inferkit 0.1.3

Pure-Dart OpenAI-compatible inference client for local and hosted LLM servers.

Changelog

0.1.3

• Added GitHub environment protection to require approval before trusted publishing to pub.dev.

0.1.2

• Added a documented release policy for version bumps, release PRs, and tag-based publishing.
• Added GitHub Actions workflows for CI verification and trusted pub.dev publishing.

0.1.1

• Added model profiles, model matching, and request descriptors for model-aware request behavior and UI introspection.
• Added request-level thinking modes, profile-driven system prompt injection, and profile-controlled multimodal content ordering.
• Added AudioPart and profile-level reasoning overrides for matched models.
• Added extraBody pass-through fields on ChatCompletionRequest and AgentOptions for server-specific top-level chat completion parameters.

0.1.0

Initial public release.

• Added a pure-Dart OpenAI-compatible inference client.
• Added chat completion requests, non-streaming responses, and SSE streaming.
• Added configurable reasoning extraction from structured fields and inline tags.
• Added typed stream events for reasoning, content deltas, tool-call deltas, usage, and completion.
• Added phase tracking for chat streams.
• Added model listing through /v1/models.
• Added typed exceptions for HTTP, timeout, network, and invalid response failures.
• Added an optional Agent layer for recursive tool-call orchestration, parallel tool execution, lifecycle events, and final-answer handling.