laravel/ai

What's Changed

• Skip provider tools the provider does not support by @yognevoy in #1080
• Support self-hosted Laya through the TypeSafe driver by @mysticseagull in #1081
• [1.x] Add Cohere text generation support by @mssayari in #1078
• Share the provider capability check in AiManager by @pushpak1300 in #1064
• Document Boost and Laravel MCP 1.0 upgrade requirements by @pushpak1300 in #1087
• Update provider default models to the latest releases by @pushpak1300 in #1095
• Add Agent Skills support by @pushpak1300 in #1054
• Allow built-in tools to require approval by @pushpak1300 in #1097
• Send web search location to OpenRouter by @pushpak1300 in #1098
• Add Azure tool search and Groq web search and code execution by @pushpak1300 in #1099
• Add a MoveFile filesystem tool by @yognevoy in #1092
• fix: resolve stateless OpenAI file search replay by @owenvoke in #1089
• [ 1.x ] Fix: Stop an edited useChat user message from being replaced by the assistant reply message by @vinitkadam03 in #1076
• Add HasSkills contract that gives agents the LoadSkill tool by @pushpak1300 in #1102
• Send ElevenLabs query parameters in the query string by @Taldres in #1103

New Contributors

• @mysticseagull made their first contribution in #1081
• @owenvoke made their first contribution in #1089
• @Taldres made their first contribution in #1103

Full Changelog: v1.0.1...v1.1.0

What's Changed

• Remove unused and redundant provider code by @pushpak1300 in #1063
• Announce a tool call before its approval request by @SantosVilanculos in #1070
• Share local and stored file content through traits by @pushpak1300 in #1066
• Add Ai::build() for on-demand provider configuration by @pushpak1300 in #1071
• Add a Collection decide macro for choosing an item by @taylorotwell in #1073
• Migration from Mockery to Double by @jasonmccreary in #1061
• Improve remote file fetching by @pushpak1300 in #1082
• Settle pending calls for tools that cannot request approval by @pushpak1300 in #1083

New Contributors

• @SantosVilanculos made their first contribution in #1070
• @jasonmccreary made their first contribution in #1061

Full Changelog: v1.0.0...v1.0.1

What's Changed

• Sync Base Brance by @pushpak1300 in #898
• Sync 0.x by @pushpak1300 in #907
• Sync 0.x by @pushpak1300 in #916
• Sync 0.x by @pushpak1300 in #922
• Add withProviderOptions support to the image builder by @pushpak1300 in #899
• Add withProviderOptions support to the audio builder by @pushpak1300 in #900
• Add withProviderOptions support to the reranking builder by @pushpak1300 in #901
• Add an ai_sdk_extra_headers provider option for custom HTTP headers by @pushpak1300 in #904
• Add AG-UI protocol streaming by @pushpak1300 in #913
• Rename Agent User Interaction protocol by @pushpak1300 in #948
• Allow runtime tool overrides via withTools by @pushpak1300 in #956
• [1.x] Stop proxies buffering native streamed responses by @pascalbaljet in #960
• [1.x] Add a continueOrStart() shortcut by @pascalbaljet in #962
• [1.x] Report the conversation rows a turn wrote by @pascalbaljet in #961
• Scope continueLastConversation to the agent by @pushpak1300 in #972
• [1.x] Report the persisted message id on RUN_FINISHED by @pascalbaljet in #976
• [1.x] Report how a tool call ended on TOOL_CALL_RESULT by @pascalbaljet in #977
• Support hosted code execution as a provider tool by @pushpak1300 in #970
• Exclude cached tokens from the reported prompt token count on the Groq and OpenAI-compatible providers by @sulimanbenhalim in #909
• Accept every storable image type in OpenAI image edits by @saifulferoz in #893
• Fix Gemini file import operation handling by @drewmt in #821
• [1.x] Report the prompt's stored row on RUN_FINISHED by @pascalbaljet in #989
• [1.x] Report a step's text as one message, not one per content block by @pascalbaljet in #988
• [0.x] Combine streamed text by step rather than by message ID by @pascalbaljet in #987
• [1.x] Persist the reasoning a turn produced by @pascalbaljet in #975
• [1.x] Report why a resume was rejected on RUN_ERROR by @pascalbaljet in #978
• [0.x] Add conversation pagination by @pascalbaljet in #990
• [0.x] Persist the sources a streamed answer cited by @pascalbaljet in #991
• Simplify provider options and stream protocol plumbing by @pushpak1300 in #963
• Convert AG-UI run input into agent chat input by @pushpak1300 in #964
• [0.x] Expose the tool calls a paused conversation is waiting on by @pascalbaljet in #992
• [0.x] Expose whether a conversation belongs to a participant by @pascalbaljet in #993
• Add AG-UI approval interrupts and snapshot hydration to streams by @pushpak1300 in #965
• [1.x] Wrap each generation step in agent middleware by @pushpak1300 in #979
• [1.x] Thread request timeouts through reranking, image, and transcription by @pushpak1300 in #994
• Stream sub-agent activity by default by @gracjankubicki in #802
• Read Cohere Bedrock embeddings input token count from response header by @serpentblade in #914
• Read the cumulative usage reported on Anthropic message_delta by @goatrenterguy in #996
• Fix xAI and OpenRouter dropping filename for unnamed documents by @yognevoy in #1001
• Add OpenRouter reranking support by @ilogus in #999
• Unescape slashes when encoding structured MCP tool output by @yognevoy in #997
• feat: add audio input for OpenRouter by @davidafsoty in #998
• Make the AWS SDK an optional dependency of the Bedrock provider by @pushpak1300 in #1003
• [1.x] Store a user message before a provider is chosen by @pascalbaljet in #1005
• Replay earlier tool steps ahead of a paused step's provider blocks by @pushpak1300 in #995
• fix(anthropic): report thinking tokens in usage by @DGarbs51 in #1006
• Hoist every top level Gemini request field out of generationConfig by @pushpak1300 in #1015
• Use PHP_EOL for multi-line string assertion in test by @GeorgeBetts in #1016
• Add coverage for cached and reasoning tokens in xAI response usage by @dfinchenko in #1008
• Add coverage for the cache token counters on Anthropic responses by @dfinchenko in #1009
• Add support for inline document attachment for Mistral Provider by @GeorgeBetts in #1017
• [1.x] Return all generated images from xAI when n is greater than one by @mssayari in #1019
• Populate response reasoning on non-streamed prompts by @pushpak1300 in #1020
• Report inclusive input and output tokens on the usage value object by @pushpak1300 in #1021
• Add TypeSafe AI classification by @pushpak1300 in #1010
• Remove the ai_sdk_extra_headers provider option by @pushpak1300 in #1026
• Preserve Gemini thought signatures across persisted tool conversations by @sulimanbenhalim in #837
• [1.x] Report usage on audio responses by @pushpak1300 in #1029
• [1.x] Report embeddings usage as a usage object by @pushpak1300 in #1028
• Retain OpenAI replay blocks regardless of the store setting by @pushpak1300 in #1034
• [1.x] Split text-only usage counts into a TextUsage subclass by @pushpak1300 in #1030
• [1.x] Report image token details on image responses by @pushpak1300 in #1024
• [1.x] Report the transcribed audio duration on transcription responses by @pushpak1300 in #1025
• [1.x] Report usage on reranking responses by @pushpak1300 in #1027
• Store conversation turns as steps by @pushpak1300 in #1035
• Replay provider reasoning state only within the open turn by @pushpak1300 in #1036
• Persist provider-hosted tool calls on each step by @pushpak1300 in #1032
• Add OpenRouter file gateway by @bram-pkg in #1043
• Resolve approval results without matching the paused turn's participant by @pushpak1300 in #1046
• Fold an approved resume into the turn it paused on by @pushpak1300 in #1045
• Add OpenRouter classification gateway by @pushpak1300 in #1044
• Replace the turn's approval timestamp with a message status column by @pushpak1300 in #1050
• Keep an OpenAI function call item id only beside its reasoning by @pushpak1300 in #1052
• Persist the steps a turn completed before it failed by @pushpak1300 in #1051
• Add a decide string macro for boolean classification by @pushpak1300 in #1049
• Move Gemini to the Interactions API by @pushpak1300 in #1056
• Default OpenAI to GPT-6 and Anthropic smartest to Opus 5.5 by @pushpak1300 in #1058
• Add the 1.0 upgrade guide by @pushpak1300 in #1002

