v0.4.0

What's Changed

• [BUG]: Fixed ToolInput macro panic by @Zahgrom34 in #204
• Fix typo in MiniMax DEFAULT_BASE_URL by @WonderLawrence in #207
• [BUG]: LlamaCpp openai compat has issue post schemers by @saivishwak in #208
• Add optional shaped Qdrant payloads with mirrored root-level fields by @SergioArrighi in #210
• [BUG]: Fix wasm compatability by @saivishwak in #212
• [FEATURE]: Add openai responses api by @saivishwak in #213
• [FEATURE]: Add codeact executor by @saivishwak in #214
• [BUG]: Fix the wasi support for core and llm by @saivishwak in #216
• [MAINT]: Add Test Cases by @saivishwak in #218
• [MAINT]: Update llammcpp version by @saivishwak in #219
• feat(llamacpp): make enable_thinking configurable via LlamaCppConfig by @kgentic-dev in #223
• [BUG]: Fix CI Issues by @saivishwak in #225
• feat(llamacpp): KV-cache prefix reuse across inference calls by @kgentic-dev in #222
• docs: add FAQ section for common questions by @meichuanyi in #226
• feat(llamacpp): per-call sampling overrides via SamplingOverrides by @kgentic-dev in #227
• feat(llamacpp): shared model constructor for multi-context deployments by @kgentic-dev in #224
• [MAINT]: Fix CI Issues by @saivishwak in #228
• feat(llm): surface usage tokens for Ollama and Google backends by @kgentic-dev in #230
• fix(llamacpp): structured output without tools (closes #220) by @kgentic-dev in #232
• feat(anthropic): wire output_config.format for structured output by @kgentic-dev in #236
• fix(llamacpp): grammar-only chat without tools returns ffi -3 (#220 follow-up) by @kgentic-dev in #240
• feat(chat): add fn model() default impl to ChatProvider + 8 backends by @kgentic-dev in #237
• docs: fix README badge typo and wording by @meichuanyi in #242
• docs: fix localized README badge typo by @meichuanyi in #243
• docs: fix typo in AGENTS guide by @meichuanyi in #244
• docs: fix typo 'availble' -> 'available' in AGENTS guidelines by @meichuanyi in #245
• [MAINT]: Stabilize main CI workflow by @saivishwak in #247
• [BUG]: Bound memory writes by @saivishwak in #262
• [BUG]: Return typed errors instead of provider panics for unsupported capabilities by @saivishwak in #263
• [BUG]: Harden Environment runtime lookup and actor streaming execution by @saivishwak in #264
• feat(protocol): add app_meta field to Task for app-provided metadata by @vootooguru in #246
• [BUG]: Add typed HTTP error mapping and retry-after semantics by @saivishwak in #265
• [BUG]: Make derive macro paths and schema failures production-safe (#256) by @saivishwak in #266
• [BUG]: Harden parse_document against SSRF and unbounded reads (#249) by @saivishwak in #267
• [FEATURE]: Add WASI Preview2 OpenAI Responses support by @bigboss2063 in #270
• [FEATURE]: Add security boundary for filesystem toolkit by @saivishwak in #272
• [BUG]: Fix CI failures by @bigboss2063 in #273
• [FEATURE] Add WASI support for model listing and embeddings in OpenAI by @bigboss2063 in #271
• [BUG]: Harden MCP process trust model by @saivishwak in #275
• [REFACTOR]: Harden the LLM secret store and playload logging by @saivishwak in #277
• [FIX]: Fix dependabot issues by @saivishwak in #278
• [CI]: Expand release readiness gates by @saivishwak in #279
• [MAINT]: Optimize CI checks by @saivishwak in #280
• [MAINT]: Add Qdrant named vector tests by @ahfoysal in #274
• [MAINT]: bump version to v0.4.0 by @saivishwak in #281

New Contributors

• @Zahgrom34 made their first contribution in #204
• @WonderLawrence made their first contribution in #207
• @kgentic-dev made their first contribution in #223
• @meichuanyi made their first contribution in #226
• @vootooguru made their first contribution in #246
• @bigboss2063 made their first contribution in #270
• @ahfoysal made their first contribution in #274

Full Changelog: v0.3.7...v0.4.0