Release v2.3.0 · antinomyhq/forgecode

Changes

feat(provider): add glm-5.1 and glm-5-turbo models @amitksingh1490 (#2726)
feat(provider): rename minimax host param and update provider url @amitksingh1490 (#2727)
feat: add minimax china endpoint support in minimax provider @Nalleyer (#2701)
feat(provider): add Novita AI as LLM provider @Alex-wuhu (#2687)
feat(provider): add alibaba_coding provider with Anthropic-compatible model catalog @amitksingh1490 (#2696)
feat: add .forge.toml support @tusharmath (#2685)

docs(prompt): tighten shell tool usage guidance wording @tusharmath (#2737)
fix: adjust doctor script severity levels @majiayu000 (#2734)
fix(forge_config): fix f32/f64 TOML serialization noise with Decimal newtype @tusharmath (#2733)
fix(bedrock): use message signature for reasoning text @amitksingh1490 (#2729)
fix: correct Bedrock model ID for Claude Opus 4.6 @majiayu000 (#2723)
fix(responses): add message phase support and preserve reasoning @amitksingh1490 (#2722)
fix(skills): avoid treating skill content as a template @daninfpj (#2703)
fix(open_ai): openai_responses_compatible full endpoint handling @lidaxian121 (#2712)
fix(ui): use bullet glyph for better Windows terminal compatibility @tusharmath (#2720)
fix(forge_main): strip trailing slash from base URL during OpenAICompatible login; fixes #2715 @mishl-dev (#2718)
fix: add model-change tracking for reasoning @amitksingh1490 (#2710)
fix(ui): scope model selection to activated provider during onboarding @tusharmath (#2691)
fix: add enable-ansi-support for windows compatibility @amitksingh1490 (#2700)
fix(forge_config): make global config file optional @tusharmath (#2699)
fix(forge_repo): retry database pool creation and connection access @amitksingh1490 (#2608)
fix(fetch): reject binary responses and harden HTML sniffing @amitksingh1490 (#2690)

perf: lazily initialize Handlebars and SyntaxHighlighter to reduce startup time @tusharmath (#2739)