New Contributors

• @pascalbaljet made their first contribution in #960
• @saifulferoz made their first contribution in #893
• @goatrenterguy made their first contribution in #996
• @yognevoy made their first contribution in #1001
• @ilogus made their first contribution in #999
• @davidafsoty made their first contribution in #998
• @GeorgeBetts made their first contribution in #1016
• @mssayari made their first contribution in #1019
• @bram-pkg made their first contribution in #1043

Full Changelog: v0.11.2...v1.0.0

What's Changed

• Copy uploaded files to a managed temp path so their contents survive the request by @sulimanbenhalim in #959
• Preserve tool result failure status across conversations by @drewmt in #812
• Fix invalid Mistral cheapest text model default by @kachelle in #980

New Contributors

• @drewmt made their first contribution in #812

Full Changelog: v0.11.1...v0.11.2

What's Changed

• Align the Vercel UI message stream with the data stream protocol by @pushpak1300 in #906
• Allow caching embeddings individually by @timacdonald in #905
• Add v0.11 upgrade guide by @pushpak1300 in #897
• Fix Vercel provider tool events and finish reasons by @pushpak1300 in #921
• Emit citation events for Gemini grounding results while streaming by @sulimanbenhalim in #912
• Emit citation events for xAI web search results while streaming by @sulimanbenhalim in #911
• Add audio (TTS) support to the Mistral provider by @SanderMuller in #926
• Fail over on Groq's flex tier capacity exceeded status code by @sulimanbenhalim in #925
• Honor the Strict attribute for structured output on the Groq, Mistral, OpenAI-compatible, and OpenRouter providers by @sulimanbenhalim in #923
• Add explicit prompt cache breakpoints for Anthropic and Bedrock by @pushpak1300 in #869
• Exclude cached tokens from the reported prompt token count on the OpenRouter provider by @sulimanbenhalim in #924
• Update default provider models by @pushpak1300 in #951
• Store uploaded files from their path instead of base64 encoding them by @lazerg in #945
• Fetch and inline remote image, document, and audio attachments for the Gemini provider by @sulimanbenhalim in #927
• Return null for a remote file mime type when the response declares no content type by @sulimanbenhalim in #953
• Send the ElevenLabs transcription language hint as the documented language_code field by @sulimanbenhalim in #954
• Map the Mistral model_length and error finish reasons to Length and Error by @sulimanbenhalim in #955
• Added Bedrock Reranking Gateway by @F1nnG in #949
• Emit citation events for OpenAI web search results while streaming by @sulimanbenhalim in #910
• Document agents and provider tools in tools method docblocks by @pushpak1300 in #958
• Support validating tool request arguments by @pushpak1300 in #968
• Document attaching MCP servers to agents by @pushpak1300 in #971

Full Changelog: v0.11.0...v0.11.1

What's Changed

• Add coverage for persisting provider content blocks on paused turns by @dfinchenko in #861
• Add coverage for storing generated audio on a filesystem disk by @dfinchenko in #863
• Add coverage for storing a generated image on a filesystem disk by @dfinchenko in #862
• Trigger failover when a provider connection fails by @JVillator0 in #781
• Fail over on transient gateway and Cloudflare upstream errors, not only 503 by @sulimanbenhalim in #810
• Execute then(...) callback when faking the queued generation of transcription/image/audio/embeddings by @gdebrauwer in #797
• Add web search support to the xAI provider by @sulimanbenhalim in #857
• Add transcription support to the OpenAI-compatible provider by @sulimanbenhalim in #858
• Add transcription support to the Groq provider by @sulimanbenhalim in #868
• Fail over when Anthropic rejects a request for hitting a usage limit by @oddvalue in #864
• Reconstruct S3 document attachments in File::fromArray by @sulimanbenhalim in #878
• Handle Mistral responses that return content as a list of blocks by @lazerg in #866
• Reset text part state after TextEnd in Responses API streams by @chrisnetonline in #854
• Fix OpenAI gateway dropping cache write tokens from usage by @aqu1les in #846
• Add hosted tool search via ToolSearch tool and stateless output replay (OpenAI + Anthropic) by @behzadsp in #697
• Repair unknown local tool calls by @pushpak1300 in #859
• Make stream errors throw instead of silently ending the run by @pushpak1300 in #870
• Thread one invocation id through an entire agent run by @pushpak1300 in #871
• Report tool invocations through a run context by @pushpak1300 in #872
• Report each generation step through the run context by @pushpak1300 in #873
• Report tool failures and how long a tool ran by @pushpak1300 in #874
• Link an agent invoked as a tool back to its parent by @pushpak1300 in #875
• Report a run's terminal failure by @pushpak1300 in #876
• Filter partially-orphaned tool calls when replaying conversation history by @filipposallemi in #758
• [BUG] [0.x] Sum streamed usage across tool-call steps in all provider gateways by @mo-inkhan in #698
• Map Anthropic refusal and model_context_window_exceeded stop reasons by @egantz in #881
• Update Gemini default text model to gemini-3.7-flash by @pushpak1300 in #887
• Keep the MCP request bound while converting generator results by @pushpak1300 in #888
• Add support for the OpenRouter web fetch server tool by @CamilleScholtz in #889
• Send text/plain for all text document sources on Anthropic by @calrloco in #886
• Fail over on Anthropic transient gateway and Cloudflare errors by @sulimanbenhalim in #884
• Strip markdown code fences from OpenAI-compatible structured output by @sulimanbenhalim in #879
• Add file search support to the xAI provider by @timmcleod in #894
• Added assertPromptedTimes test method by @F1nnG in #891
• Surface Anthropic web fetch citations on the response by @sulimanbenhalim in #892
• Exclude DeepSeek cache-hit tokens from the reported prompt token count by @sulimanbenhalim in #896

New Contributors

• @gdebrauwer made their first contribution in #797
• @oddvalue made their first contribution in #864
• @lazerg made their first contribution in #866
• @chrisnetonline made their first contribution in #854
• @aqu1les made their first contribution in #846
• @behzadsp made their first contribution in #697
• @filipposallemi made their first contribution in #758
• @mo-inkhan made their first contribution in #698
• @egantz made their first contribution in #881
• @CamilleScholtz made their first contribution in #889
• @calrloco made their first contribution in #886
• @F1nnG made their first contribution in #891

