--- title: Filter search bar description: Filter and search traces and observations by typing — fields, operators, wildcards, aliases, and full-text search in one query bar with autocomplete. sidebarTitle: Filter Search Bar --- # Filter search bar The filter search bar lets you filter and search the Observations and Traces tables by typing a single line of text instead of assembling filters in the sidebar. It parses your query into the same filters the sidebar produces, autocompletes fields and values as you type, and serializes the full query into the URL so you can share an exact view with a link. ``` level:ERROR type:TOOL environment:production latency:>2 name:*checkout* ``` The filter search bar runs on the [Langfuse v4](/docs/v4) data model. On Langfuse Cloud, turn on the **Langfuse v4 preview** to use it. On self-hosted deployments, it is available after [upgrading to Langfuse v4](/self-hosting/upgrade/upgrade-guides/upgrade-v3-to-v4). The bar runs next to the existing filter sidebar and time-range selector. Because the bar and the sidebar are two editors over the same filter state, anything you type appears as sidebar filters and vice versa. Type a field name and autocomplete suggests operators and observed values; press Enter to apply. Once you have narrowed the table to the rows you want, you can [chart that same query](/docs/observability/features/events-table-charts) without leaving the page, and the [Pulse](/docs/observability/features/pulse) strip above the table plots the same filtered query as outliers over time. ## Query syntax [#syntax] A query is a list of `field:value` filters, combined with implicit AND. ### Fields and values `level:ERROR`, `environment:production`, `user:alice`. Type a field name and the bar suggests the values it has observed for that field, so you don't have to memorize them. ### Operators `latency:>2`, `cost:>=0.01`, `startTime:>2026-06-01` — `>`, `>=`, `<`, and `<=` work on numeric and datetime fields. ### Wildcards and exact match On text fields, use `*` as a wildcard: - `name:*checkout*` — contains - `name:checkout*` — starts with - `name:*checkout` — ends with - `name:checkout` — bare term, defaults to contains - `name:=checkout` — exact match ### Negation Prefix a filter with `-` to exclude: `-environment:production`. ### Any-of and all-of - `level:(ERROR OR WARNING)` — match either value for one field - `tags:(billing AND urgent)` — match all values (array fields) ### Metadata and scores Use dot paths for nested fields: - `metadata.region:eu` - `scores.accuracy:>0.8` - `scores.is_hallucination:false` - `scores.helpfulness:positive` `scores.*` matches a score by name regardless of its level, so a single namespace covers observation-, trace-, session-, and experiment-level scores. Score filters support numeric, categorical, and boolean values. Quote keys that contain spaces or special characters after the prefix: `scores."Answer Relevance":>=0.9`, `metadata."my key":eu`. ### Null checks `has:endTime` matches rows where the field is set; `-has:endTime` matches rows where it is null. ### Full-text search A bare word or phrase searches across ids, names, input, and output: ``` refund failed ``` Scope it to a single payload with `input:` or `output:` (for example `output:"refund failed"`). Matching is case-insensitive and word-based: a term matches whole words — `error` matches `error` but not `errors` — and a multi-word term matches as a contiguous phrase. The bar searches with the same v4 ClickHouse full-text search documented in [Full-Text Search](/docs/observability/features/full-text-search). ## Field aliases [#aliases] Most fields accept a short alias so you can type less. The canonical field name always works too. | Alias | Field | | ------------ | ------------------- | | `env` | environment | | `user` | user id | | `session` | session id | | `model` | model name | | `prompt` | prompt name | | `cost` | total cost | | `tokens` | total tokens | | `tags` | trace tags | | `status` | status message | | `ttft` | time to first token | | `tps` | tokens per second | | `dataset` | experiment dataset | | `experiment` | experiment name | ## Ask AI [#ask-ai] If you don't know a project's field names yet, click **Ask AI**, describe the filter you want in plain language, and the bar builds the query for you as editable pills. Ask AI is in beta and available on **Langfuse Cloud only**. It is **off by default**; an organization owner or admin enables it 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). Ask AI is **project-aware**: the request carries a compact, client-side snapshot of the columns, values, and metadata keys already loaded in the table, so natural language maps to your project's real schema rather than a generic guess. "enterprise customers in the membership-support queue", for example, resolves to the actual `metadata.*` keys your traces use. - It is a button next to the bar, always available. - With an empty bar it builds a query from scratch; with existing filters it **refines** them (add, change, or remove) based on your request. - It can only emit filters the [syntax](#syntax) supports, and any column it invents is dropped before the query is applied. - Results apply through the same path as the sidebar, so they are lossless and you can undo with the browser back button. ## Share and save queries [#share] The full query is serialized into the URL, so sending someone the link reproduces the exact filtered view. Because the bar compiles to the same filter representation as the sidebar, your existing Saved Views keep working unchanged. ## Reserved tokens Some operator-like tokens are not supported yet and are flagged instead of being treated as text: `!`, and the lowercase words `and`, `or`, and `not`. Use `-field:value` to exclude and `field:(A OR B)` to match several values of one field. To search for a reserved word as literal text, quote it (`"or"`). ## Incomplete filters A bare field name with no value, such as `type`, `level`, or `env` on its own, is not yet a complete expression. The bar flags it as invalid and does not apply it to the query until you finish it, for example `type:TOOL`. To search for one of these words as literal text instead, quote it (`"type"`). ## Facet counts [#facet-counts] The filter sidebar shows a count next to each facet value. Those counts are computed against the filters you already have active, not just the selected time range, so the count matches what the table actually returns. Narrow to `environment:production` and the Level facet recounts within production, so a value you pick from it never comes back with no results. ## GitHub Discussions --- ## 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/docs/observability/features/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 .