ai from 8.7.1 to 8.10.1.

Release notes

Sourced from @​posthog/ai's releases.

@​posthog/ai@​8.10.1

8.10.1

Patch Changes

• #4736 c16a7ce Thanks @​bernatixer! - Only terminal Responses API statuses become $ai_stop_reason (a queued or in-progress background run no longer records a lifecycle state as its stop reason), and the native OpenAI wrapper now names a truncated run by incomplete_details.reason (e.g. max_output_tokens) instead of the bare incomplete, matching the LangChain callback (2026-09-03)
• Updated dependencies [dbbb58e]:
  • @​posthog/core@​1.50.3

@​posthog/ai@​8.10.0

8.10.0

Minor Changes

• #4687 657b173 Thanks @​AhmadHammad21! - Add a Google ADK (@google/adk) observability adapter. @posthog/ai/adk exposes PostHogADKPlugin, an ADK BasePlugin that captures a full $ai_generation event (input, output, model, token usage, latency, finish reason, trace id, session id, distinct id and groups) for every model call an ADK agent makes, funnelling through the shared captureAiGeneration primitive so PostHog derives cost from the model and tokens. (2026-09-02)

@​posthog/ai@​8.9.3

8.9.3

Patch Changes

• #4700 a105b77 Thanks @​bernatixer! - Capture $ai_stop_reason from LangChain runs that use the OpenAI Responses API.

  The callback only understood Chat Completions vocabulary (finish_reason / stop_reason), so Responses API runs, which report status and incomplete_details.reason instead, never carried a stop reason. The Chat Completions keys keep priority, and incomplete_details.reason outranks status, so an early stop names its cause (for example max_output_tokens) instead of just incomplete. (2026-09-01)

• Updated dependencies [e87820a]:

  • @​posthog/core@​1.50.1

@​posthog/ai@​8.9.2

8.9.2

Patch Changes

• #4664 86fe9da Thanks @​bernatixer! - Interrupted or cancelled streams now report the token usage and latency they actually observed, instead of zeros, across the OpenAI, Anthropic, Gemini, Azure and Vercel wrappers. When usage was never reported, token counts and override costs are omitted entirely, so $ai_input_tokens can be absent where it was previously always 0. (2026-09-01)

@​posthog/ai@​8.9.1

8.9.1

Patch Changes

• #4698 5ae5c29 Thanks @​dependabot! - Preserve Anthropic workspace IDs on wrapped provider promises and support @anthropic-ai/sdk versions from 0.112.3 through 0.120.x. (2026-09-01)
• Updated dependencies [444bf35, 6ae173f, 6723395]:
  • @​posthog/core@​1.50.0
  • posthog-node@5.51.5

Changelog

Sourced from @​posthog/ai's changelog.

8.10.1

Patch Changes

• #4736 c16a7ce Thanks @​bernatixer! - Only terminal Responses API statuses become $ai_stop_reason (a queued or in-progress background run no longer records a lifecycle state as its stop reason), and the native OpenAI wrapper now names a truncated run by incomplete_details.reason (e.g. max_output_tokens) instead of the bare incomplete, matching the LangChain callback (2026-09-03)
• Updated dependencies [dbbb58e]:
  • @​posthog/core@​1.50.3

8.10.0

Minor Changes

• #4687 657b173 Thanks @​AhmadHammad21! - Add a Google ADK (@google/adk) observability adapter. @posthog/ai/adk exposes PostHogADKPlugin, an ADK BasePlugin that captures a full $ai_generation event (input, output, model, token usage, latency, finish reason, trace id, session id, distinct id and groups) for every model call an ADK agent makes, funnelling through the shared captureAiGeneration primitive so PostHog derives cost from the model and tokens. (2026-09-02)

8.9.3

Patch Changes

• #4700 a105b77 Thanks @​bernatixer! - Capture $ai_stop_reason from LangChain runs that use the OpenAI Responses API.

  The callback only understood Chat Completions vocabulary (finish_reason / stop_reason), so Responses API runs, which report status and incomplete_details.reason instead, never carried a stop reason. The Chat Completions keys keep priority, and incomplete_details.reason outranks status, so an early stop names its cause (for example max_output_tokens) instead of just incomplete. (2026-09-01)

• Updated dependencies [e87820a]:

  • @​posthog/core@​1.50.1

8.9.2

Patch Changes

• #4664 86fe9da Thanks @​bernatixer! - Interrupted or cancelled streams now report the token usage and latency they actually observed, instead of zeros, across the OpenAI, Anthropic, Gemini, Azure and Vercel wrappers. When usage was never reported, token counts and override costs are omitted entirely, so $ai_input_tokens can be absent where it was previously always 0. (2026-09-01)

8.9.1

Patch Changes

• #4698 5ae5c29 Thanks @​dependabot! - Preserve Anthropic workspace IDs on wrapped provider promises and support @anthropic-ai/sdk versions from 0.112.3 through 0.120.x. (2026-09-01)
• Updated dependencies [444bf35, 6ae173f, 6723395]:
  • @​posthog/core@​1.50.0
  • posthog-node@5.51.5

8.9.0

Minor Changes

• #4556 a47b232 Thanks @​gouveags! - Add LangChain v1 agent middleware for AI observability. (2026-08-27)

... (truncated)

Commits

• 1e5ddbe chore: update versions and lockfile [version bump]
• e9a47d5 chore: migrate unit tests from Jest to Vitest (#4739)
• c90699a chore: migrate AI and web builds to tsdown (#4746)
• c16a7ce fix(ai): only capture terminal Responses statuses as stop reasons (#4736)
• 285a2cb chore: migrate remaining lint rules to Oxlint (#4738)
• 1278ba2 chore: update versions and lockfile [version bump]
• 657b173 feat(ai): add Google ADK (@​google/adk) observability adapter (#4687)
• 3fabfbc chore: adopt oxfmt and oxlint incrementally (#4722)
• 9a79809 chore: update versions and lockfile [version bump]
• a105b77 fix(ai): capture the stop reason from LangChain Responses API runs (#4700)
• Additional commits viewable in compare view

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[greptile-apps]

PR author is in the excluded authors list.

[github-actions]

React Doctor skipped this pull request — it changed no React files.

_{Reviewed by React Doctor for commit 91972e9.}