Full Changelog: v0.10.3...v0.11.0

What's Changed

• Change minimum required AWS SDK version by @stevencsimpson in #808
• Add Vercel protocol coverage for text and reasoning stream events by @dfinchenko in #845
• Add Gemini tool-loop coverage for thought signature preservation by @sulimanbenhalim in #834
• Azure OpenAI - Add web search support by @jesse-bos in #843
• Parse Cohere Embed v4's type-keyed embeddings shape on Bedrock by @serpentblade in #725
• Expose raw http response in model API response by @dumbbellcode in #714
• Add embeddings support to the OpenAI-compatible provider by @ademola-emmanuel in #841
• Add Vercel protocol coverage for citation, error, and default tool error parts by @dfinchenko in #849
• Support custom HTTP headers via provider connection config by @AngadSethi in #838
• Run the test suite in parallel by @pushpak1300 in #851
• Preserve header casing when merging configured headers by @pushpak1300 in #852

New Contributors

• @stevencsimpson made their first contribution in #808
• @jesse-bos made their first contribution in #843
• @serpentblade made their first contribution in #725
• @AngadSethi made their first contribution in #838

Full Changelog: v0.10.2...v0.10.3

What's Changed

• Preserve conversation title language by @yuters in #814
• Bump actions/checkout from 7.0.0 to 7.0.1 in the github-actions group by @dependabot[bot] in #830
• Fix Anthropic structured output rejecting JSON Schema constraints by @sulimanbenhalim in #718
• Bump postcss from 8.5.16 to 8.5.23 in /workbench by @dependabot[bot] in #831
• Fix colliding embeddings cache keys for distinct string batches by @ademola-emmanuel in #832
• Replace deprected DeepSeek model name with deepseek-v4-flash by @peter-nassef in #829
• Fix: SimilaritySearch tool output with additional JSON flags (for other languages) by @yuters in #820

New Contributors

• @yuters made their first contribution in #814
• @ademola-emmanuel made their first contribution in #832
• @peter-nassef made their first contribution in #829

Full Changelog: v0.10.1...v0.10.2

• [0.x] Update .gitattributes by @jackbayliss in #806
• Update Gemini default models by @pushpak1300 in #807

• Add Rector and Laravel Rector, apply automated refactors by @pushpak1300 in #784
• Support multimodal for Gemini and VoyageAI by @paulinevos in #798
• Bedrock: support cross-account AssumeRole credential provider by @amjosuesanchez in #566
• Add summarize macro for text summarization by @pushpak1300 in #800
• Add human-in-the-loop by @pushpak1300 in #773
• Store conversation participants polymorphically by @pushpak1300 in #795

What's Changed

• Remove dead dall-e-3 image integration dataset by @mrdzen in #779
• Forward attachment provider options to OpenAI text generation requests by @mrdzen in #768
• Add tool choice support for Gemini, OpenAI, and Anthropic by @JVillator0 in #780
• Add tests for the image portrait and custom size helpers by @dfinchenko in #783
• Add interactive agent console command for the workbench by @pushpak1300 in #786
• Add web fetch tool coverage for the Anthropic provider by @dfinchenko in #792
• [0.x] Validate blank base64 input consistently across file classes by @Button99 in #788
• Add coverage for iterating the embeddings response by @dfinchenko in #790
• Add web search user location coverage for the Anthropic provider by @dfinchenko in #791
• FixTextDelta::combine glues multi-step streamed text together mid-sentence by @crishoj in #789
• Fix empty Bedrock stream content blocks by @iAmKevinMcKee in #785

New Contributors

• @crishoj made their first contribution in #789
• @iAmKevinMcKee made their first contribution in #785

Full Changelog: v0.9.0...v0.9.1

Upgrade guide: https://github.com/laravel/ai/blob/0.x/UPGRADE.md

• Add missing return type generics to embeddings and reranking responses by @dfinchenko in #701
• Allow disabling the embeddings cache per request by @pushpak1300 in #706
• Bump shivammathur/setup-php from 2.37.1 to 2.37.2 in the github-actions group by @dependabot[bot] in #708
• Use native Anthropic structured outputs by default by @paix26875 in #707
• Add collection generics to TextResponse properties by @dfinchenko in #709
• Merge all plain-object anyOf branches instead of keeping only the first by @pushpak1300 in #703
• Add file and store support for Azure OpenAI by @markusjwetzel in #676
• Add collection generics to AssistantMessage toolCalls by @dfinchenko in #720
• Extract recursive tool loop into TextGenerationLoop and convert OpenAI by @vinitkadam03 in #652
• Add collection generics to ToolResultMessage toolResults by @dfinchenko in #735
• Migrate Anthropic gateway to StepTextGateway via TextGenerationLoop by @vinitkadam03 in #726
• Migrate xAI gateway to StepTextGateway via TextGenerationLoop by @vinitkadam03 in #728
• Migrate Groq gateway to StepTextGateway via TextGenerationLoop by @pushpak1300 in #729
• Convert Ollama text gateway to TextGenerationLoop by @pushpak1300 in #730
• Convert DeepSeek to TextGenerationLoop and StepTextGateway by @pushpak1300 in #731
• Convert Gemini gateway to the shared TextGenerationLoop by @pushpak1300 in #732
• Convert OpenRouter text gateway to TextGenerationLoop by @pushpak1300 in #733
• Convert Mistral to TextGenerationLoop by @pushpak1300 in #736
• Migrate Bedrock text gateway onto TextGenerationLoop by @pushpak1300 in #734
• Support native anyOf schemas by @dbpolito in #710
• Inline single-use step traits into Gemini and OpenRouter gateways by @pushpak1300 in #741
• Add collection generics to streamed agent response events by @dfinchenko in #744
• Add collection generics to image, transcription, and meta responses by @dfinchenko in #745
• Bump actions/checkout from 6.0.3 to 7.0.0 in the github-actions group by @dependabot[bot] in #747
• Route top-level Ollama provider options out of the options object by @cjmcintyre in #722
• Remove redundant global use statements in OpenRouter transcription test by @pushpak1300 in #751
• Add withProviderOptions() to storable files by @mohali-id in #717
• Add filesystem tools for AI agents by @pushpak1300 in #651
• Fix stored tool conversation replay order by @gracjankubicki in #621
• fix: strip markdown code fences from structured output before decoding by @vaishnavyogesh in #754
• Add web search and citation support to OpenRouterProvider by @pietervanleuven in #738
• Default Anthropic text model to Claude Sonnet 5 by @pushpak1300 in #756
• Standardize withProviderOptions across the provider option API by @pushpak1300 in #753
• Add Workbench setup with a chat example for manual testing by @pushpak1300 in #762
• Fix OpenRouter test for new provider options API by @mrdzen in #766
• Add configurable OpenAI-compatible provider by @pushpak1300 in #742
• Add WithoutBroadcasting attribute to skip oversized stream events by @pushpak1300 in #752
• Support PHPStan static analysis at level 2 by @pushpak1300 in #755
• Remove TextGateway contract in favor of StepTextGateway by @pushpak1300 in #743

Full Changelog: v0.8.1...v0.9.0

What's Changed

• Normalize MCP tool schemas into the deserializable JSON Schema subset by @pushpak1300 in #696

Full Changelog: v0.8.0...v0.8.1

What's Changed

• Fix Mistral transcription requests by @NoelDeMartin in #671
• Bedrock support for remote s3 document by @dumbbellcode in #476
• Migrate OpenAI web_search_preview to web_search with provider options by @SergiiKhrystenko in #662
• Add Dependabot cooldown of 5 days by @nunomaduro in #685
• Enable Dependabot auto-merge by @nunomaduro in #687
• Add contains method to QueuedAgentPrompt by @balboacodes in #678
• Recursively remove additionalProperties for Gemini tools by @Edward144 in #682
• Fix transcription usage mapping for output tokens by @Anoop-Kadachi in #637
• Fix ImageResponse::toHtml() data URI when image mime is null by @Anoop-Kadachi in #677
• Pass the original prompt to AgentStreamed when a stream is short-circuited by @FayazK in #684
• Omit null header and multipart values for Laravel 13 compatibility by @pushpak1300 in #688
• Add MCP client and server tool support by @pushpak1300 in #655
• Bump actions/checkout from 6.0.2 to 6.0.3 in the github-actions group by @dependabot[bot] in #692
• Support OpenAI Zero Data Retention via encrypted reasoning by @Junveloper in #622

