Skip to content

Releases

openai v1.1.0

Choose a tag to compare

@github-actions github-actions released this 14 Nov 07:39

Added

  • Fast mode for OpenAI API-key connections and OpenAI Codex, including /fast controls and response-tier status.
  • Optional openai.optimizeLatency setting to reduce repeated history uploads and request latency in long conversations when signed in with ChatGPT.

webfetch v1.1.7

Choose a tag to compare

@github-actions github-actions released this 11 Nov 09:43

Fixed

  • Restored direct MP4 links for X/Twitter video posts after X changed its page data format.

websearch v1.3.4

Choose a tag to compare

@github-actions github-actions released this 06 Nov 10:39

Fixed

  • Allow OpenAI image generation settings and other unrelated OpenAI options without reporting a native-search configuration error.

openai v1.0.0

Choose a tag to compare

@github-actions github-actions released this 06 Nov 10:39

Added

  • Image generation and editing through Codex OAuth or the OpenAI Images API, available with any chat model.
  • Batch generation, output formats, sizes and editing options supported by the selected backend and model.
  • Save images to new files without overwriting existing files, report actual raster dimensions, and support cancellation.

Changed

  • Image generation is available as a replacement for the deprecated @pi-lab/codex-image CLI integration. Configure openai.imageGeneration and remove the old extension before reloading.

webfetch v1.1.6

Choose a tag to compare

@github-actions github-actions released this 02 Nov 05:44

Fixed

  • Respect PI_CODING_AGENT_DIR when saving binary downloads.

Changed

  • Local installations no longer require a separate build.

Known issues

  • Some X/Twitter video posts omit direct MP4 links even though their inline scripts contain video URLs. Page content and inline script access remain available.

env v1.1.0

Choose a tag to compare

@github-actions github-actions released this 02 Nov 05:44

Added

  • Reference existing environment variables in settings with ${VAR}, use literal defaults with ${VAR:-default}, and escape references with $${VAR}. Missing references remain literal and warn without exposing values.

Fixed

  • Respect PI_CODING_AGENT_DIR for dotenv files, including the legacy fallback location.
  • Ignore project environment settings when project trust cannot be confirmed.

Changed

  • Local installations no longer require a separate build.

websearch v1.3.3

Choose a tag to compare

@github-actions github-actions released this 01 Nov 14:40

Fixed

  • Honor PI_CODING_AGENT_DIR when reading user-level provider and OpenAI native-search settings. (#11)

websearch v1.3.2

Choose a tag to compare

@github-actions github-actions released this 24 Oct 17:44

Fixed

  • Read OpenAI native-search settings from Pi's user and project settings.json files so standalone search yields to native search consistently with @pi-lab/openai.

openai v0.1.2

Choose a tag to compare

@github-actions github-actions released this 24 Oct 17:44

Changed

  • Read the openai section from Pi's user and project settings.json files instead of separate openai.json files.

Fixed

  • Apply Fast mode consistently to eligible requests and omit the service tier when Fast mode is off.
  • Preserve native search actions and citations across subsequent requests and session reloads without changing assistant answers.
  • Retain streamed search records even when the final response contains an empty output array, and prevent overlapping requests from mixing their records.
  • Include recorded native search actions and available source data in context-compaction input while respecting branch boundaries and context edits.

websearch v1.3.0

Choose a tag to compare

@github-actions github-actions released this 23 Oct 16:39

Changed

  • Publish TypeScript source directly for Pi to load without compilation; replace the build-time bundler with standard bundledDependencies carrying private utils source, excluding tests.
  • Withdraw standalone websearch while OpenAI native search is enabled for the current model, independently reading the same configuration and applying matching eligibility rules without extension communication.