v0.3.0

0.3.0 (2026-07-10)

Features

• site: serve LLM-friendly markdown twins via Cloudflare worker (#11) (4569d02)
• xai: add xAI Grok provider with SuperGrok OAuth login (#8) (a8540c1)

Bug Fixes

• responses: rewrite context-overflow errors to Anthropic wording (#9) (8ef8746)

v0.2.0

0.2.0 (2026-07-10)

Features

• add GET /health healthcheck and GET / landing endpoints (#4) (3618779)
• config: standard config-file fallback chain and strict --config (#5) (66fa78b)

v0.1.0

0.1.0 (2026-07-09)

Features

• add M0 pass-through Anthropic Messages gateway (bacda61)
• add shunt token subcommand for Claude subscription apiKeyHelper (7309006)
• auth: add inbound client tokens for shared gateways (M4) (fc6f085)
• default count_tokens to tiktoken for responses providers (75f0c43)
• M1 — Anthropic Messages <-> OpenAI Responses translation (4ec674d)
• M2 — codex/chatgpt provider via reused ChatGPT OAuth (ac92b9d)
• M3 — GET /v1/models discovery endpoint (c31982f)
• map output_config.effort to responses reasoning.effort (119c08b)
• opt-in tiktoken count_tokens for responses providers (de3b6d6)
• responses: render tool_reference blocks as loaded-tool text (ef9e70b)
• responses: round-trip reasoning and enrich request/response mapping (#2) (acdc0cd)
• short-circuit count_tokens for responses-routed models (a28e281)
• sse: inject keepalive pings on idle streams (M5) (4091fa9)
• support gpt-5.6 codex slugs and their max reasoning level (8fee803)

Bug Fixes

• forward prompt token usage so context shows for Responses models (f6f524b)
• map system-role messages to developer for the responses backend (c591a1c)
• responses: drop max_output_tokens for the ChatGPT/Codex backend (2522ede)
• responses: forward upstream Retry-After through mapped errors (65b6acc)
• surface upstream error detail from the responses backend (86d8c8f)

---

This PR was generated with Release Please. See documentation.