New Contributors

• @SergiiKhrystenko made their first contribution in #662
• @balboacodes made their first contribution in #678
• @Edward144 made their first contribution in #682
• @FayazK made their first contribution in #684

Full Changelog: v0.7.2...v0.8.0

What's Changed

• Revert "Apply model attribute and explicit model to all providers in a failover list" by @taylorotwell in #667
• Set claude-opus-4-8 as Anthropic's smartest text model by @pushpak1300 in #670

Full Changelog: v0.7.1...v0.7.2

What's Changed

• Set gemini-3.5-flash as Gemini's smartest text model by @pushpak1300 in https://github.com/laravel/ai/pull/640
• BedrockTextGateway: Add cache usage by @liambenson in https://github.com/laravel/ai/pull/642
• Set invocation id on fake stream events by @Anoop-Kadachi in https://github.com/laravel/ai/pull/635
• Apply model attribute and explicit model to all providers in a failover list by @sumaiazaman in https://github.com/laravel/ai/pull/633
• Add JSON_UNESCAPED_UNICODE to schema instruction encoding by @sumaiazaman in https://github.com/laravel/ai/pull/644
• Updates deprecated gemini model to successor by @nicolasbuch in https://github.com/laravel/ai/pull/657
• Fix Gemini image generation when text part precedes image data by @pushpak1300 in https://github.com/laravel/ai/pull/658
• Update deprecated Gemini default to gemini-3.5-flash by @pushpak1300 in https://github.com/laravel/ai/pull/661

New Contributors

• @liambenson made their first contribution in https://github.com/laravel/ai/pull/642
• @nicolasbuch made their first contribution in https://github.com/laravel/ai/pull/657

Full Changelog: https://github.com/laravel/ai/compare/v0.7.0...v0.7.1

• Reject empty reranking document lists by @Anoop-Kadachi in https://github.com/laravel/ai/pull/563
• Support closure provider options in queued embeddings by @Anoop-Kadachi in https://github.com/laravel/ai/pull/562
• Add PHPStan by @pushpak1300 in https://github.com/laravel/ai/pull/554
• Make OpenAI strict mode opt-in via Strict attribute by @pushpak1300 in https://github.com/laravel/ai/pull/530
• Add providerOptions to Transcription API by @ihxnnxs in https://github.com/laravel/ai/pull/31
• Validate empty inputs across generation entry points by @Anoop-Kadachi in https://github.com/laravel/ai/pull/564
• Add CHANGELOG and update-changelog workflow by @pushpak1300 in https://github.com/laravel/ai/pull/567
• Add unit test for Strict attribute by @mosabbirrakib in https://github.com/laravel/ai/pull/570
• Add test for Anthropic 529 overloaded response by @mosabbirrakib in https://github.com/laravel/ai/pull/574
• Document @throws on Image::of() and Embeddings::for() by @mosabbirrakib in https://github.com/laravel/ai/pull/576
• Add test for structured agent without Strict sends strict false by @mosabbirrakib in https://github.com/laravel/ai/pull/573
• Add unit test for HandlesFailoverErrors trait by @mosabbirrakib in https://github.com/laravel/ai/pull/572
• Add tests for Anthropic insufficient credits patterns by @mosabbirrakib in https://github.com/laravel/ai/pull/575
• Retrieve conversation list from ConversationStore and agent trait by @barryvdh in https://github.com/laravel/ai/pull/236
• Pin GitHub Actions to commit SHAs and add Dependabot config by @joetannenbaum in https://github.com/laravel/ai/pull/579
• Add generic types to conversation model relationships by @mosabbirrakib in https://github.com/laravel/ai/pull/582
• Document @throws on AzureOpenAiGateway::generateImage() by @mosabbirrakib in https://github.com/laravel/ai/pull/585
• Respect configured database connection in Conversation and ConversationMessage models by @sumaiazaman in https://github.com/laravel/ai/pull/589
• Reject blank or non-string embeddings inputs by @Anoop-Kadachi in https://github.com/laravel/ai/pull/590
• Consistent provider key resolution in provider options by @dumbbellcode in https://github.com/laravel/ai/pull/586
• Use consistent X generation wording by @mosabbirrakib in https://github.com/laravel/ai/pull/584
• Reject blank or non-string reranking documents by @Anoop-Kadachi in https://github.com/laravel/ai/pull/591
• Rehydrate attachments when loading conversation history by @sumaiazaman in https://github.com/laravel/ai/pull/587
• Reject blank model or column in SimilaritySearch::usingModel by @Anoop-Kadachi in https://github.com/laravel/ai/pull/592
• Reject blank image size and quality by @Anoop-Kadachi in https://github.com/laravel/ai/pull/593
• Document @throws RuntimeException on file content() methods by @mosabbirrakib in https://github.com/laravel/ai/pull/601
• Revert "Reject blank image size and quality" by @pushpak1300 in https://github.com/laravel/ai/pull/603
• Reject blank path in LocalDocument and StoredDocument constructors by @mosabbirrakib in https://github.com/laravel/ai/pull/600
• Add tests for blank file path constructor validation by @Anoop-Kadachi in https://github.com/laravel/ai/pull/604
• Bump PHPStan to level 1 and fix type errors by @pushpak1300 in https://github.com/laravel/ai/pull/608
• Throw InsufficientCreditsException for OpenRouter and DeepSeek credit errors by @sumaiazaman in https://github.com/laravel/ai/pull/606
• Capture missing usage details across OpenAI-shaped providers by @sumaiazaman in https://github.com/laravel/ai/pull/610
• Bump shivammathur/setup-php from 2.37.0 to 2.37.1 in the github-actions group by @dependabot[bot] in https://github.com/laravel/ai/pull/615
• Reject blank URL in RemoteDocument, RemoteImage, and RemoteAudio constructors by @sumaiazaman in https://github.com/laravel/ai/pull/613
• Pass providerOptions through to Bedrock Converse and capture reasoning content by @Junveloper in https://github.com/laravel/ai/pull/611
• Remove image+text restriction on openrouter image generation by @Themodem in https://github.com/laravel/ai/pull/595
• Document @throws LogicException on OpenAiGateway::generateTranscription by @mosabbirrakib in https://github.com/laravel/ai/pull/626

