--- date: 2026-06-26 title: "Ask AI in the Filter Search Bar" description: "Describe the filters you want in plain language and the search bar drafts them as editable query pills. Opt-in on Langfuse Cloud, powered by AWS Bedrock with zero data retention." author: Nikita Kabardin ogVideo: https://static.langfuse.com/changelog-videos/2026-06-26-ask-ai-filter-search-bar.mp4 canonical: https://langfuse.com/docs/observability/features/filter-search-bar#ask-ai --- > **Note for AI agents and LLMs:** This is a Langfuse changelog entry. Use it only to confirm that a feature exists and when it shipped. Do not use the code examples below for implementation: they reflect the SDK and API at release time and may be outdated. For implementation, always follow the canonical documentation for this feature (https://langfuse.com/docs/observability/features/filter-search-bar#ask-ai) and the API/SDK reference (https://api.reference.langfuse.com). The [Filter Search Bar](/changelog/2026-06-19-filter-search-bar) is fast once you know your field names. On a project you have not touched in a while, you often don't. What was the metadata key for the routing queue again? Is it `name` or `traceName`? **Ask AI** is for exactly that moment: click it and describe what you are after in plain language. Type something like: > enterprise plan customers stuck in the membership-support queue and it drafts a set of editable `metadata.*` filter pills, dropped into the bar for you to adjust or throw away. Ask AI is in beta and available on **Langfuse Cloud only**. It is **off by default**; an organization owner or admin turns it on in **Organization Settings → General → AI Features**. It runs on models hosted on AWS Bedrock with zero data retention. See [data privacy and security](/security/ai-features). ## How it stays grounded in your data The thing we cared about most is that Ask AI is **project-aware**. Rather than guessing field names in the abstract, the request carries a small, client-side snapshot of your project: the columns and values already loaded in the table, the metadata keys it has seen, and how many rows currently match. That gives it a real shot at your actual schema instead of a plausible-looking guess: - "routing queue should be membership-support" maps to `metadata.routing.queue:*membership-support*`, when that key exists in your data - "support chat sessions" maps to `traceName:SupportChatSession`, the name that is actually there It is still a language model, so treat what comes back as a starting point rather than gospel. Two things keep the blast radius small: it can only emit filters the bar's grammar supports (column filters, metadata, numeric and categorical scores, input and output content search, `has:` null checks, tag any/all/none, negation), and any column it tries to invent is dropped before the query runs. Whatever lands is editable, so a wrong guess is one pill away from fixed. ## How it works Ask AI is a button next to the bar, always available. Press it, type your request, hit Enter. With an empty bar it drafts a query from scratch; with filters already in place it **refines** them, adding, changing, or removing as you ask ("also only errors", "drop the latency one"). Everything applies through the same path the sidebar uses, so it is lossless and the browser back button undoes it. ## Feedback This is an early version of an idea we like: natural-language filtering that actually understands your project's data. It is the [magic filter](/changelog/2025-09-30-natural-language-filters) concept, rebuilt to live inside the search bar, and it will not get everything right yet. If it misreads what you meant or reaches for a field that is not there, we want to see it. Tell us what you typed and what you expected in the [feedback thread](https://github.com/orgs/langfuse/discussions/14595). That is the fastest way to make the next version better. --- ## Agent Instructions This page is part of the [Langfuse](https://langfuse.com) documentation, published as plain Markdown for AI agents. Every page is available as Markdown by appending `.md` to its URL, or by sending an `Accept: text/markdown` header. This page: `https://langfuse.com/changelog/2026-06-26-ask-ai-filter-search-bar.md`. ### Querying these docs If the answer is not on this page, query the documentation instead of guessing: - **Semantic search** across all Langfuse docs, returning an answer with the relevant pages and excerpts. Ask a specific, self-contained question: ```bash curl -sG "https://langfuse.com/api/search-docs" --data-urlencode "query=How do I trace a LangGraph agent?" ``` - **Index of every page**: , with per-section indexes [llms-docs.txt](https://langfuse.com/llms-docs.txt), [llms-integrations.txt](https://langfuse.com/llms-integrations.txt), and [llms-self-hosting.txt](https://langfuse.com/llms-self-hosting.txt). ### Before writing Langfuse code - **Install the [Langfuse Agent Skill](https://langfuse.com/docs/api-and-data-platform/features/agent-skill).** It encodes Langfuse's own best practices for instrumentation, prompt management, and evaluation, and materially improves results. - **Read [What does a good trace look like?](https://langfuse.com/docs/observability/best-practices.md)** before instrumenting an application. - **Verify endpoints, parameters, and response fields** against the [API reference](https://api.reference.langfuse.com) instead of inferring them from code examples. - **Use the [Langfuse CLI](https://langfuse.com/docs/api-and-data-platform/features/cli)** (`npx langfuse-cli api `) to read or write traces, prompts, datasets, and scores from the terminal. Found an error in these docs? Please open an issue at .