--- title: Releasing Airflow AI Provider 1.0: Agents, Sandboxes,... description: Apache Airflow's Common AI provider has reached version 1.0 and now installs on Airflow 2.11 as well as Airflow 3. New capabilities include human-approval... canonical: https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn copyright: © 2026 Daily Dev Ltd. terms: https://daily.dev/tos/ twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Releasing Airflow AI Provider 1.0: Agents, Sandboxes, Human Approval and Smarter Retries | daily.dev og:description: Apache Airflow's Common AI provider has reached version 1.0 and now installs on Airflow 2.11 as well as Airflow 3. New capabilities include human-approval... og:url: https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn og:image: https://api.daily.dev/og/posts/kF8gTEqvN.png og:image:alt: Releasing Airflow AI Provider 1.0: Agents, Sandboxes, Human Approval and Smarter Retries og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Releasing Airflow AI Provider 1.0: Agents, Sandboxes, Human Approval and Smarter Retries **[Apache Airflow](https://daily.dev/sources/airflow)** · 29 min read · 5 upvotes · 0 comments ## Summary Apache Airflow's Common AI provider has reached version 1.0 and now installs on Airflow 2.11 as well as Airflow 3. New capabilities include human-approval gates for tool calls (waiting for a person to approve a Slack post or similar action), model-driven retry policies that let an LLM decide whether a failed task should retry, sandboxed code execution for agent-written code, Agent Skills support, durable execution that replays recorded steps on retry, batch prompting at roughly half the per-token cost, and framework-neutral tool adapters for Pydantic AI, Strands, Google ADK, OpenAI Agents SDK, Anthropic SDK, and LangChain. The release documents which features are now stable under semantic versioning versus experimental, and lists concrete migration steps from the 0.x releases. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I make an Airflow agent pause and wait for human approval before calling a tool like Slack? Mark a toolset with .approval_required() on HookToolset, which requires Airflow 3.3 or later; the task enters an awaiting_input state holding no worker slot, and the request appears on the task's Required Actions tab. Approve runs the call, while Reject does not fail the task but tells the agent the call was denied with the reviewer's reason, so it can act accordingly. Set tool_approval_timeout to fail or auto-deny after a wait. _Teams wiring approval gates into agentic pipelines can track Airflow provider changes like this on daily.dev._ ### Can I let an LLM decide whether a failed Airflow task should retry instead of using fixed retry logic? Yes, using LLMRetryPolicy or ClassifierRetryPolicy from the apache-airflow-providers-common-ai provider, available on Airflow 3.3 and later. The model reads the exception's class and message, names a failure category, and decides to retry after a delay or fail immediately; for example, a revoked API key fails immediately while a rate-limit error retries after 60 seconds. A fallback_policy or fallback_rules handles cases the model can't classify. _Developers tuning retry behavior for flaky pipelines can follow updates like this through daily.dev._ ### Does the Apache Airflow Common AI provider work on Airflow 2.11, or do I need Airflow 3? Version 1.0 of apache-airflow-providers-common-ai installs on Airflow 2.11 and later, even though Airflow 2 reached end of life on 22 April 2026. Core operators, decorators, hooks, and toolsets work the same as on Airflow 3.0, but features like tool approval that pauses a task, retry policies, HITL review, and durable execution without extra configuration require Airflow 3.1 through 3.3. _Teams planning an Airflow 2-to-3 migration while adopting AI features can keep tabs on provider support via daily.dev._ ## Similar posts on daily.dev - [Introducing the Common AI Provider: LLM and AI Agent Support for Apache Airflow](https://daily.dev/posts/introducing-the-common-ai-provider-llm-and-ai-agent-support-for-apache-airflow-0vn8ziyl5) · Apache Airflow · 0 upvotes · 0 comments - [Cloud Composer supports Apache Airflow 3.1](https://daily.dev/posts/cloud-composer-supports-apache-airflow-3-1-tfxltyyoe) · Google Cloud · 1 upvotes · 0 comments - [Apache Flink Agents 0.3.0 Release Announcement](https://daily.dev/posts/apache-flink-agents-0-3-0-release-announcement-oryt2tsr5) · Flink · 8 upvotes · 2 comments - [foojay – a place for friends of OpenJDK](https://daily.dev/posts/foojay-a-place-for-friends-of-openjdk-oouz4lj4e) · Foojay.io · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#apache-airflow](https://daily.dev/tags/apache-airflow) [View this post on daily.dev](https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Releasing Airflow AI Provider 1.0: Agents, Sandboxes, Human Approval and Smarter Retries","url":"https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn"},"datePublished":"2026-10-20T00:06:15.784Z","dateModified":"2026-10-20T00:07:05.561Z","description":"Apache Airflow's Common AI provider has reached version 1.0 and now installs on Airflow 2.11 as well as Airflow 3. New capabilities include human-approval...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/dc011ddf9adab9fd441fc8a5bb797f50?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/dc011ddf9adab9fd441fc8a5bb797f50?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Apache Airflow","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Apache Airflow","logo":"https://media.daily.dev/image/upload/s--Dqa6irgE--/f_auto/v1752573492/logos/airflow","url":"https://daily.dev/sources/airflow"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,apache-airflow","timeRequired":"PT29M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Apache Airflow","item":"https://daily.dev/sources/airflow"},{"@type":"ListItem","position":3,"name":"Releasing Airflow AI Provider 1.0: Agents, Sandboxes, Human Approval and Smarter Retries"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/releasing-airflow-ai-provider-1-0-agents-sandboxes-human-approval-and-smarter-retries-kf8gteqvn#faq","mainEntity":[{"@type":"Question","name":"How do I make an Airflow agent pause and wait for human approval before calling a tool like Slack?","acceptedAnswer":{"@type":"Answer","text":"Mark a toolset with .approval_required() on HookToolset, which requires Airflow 3.3 or later; the task enters an awaiting_input state holding no worker slot, and the request appears on the task's Required Actions tab. Approve runs the call, while Reject does not fail the task but tells the agent the call was denied with the reviewer's reason, so it can act accordingly. Set tool_approval_timeout to fail or auto-deny after a wait. Teams wiring approval gates into agentic pipelines can track Airflow provider changes like this on daily.dev."}},{"@type":"Question","name":"Can I let an LLM decide whether a failed Airflow task should retry instead of using fixed retry logic?","acceptedAnswer":{"@type":"Answer","text":"Yes, using LLMRetryPolicy or ClassifierRetryPolicy from the apache-airflow-providers-common-ai provider, available on Airflow 3.3 and later. The model reads the exception's class and message, names a failure category, and decides to retry after a delay or fail immediately; for example, a revoked API key fails immediately while a rate-limit error retries after 60 seconds. A fallback_policy or fallback_rules handles cases the model can't classify. Developers tuning retry behavior for flaky pipelines can follow updates like this through daily.dev."}},{"@type":"Question","name":"Does the Apache Airflow Common AI provider work on Airflow 2.11, or do I need Airflow 3?","acceptedAnswer":{"@type":"Answer","text":"Version 1.0 of apache-airflow-providers-common-ai installs on Airflow 2.11 and later, even though Airflow 2 reached end of life on 22 April 2026. Core operators, decorators, hooks, and toolsets work the same as on Airflow 3.0, but features like tool approval that pauses a task, retry policies, HITL review, and durable execution without extra configuration require Airflow 3.1 through 3.3. Teams planning an Airflow 2-to-3 migration while adopting AI features can keep tabs on provider support via daily.dev."}}]} ```