• Delete update-changelog.yml by @pushpak1300 in https://github.com/laravel/ai/pull/525
• Add error handling tests for VoyageAI reranking by @mosabbirrakib in https://github.com/laravel/ai/pull/527
• Add error handling tests for OpenAI image generation by @mosabbirrakib in https://github.com/laravel/ai/pull/528
• Add error handling tests for VoyageAI embeddings by @mosabbirrakib in https://github.com/laravel/ai/pull/526
• Handle Anthropic pause_turn server-tool continuations by @CodeWrap in https://github.com/laravel/ai/pull/493
• Add sub-agent support as tools by @JVillator0 in https://github.com/laravel/ai/pull/348
• Make conversation/message table names configurable by @timmcleod in https://github.com/laravel/ai/pull/484
• Add error handling tests for image generation (Gemini, AzureOpenAI, OpenRouter) by @mdalikadar in https://github.com/laravel/ai/pull/545
• Add rate limit and overloaded error tests for ElevenLabs audio by @mosabbirrakib in https://github.com/laravel/ai/pull/542
• Add rate limit and overloaded error tests for Cohere reranking by @mosabbirrakib in https://github.com/laravel/ai/pull/540
• Add rate limit and overloaded error tests for Jina reranking by @mosabbirrakib in https://github.com/laravel/ai/pull/541
• Add tests for Gemini image generation safety blocks by @mosabbirrakib in https://github.com/laravel/ai/pull/547
• Add error handling tests for embeddings (OpenAI, Gemini) by @mdalikadar in https://github.com/laravel/ai/pull/546
• Fix parameter name in GeneratesEmbeddings PHPDoc by @mosabbirrakib in https://github.com/laravel/ai/pull/549
• Remove stale @param and unused Request import in Vercel protocol trait by @mosabbirrakib in https://github.com/laravel/ai/pull/551
• Fix nonexistent ImageAttachment type in image-related PHPDoc by @mosabbirrakib in https://github.com/laravel/ai/pull/550
• Add error handling tests for AzureOpenAI image generation by @mosabbirrakib in https://github.com/laravel/ai/pull/537
• Mark nullable size and quality docblocks across image gateways by @mosabbirrakib in https://github.com/laravel/ai/pull/552
• Broadcast stream_failed event when BroadcastAgent job fails by @sumaiazaman in https://github.com/laravel/ai/pull/536
• Gemini context cache support via cachedContent param in agent provider options by @dumbbellcode in https://github.com/laravel/ai/pull/556
• Fix: DeepSeek 400 Bad Request on multi-turn tool-calls by preserving reasoning_content by @k1rana in https://github.com/laravel/ai/pull/534
• OpenRouter TTS and STT support by @ondrejehrlich in https://github.com/laravel/ai/pull/559
• Support provider options in embeddings by @dumbbellcode in https://github.com/laravel/ai/pull/555
• Enable failover during stream iteration by @kachelle in https://github.com/laravel/ai/pull/279

What's Changed

• Add coding standards workflow by @pushpak1300 in https://github.com/laravel/ai/pull/487
• Bind AiManager as singleton to preserve driver extensions across jobs by @ejtmicroventures in https://github.com/laravel/ai/pull/411
• fix(anthropic): preserve provider content blocks on assistant replay (refs #298) by @CodeWrap in https://github.com/laravel/ai/pull/392
• Drop redundant tool_config from Gemini text requests by @pushpak1300 in https://github.com/laravel/ai/pull/490
• Update AssistantMessage.php by @pushpak1300 in https://github.com/laravel/ai/pull/491
• Add GitHub community health files and workflows by @pushpak1300 in https://github.com/laravel/ai/pull/488
• Add error handling tests for Mistral embeddings by @sumaiazaman in https://github.com/laravel/ai/pull/495
• Add error handling tests for Ollama embeddings by @sumaiazaman in https://github.com/laravel/ai/pull/496
• Add diarize and error handling tests for Mistral transcription by @sumaiazaman in https://github.com/laravel/ai/pull/494
• Add error handling tests for Azure OpenAI embeddings by @sumaiazaman in https://github.com/laravel/ai/pull/497
• Add error handling tests for OpenAI transcription by @sumaiazaman in https://github.com/laravel/ai/pull/501
• Add error handling tests for OpenRouter embeddings by @sumaiazaman in https://github.com/laravel/ai/pull/498
• Add error handling tests for xAI image generation by @sumaiazaman in https://github.com/laravel/ai/pull/502
• Add toArray and jsonSerialize to StructuredStep by @plusemon in https://github.com/laravel/ai/pull/504
• Bedrock : support Base64 image for BedrockTextGateway by @DenneulinThomas in https://github.com/laravel/ai/pull/500
• [0.x] Add image generation support to AzureOpenAiProvider by @maherelgamil in https://github.com/laravel/ai/pull/505
• Fix OpenAI citation deduplication by URL by @Anoop-Kadachi in https://github.com/laravel/ai/pull/509
• Remove URL overrides from providers that rarely need them by @pushpak1300 in https://github.com/laravel/ai/pull/492
• Preserve OpenAI url citation span indices and stop deduping by @pushpak1300 in https://github.com/laravel/ai/pull/510
• Add Gemini TTS and STT support by @allurco in https://github.com/laravel/ai/pull/81
• Allow configuring the database connection for DatabaseConversationStore by @sumaiazaman in https://github.com/laravel/ai/pull/469
• Align xAI URL citations with OpenAI response parsing by @Anoop-Kadachi in https://github.com/laravel/ai/pull/513
• Add missing type declaration to ToolResult $result property by @mahfuz-rahman007 in https://github.com/laravel/ai/pull/518
• Test Anthropic streaming tool_use finish reason without tool blocks by @Anoop-Kadachi in https://github.com/laravel/ai/pull/515
• Add missing type hint to $quality parameter in defaultImageOptions by @mahfuz-rahman007 in https://github.com/laravel/ai/pull/516
• Add missing parameter name in EmbeddingsResponse PHPDoc by @mahfuz-rahman007 in https://github.com/laravel/ai/pull/517
• Add config option to disable LLM-generated conversation titles by @sumaiazaman in https://github.com/laravel/ai/pull/474
• Sync conversation metadata after streamed responses by @dhrupo in https://github.com/laravel/ai/pull/434
• Expand streaming finish-reason parity coverage across providers by @Anoop-Kadachi in https://github.com/laravel/ai/pull/521
• Expand OpenAI-family streaming finish reason parity tests by @Anoop-Kadachi in https://github.com/laravel/ai/pull/523

New Contributors

• @ejtmicroventures made their first contribution in https://github.com/laravel/ai/pull/411
• @plusemon made their first contribution in https://github.com/laravel/ai/pull/504
• @DenneulinThomas made their first contribution in https://github.com/laravel/ai/pull/500
• @Anoop-Kadachi made their first contribution in https://github.com/laravel/ai/pull/509
• @allurco made their first contribution in https://github.com/laravel/ai/pull/81
• @mahfuz-rahman007 made their first contribution in https://github.com/laravel/ai/pull/518
• @dhrupo made their first contribution in https://github.com/laravel/ai/pull/434

Full Changelog: https://github.com/laravel/ai/compare/v0.6.6...v0.6.7

What's Changed

• Add transcription feature tests for OpenAI provider by @sumaiazaman in https://github.com/laravel/ai/pull/468
• Add audio feature tests for OpenAI provider by @sumaiazaman in https://github.com/laravel/ai/pull/471
• Fix reasoning data lost when restoring conversation history by @Cbrad24 in https://github.com/laravel/ai/pull/301
• Add top_p configuration option for text generation requests by @Duetro in https://github.com/laravel/ai/pull/306
• Allow tools to declare their name dynamically by @seankndy in https://github.com/laravel/ai/pull/420
• feat: enhance assistant message mapping and extend tool call attributes by @mohali-id in https://github.com/laravel/ai/pull/461
• Pass through provider specific Gemini image aspect ratios by @morcken in https://github.com/laravel/ai/pull/282
• Add image generation support to OpenRouterProvider by @billyfranklim1 in https://github.com/laravel/ai/pull/333
• Throw helpful error when ai.default config is an array by @JoshSalway in https://github.com/laravel/ai/pull/327
• [0.x] Fix conversational message iterable handling in text generation by @Button99 in https://github.com/laravel/ai/pull/345
• Surface usage tokens in OpenAI and Gemini image responses by @maherelgamil in https://github.com/laravel/ai/pull/482
• Reindex tool calls and results before persisting to prevent JSON object encoding by @sumaiazaman in https://github.com/laravel/ai/pull/473

New Contributors

• @Cbrad24 made their first contribution in https://github.com/laravel/ai/pull/301
• @Duetro made their first contribution in https://github.com/laravel/ai/pull/306
• @seankndy made their first contribution in https://github.com/laravel/ai/pull/420
• @mohali-id made their first contribution in https://github.com/laravel/ai/pull/461
• @morcken made their first contribution in https://github.com/laravel/ai/pull/282
• @billyfranklim1 made their first contribution in https://github.com/laravel/ai/pull/333
• @JoshSalway made their first contribution in https://github.com/laravel/ai/pull/327

Full Changelog: https://github.com/laravel/ai/compare/v0.6.5...v0.6.6

What's Changed

• fix(bedrock): bedrock converse API parameter edge cases by @dumbbellcode in https://github.com/laravel/ai/pull/463
• fix(bedrock): incorrect name format in user attachments for bedrock provider by @dumbbellcode in https://github.com/laravel/ai/pull/465
• fix(bedrock): silent tool call failures by @dumbbellcode in https://github.com/laravel/ai/pull/464
• Support configurable base URL for Jina provider by @sumaiazaman in https://github.com/laravel/ai/pull/470
• Support configurable base URL for ElevenLabs provider by @sumaiazaman in https://github.com/laravel/ai/pull/472
• Add embedding feature tests for OpenAI provider by @sumaiazaman in https://github.com/laravel/ai/pull/467
• fix(openai, anthropic): cast empty tool args to object by @CodeWrap in https://github.com/laravel/ai/pull/419

Full Changelog: https://github.com/laravel/ai/compare/v0.6.4...v0.6.5

What's Changed

• Add tool mapping test suite for DeepSeek provider by @sumaiazaman in https://github.com/laravel/ai/pull/440
• Add provider options test suite for DeepSeek provider by @sumaiazaman in https://github.com/laravel/ai/pull/439
• Add message mapping test suite for DeepSeek provider by @sumaiazaman in https://github.com/laravel/ai/pull/438
• Fix duplicate helper function causing fatal test bootstrap error by @pushpak1300 in https://github.com/laravel/ai/pull/443
• use getter functions in toArray method of file classes by @GigaGiorgadze in https://github.com/laravel/ai/pull/436
• fix: detect LocalImage MIME type across native gateways by @maherelgamil in https://github.com/laravel/ai/pull/430
• fix: only send explicitly set image options in OpenAI provider by @navneetkumar-pim-webkul in https://github.com/laravel/ai/pull/415
• Add toAudio macro to Stringable by @nhedger in https://github.com/laravel/ai/pull/264
• Adding agent methods for text generation options by @MaestroError in https://github.com/laravel/ai/pull/198
• Add configurable URL support and base URL tests for Cohere provider by @sumaiazaman in https://github.com/laravel/ai/pull/452
• Add error handling feature tests for ElevenLabs, Cohere, Jina, and VoyageAi by @sumaiazaman in https://github.com/laravel/ai/pull/448
• Add image generation feature tests for xAI provider by @sumaiazaman in https://github.com/laravel/ai/pull/447
• Add embedding feature tests for Gemini provider by @sumaiazaman in https://github.com/laravel/ai/pull/450
• Add base URL feature tests for VoyageAi provider by @sumaiazaman in https://github.com/laravel/ai/pull/451
• Align Bedrock image and error handling with AWS API by @pushpak1300 in https://github.com/laravel/ai/pull/442
• Add image generation feature tests for Gemini provider by @sumaiazaman in https://github.com/laravel/ai/pull/449
• fix(bedrock): bedrock text gateway response messages and steps by @dumbbellcode in https://github.com/laravel/ai/pull/458
• fix(bedrock): correct default model IDs by @catatsumuri in https://github.com/laravel/ai/pull/460

New Contributors

• @GigaGiorgadze made their first contribution in https://github.com/laravel/ai/pull/436
• @navneetkumar-pim-webkul made their first contribution in https://github.com/laravel/ai/pull/415
• @MaestroError made their first contribution in https://github.com/laravel/ai/pull/198
• @dumbbellcode made their first contribution in https://github.com/laravel/ai/pull/458
• @catatsumuri made their first contribution in https://github.com/laravel/ai/pull/460

Full Changelog: https://github.com/laravel/ai/compare/v0.6.3...v0.6.4

• fix(gateway): fix DeepSeek usage stats by @aaronlei in https://github.com/laravel/ai/pull/435
• Add agent fake test suite for DeepSeek provider by @sumaiazaman in https://github.com/laravel/ai/pull/437
• Bedrock provider by @tott in https://github.com/laravel/ai/pull/270

Full Changelog: https://github.com/laravel/ai/compare/v0.6.1...v0.6.2

• Add HandlesFailoverErrors to VoyageAiGateway by @sumaiazaman in https://github.com/laravel/ai/pull/424
• Add HandlesFailoverErrors to CohereGateway and JinaGateway by @sumaiazaman in https://github.com/laravel/ai/pull/423
• Add test suite for CohereProvider (embeddings and reranking) by @maherelgamil in https://github.com/laravel/ai/pull/425
• Add test suite for ElevenLabsProvider (audio and transcription) by @maherelgamil in https://github.com/laravel/ai/pull/426
• Memoize ElevenLabs gateway instances by @maherelgamil in https://github.com/laravel/ai/pull/427
• Add test suite for JinaProvider (embeddings and reranking) by @maherelgamil in https://github.com/laravel/ai/pull/418
• Remove redundant match expressions in ElevenLabsGateway by @sumaiazaman in https://github.com/laravel/ai/pull/422
• Apply Files::put(name:) override to StorableFile inputs by @maherelgamil in https://github.com/laravel/ai/pull/428
• Remove unused prism-php/prism dependency by @pushpak1300 in https://github.com/laravel/ai/pull/414
• [0.x] Add timeout parameter to Agent contract by @Tiagospem in https://github.com/laravel/ai/pull/412

What's Changed

• fix(anthropic): cast server_tool_use.input to object by @CodeWrap in https://github.com/laravel/ai/pull/389
• Fix Mistral transcription with diarize() by @NoelDeMartin in https://github.com/laravel/ai/pull/385
• [0.x] Add feature tests into CI by @jackbayliss in https://github.com/laravel/ai/pull/384
• Update StoreTest.php by @pushpak1300 in https://github.com/laravel/ai/pull/394
• Handle nullable tool parameter types for Gemini gateway by @pushpak1300 in https://github.com/laravel/ai/pull/386
• Fix HasStructuredOutput agents crashing with empty schema by @pushpak1300 in https://github.com/laravel/ai/pull/391
• Move test fixture classes and files into tests/Fixtures/ by @pushpak1300 in https://github.com/laravel/ai/pull/397
• [0.x] Fix actions deprecation warnings by @jackbayliss in https://github.com/laravel/ai/pull/401
• Add OpenRouter gateway for direct API integration by @pushpak1300 in https://github.com/laravel/ai/pull/395
• Migrate VoyageAi off Prism to native HTTP gateway by @pushpak1300 in https://github.com/laravel/ai/pull/398
• Add Ollama gateway for direct API integration by @pushpak1300 in https://github.com/laravel/ai/pull/396
• Migrate DeepSeek provider to a native gateway by @pushpak1300 in https://github.com/laravel/ai/pull/399
• Add correct finish reason to Gemini streaming text responses by @dash8x in https://github.com/laravel/ai/pull/403
• Refine document attachment mapping and parameterize agent tests by @pushpak1300 in https://github.com/laravel/ai/pull/408
• [0.x] Fix HasTools::tools() return type to accept ProviderTool by @sumaiazaman in https://github.com/laravel/ai/pull/353
• Fix structured output and tool-call handling in native gateways by @pushpak1300 in https://github.com/laravel/ai/pull/409
• Migrate AzureOpenAiProvider from Prism to dedicated gateway by @pushpak1300 in https://github.com/laravel/ai/pull/404
• Update Anthropic smartest model default to Claude Opus 4.7 by @pushpak1300 in https://github.com/laravel/ai/pull/413

New Contributors

• @CodeWrap made their first contribution in https://github.com/laravel/ai/pull/389
• @dash8x made their first contribution in https://github.com/laravel/ai/pull/403
• @sumaiazaman made their first contribution in https://github.com/laravel/ai/pull/353

Full Changelog: https://github.com/laravel/ai/compare/v0.5.1...v0.6.0

What's Changed

• Handle provider overload errors in native gateways for failover by @pushpak1300 in https://github.com/laravel/ai/pull/378
• Fix MistralProvider ignoring injected fake gateways by @pushpak1300 in https://github.com/laravel/ai/pull/375
• Migrate test suite from PHPUnit to Pest PHP by @pushpak1300 in https://github.com/laravel/ai/pull/366
• Move Gemini file and store gateways into Gemini namespace by @pushpak1300 in https://github.com/laravel/ai/pull/382

Full Changelog: https://github.com/laravel/ai/compare/v0.5.0...v0.5.1

• Fix Mistral default transcription model by @NoelDeMartin in https://github.com/laravel/ai/pull/325
• Update prism-php/prism dependency version by @rspahni in https://github.com/laravel/ai/pull/322
• Switch Gemini structured output to response_json_schema by @ipalaus in https://github.com/laravel/ai/pull/364
• Preserve configured options in tool loop follow-up requests by @pushpak1300 in https://github.com/laravel/ai/pull/363
• Add Mistral gateway for direct API integration by @pushpak1300 in https://github.com/laravel/ai/pull/368
• Add xAI gateway for Responses API by @pushpak1300 in https://github.com/laravel/ai/pull/310
• Add missing gateway tests for Groq and OpenAI providers by @pushpak1300 in https://github.com/laravel/ai/pull/370

• Fix ProviderDocument OpenAI attachment mapping by @qwertyquest in https://github.com/laravel/ai/pull/360
• Add Gemini gateway for direct API integration by @pushpak1300 in https://github.com/laravel/ai/pull/312

• Add Groq gateway for Chat Completions API by @pushpak1300 in https://github.com/laravel/ai/pull/311
• Add direct Anthropic gateway for Messages API by @pushpak1300 in https://github.com/laravel/ai/pull/309
• Fix OpenAI strict tool parameters and persist providerOptions in tool loops by @pushpak1300 in https://github.com/laravel/ai/pull/340
• Fix missing additionalProperties on nested objects in strict schemas by @pushpak1300 in https://github.com/laravel/ai/pull/357

• Apply missing providerOptions and align maxSteps in OpenAI gateway by @pushpak1300 in https://github.com/laravel/ai/pull/338
• Add missing filename to OpenAI input_file attachment mappings by @pushpak1300 in https://github.com/laravel/ai/pull/324

• fix: use correct file extension for audio transcription uploads by @radumetes in https://github.com/laravel/ai/pull/316
• [0.x] Fix type error in fileMatchingCallback when fileId is null by @Button99 in https://github.com/laravel/ai/pull/295

• Respect configured OpenAI base URL and fixes #314 by @AnnoyingTechnology in https://github.com/laravel/ai/pull/315

• Add OpenAI gateway for Responses API by @pushpak1300 in https://github.com/laravel/ai/pull/275
• Update skill Content by @pushpak1300 in https://github.com/laravel/ai/pull/287

Full Changelog: https://github.com/laravel/ai/compare/v0.3.1...v0.3.2

• Update default OpenRouter text models to Anthropics 4.6 versions by @fosron in https://github.com/laravel/ai/pull/268
• [0.x] Remove redundant Collection wrapping in StreamEnd::combineUsage(). by @Button99 in https://github.com/laravel/ai/pull/272
• Add configurable timeout for audio requests by @nhedger in https://github.com/laravel/ai/pull/263
• [0.x] Fix file_get_contents() return value not handled if the file does not exist in Local file classes by @Button99 in https://github.com/laravel/ai/pull/267

• [0.x] Adjust AddsToolsToPrismRequestsTest by @jackbayliss in https://github.com/laravel/ai/pull/256
• Add configurable timeout for embedding requests by @nhedger in https://github.com/laravel/ai/pull/262
• Fix conversation leakage in RemembersConversations::forUser() by @Button99 in https://github.com/laravel/ai/pull/260
• [0.x] Create a CI test workflow by @jackbayliss in https://github.com/laravel/ai/pull/258
• [0.x] rename anonymous class helper methods to avoid Pint renaming by @jackbayliss in https://github.com/laravel/ai/pull/257
• Add failover support for insufficient credits and quota errors by @meirdick in https://github.com/laravel/ai/pull/249
• Fix missing audio file types in PrismMessages attachment conversion by @sulimanbenhalim in https://github.com/laravel/ai/pull/247
• Add embedding support for OpenRouter provider by @remcom in https://github.com/laravel/ai/pull/237
• Fix ErrorEvent property access in PrismStreamEvent by @Husseinadq in https://github.com/laravel/ai/pull/217
• fix: structured output schema definition by @ralphjsmit in https://github.com/laravel/ai/pull/252
• Add Ollama dimensions option to embeddings by @WoutervdBrink in https://github.com/laravel/ai/pull/190
• Fix tool call history not round-tripped in conversation context by @nicodevs in https://github.com/laravel/ai/pull/203
• [0.x] Add support for configuring provider options on agents by @shafimsp in https://github.com/laravel/ai/pull/166

• [0.x] Makes imports consistent by @nunomaduro in https://github.com/laravel/ai/pull/253

• refactor: add Lab enum as a type to the $provider parameter in agent methods in Agent contract. by @delabon in https://github.com/laravel/ai/pull/245
• refactor: remove the $options parameter from addProviderTools method calls by @delabon in https://github.com/laravel/ai/pull/244
• Fix StreamableAgentResponse signature to accept conversation user by @Button99 in https://github.com/laravel/ai/pull/246
• fix: pass outputDimension for VoyageAI embeddings by @markritterman in https://github.com/laravel/ai/pull/243
• Fix BroadcastAgent passing wrong response type to then() callbacks by @sulimanbenhalim in https://github.com/laravel/ai/pull/248

• Rename Skill by @pushpak1300 in https://github.com/laravel/ai/pull/227
• [0.x] Allow for arrays and enum for providers in the macros by @josephkerkhof in https://github.com/laravel/ai/pull/242
• [0.x] Add Message import and @return docblock to agent stubs by @josephkerkhof in https://github.com/laravel/ai/pull/240
• Renamed image to image_size in geminiprovider according to gemini api by @Jeroen-chu in https://github.com/laravel/ai/pull/235
• [0.x] Fix middleware stub naming conflict by @mt-shihab26 in https://github.com/laravel/ai/pull/234
• Improve base URL logic to prevent double path segments by @Danni2901 in https://github.com/laravel/ai/pull/226
• Fix undefined variable $e in withModelFailover when providers array is empty by @sulimanbenhalim in https://github.com/laravel/ai/pull/221
• Fix undefined variable $e in PendingResponse failover loops by @sulimanbenhalim in https://github.com/laravel/ai/pull/225

• Revert "Fix streaming generators under Octane" by @taylorotwell in https://github.com/laravel/ai/pull/220

• [0.x] Fix missing Base64Document import in Store. by @Button99 in https://github.com/laravel/ai/pull/207
• Chore: Fix Pint issues by @delabon in https://github.com/laravel/ai/pull/196
• Fix ReasoningDelta Vercel serialization to match Prism delta events by @christopherarter in https://github.com/laravel/ai/pull/202

• Add PHP 8.3 support by @pushpak1300 in https://github.com/laravel/ai/pull/212

• Add PHP 8.3 support by @pushpak1300 in https://github.com/laravel/ai/pull/212

• Make provider default models configurable via ai config by @pfrug in https://github.com/laravel/ai/pull/183
• Fix typo in comment for tool calls storagefix: correct typo in CanStr… by @wuwx in https://github.com/laravel/ai/pull/181
• Add support for timeouts in transcription by @NietThijmen in https://github.com/laravel/ai/pull/176
• [0.x] Add make:agent-middleware command by @mt-shihab26 in https://github.com/laravel/ai/pull/177
• [0.x] Allow nullable user IDs in conversation tables by @Button99 in https://github.com/laravel/ai/pull/175
• Update parameter name in EmbeddingProvider contract for consistency by @josephkerkhof in https://github.com/laravel/ai/pull/168
• [0.x] Add return type annotations to interfaces by @josephkerkhof in https://github.com/laravel/ai/pull/167
• Allow Lab enum for Provider attribute by @olavocneto in https://github.com/laravel/ai/pull/169
• Fix missing exception handling by @mubbi in https://github.com/laravel/ai/pull/161
• chore: document thrown exceptions by @mubbi in https://github.com/laravel/ai/pull/160
• Fix invalid Message role and null response in HandlesRateLimiting by @monayemislam in https://github.com/laravel/ai/pull/150
• chore: add return/parameter types by @mubbi in https://github.com/laravel/ai/pull/157
• [0.x] Ensure StreamEvent broadcast doesn't double prefix private channels by @jackbayliss in https://github.com/laravel/ai/pull/148

Full Changelog: https://github.com/laravel/ai/compare/v0.1.5...v0.2.0

Full Changelog: https://github.com/laravel/ai/compare/v0.1.4...v0.1.5

• [0.x] Fix ThinkingComplete stream event passing undefined $delta property by @jkbennemann in https://github.com/laravel/ai/pull/88
• Add missing chunking_strategy provider option for diarized transcriptions by @kachelle in https://github.com/laravel/ai/pull/87
• Return $this when calling location method in WebSearch by @lcorbett89 in https://github.com/laravel/ai/pull/84
• [0.x] Sort arguments by @lucasmichot in https://github.com/laravel/ai/pull/74
• [0x] Use new Collection and new Stringable by @lucasmichot in https://github.com/laravel/ai/pull/73
• [0.x] Use contains and doesntCountain for all assertions by @lucasmichot in https://github.com/laravel/ai/pull/70
• [0.x] Prefer ::class notation by @lucasmichot in https://github.com/laravel/ai/pull/69
• Support user string IDs by @bensherred in https://github.com/laravel/ai/pull/66
• [0.x] Normalize composer.json by @lucasmichot in https://github.com/laravel/ai/pull/68
• Add support for anonymous users by @bensherred in https://github.com/laravel/ai/pull/122
• chore(config): remove unused import from config by @monurakkaya in https://github.com/laravel/ai/pull/116
• fix: fall back to raw arguments when schema_definition key is missing by @MaximeWillinger in https://github.com/laravel/ai/pull/115
• fix: handle nullable usage in stream end events by @MaximeWillinger in https://github.com/laravel/ai/pull/114
• Add API key to provider configuration in image and audio generation by @Claudio-Emmolo in https://github.com/laravel/ai/pull/110
• Fix fromUpload() ignoring $mime parameter by @mubbi in https://github.com/laravel/ai/pull/108
• Fix PrismException::toAiException() crash by @mubbi in https://github.com/laravel/ai/pull/106
• Adds missing Tool import to stubs by @florianraith in https://github.com/laravel/ai/pull/103
• [0.x] Prefer specific PHPUnit assertions by @jackbayliss in https://github.com/laravel/ai/pull/96
• [0.x] Forward dimensions to Prism embedding requests by @alaminfirdows in https://github.com/laravel/ai/pull/97
• [0.0.x] Add array return type to toArray by @jackbayliss in https://github.com/laravel/ai/pull/95
• Fix AgentPrompted event null prompt when middleware short-circuits by @SanderMuller in https://github.com/laravel/ai/pull/100
• [0.x] Adjust assertEquals order by @jackbayliss in https://github.com/laravel/ai/pull/94
• Fix streaming generators under Octane by @csfh in https://github.com/laravel/ai/pull/118
• Add Reranking to Voyage AI by @eznix86 in https://github.com/laravel/ai/pull/63
• Add string representation to Provider to prevent fatal error in unsupported provider path by @boynet in https://github.com/laravel/ai/pull/142
• fix: add missing arguments fallback in PrismTool object code paths by @anilcancakir in https://github.com/laravel/ai/pull/140
• Fix docblock for transcribe() in GeneratesTranscriptions by @monayemislam in https://github.com/laravel/ai/pull/136
• Fix typos in AiManager DocBlocks by @jasim-k in https://github.com/laravel/ai/pull/129
• Fix undefined array key error in queued agent assertions by @mubbi in https://github.com/laravel/ai/pull/126
• Chore: Add unit test suite to phpunit.xml.dist by @delabon in https://github.com/laravel/ai/pull/124
• Add Azure OpenAI support by @imYashGupta in https://github.com/laravel/ai/pull/60
• Allow stream provider failover by @Button99 in https://github.com/laravel/ai/pull/135
• Add AiProvider enum for provider identifiers by @pushpak1300 in https://github.com/laravel/ai/pull/24
• feat: add custom base URL support for AI providers by @sadhakbj in https://github.com/laravel/ai/pull/52

• Feature/deepseek integration by @ihxnnxs in https://github.com/laravel/ai/pull/29
• [0.x] Update .gitignore by @jackbayliss in https://github.com/laravel/ai/pull/43
• Fix ToolCall arguments passed as JSON string instead of array by @nickgerrer in https://github.com/laravel/ai/pull/37
• Fix/typos param order by @CaReS0107 in https://github.com/laravel/ai/pull/35
• Allow passing connection config parameters in provider configuration (#32) by @ihxnnxs in https://github.com/laravel/ai/pull/34
• Add Model attribute by @dwightwatson in https://github.com/laravel/ai/pull/57
• Support dynamic tool names via method_exists fallback by @gracjankubicki in https://github.com/laravel/ai/pull/56
• [0.x] Add VoyageAI driver by @jackbayliss in https://github.com/laravel/ai/pull/47
• [0.x] Add Mistral AI support by @jackbayliss in https://github.com/laravel/ai/pull/40

Full Changelog: https://github.com/laravel/ai/compare/v0.1.1...v0.1.2

• [0.x] use array_any by @lucasmichot in https://github.com/laravel/ai/pull/28
• Rename index for agent conversations table by @leek in https://github.com/laravel/ai/pull/27