> ## Documentation Index > Fetch the complete documentation index at: https://docs.requesty.ai/llms.txt > Use this file to discover all available pages before exploring further. # Logs & Traces > Inspect every LLM request in real time. Filter, drill into traces, compare interactions, and debug with full message visibility. Requesty logs every API request that flows through the gateway. The Logs view gives you a searchable, filterable table of every completion, plus a Traces mode that groups related requests into end-to-end flows. Click any row to open a detail panel with full message history, metadata, and performance metrics. **[Open the Logs view](https://app.requesty.ai/logs)** in the Requesty Console. Trace detail panel showing a Claude Opus multi-step agent flow ## Two View Modes Toggle between **Logs** and **Traces** using the switcher in the top-right corner. | Mode | What it shows | | ---------- | ---------------------------------------------------------------------------------------- | | **Logs** | Flat list of every individual API request, newest first | | **Traces** | Requests grouped by `trace_id` into multi-step flows (agent runs, conversations, chains) | The Traces tab appears automatically when your requests include a `trace_id`. ## Logs View The Logs table displays one row per API request with key metrics at a glance. ### Columns Configure which columns are visible using the **Columns** button. Available columns include: | Column | Description | | ----------------- | ---------------------------------------------------------- | | **Status** | Success or failure indicator | | **Time** | Timestamp of the request | | **Model** | Model that served the request (e.g. `claude-sonnet-4-5`) | | **Provider** | Underlying provider (e.g. `anthropic`, `openai`, `vertex`) | | **Tokens** | Input and output token counts | | **Cost** | Total cost in USD (router + provider) | | **Latency** | End-to-end request duration | | **Cached** | Cache hit percentage | | **Finish Reason** | How the response ended (`stop`, `length`, `tool_calls`) | | **API Key** | Which API key made the request | | **Status Code** | HTTP status code | | **Tags** | Custom tags attached to the request | | **Feedback** | User feedback data (thumbs, ratings) | | **Extra** | Custom metadata fields | Column preferences persist across sessions. ### Click-to-Filter Click any value in the table to instantly filter by that field. For example, click a model name to see only requests for that model, or click a provider to filter by provider. Active filters appear as blue chips above the table. Remove a filter by clicking the **x** on its chip, or clear all filters at once. ### Inline Filters Use the **Add Filter** button to filter by: * Model Requested / Model Used * Provider Requested / Provider Used * Status (success / failure) * API Key Name * User Email / User ID * Trace ID / Request ID * Finish Reason / Status Code * Custom feedback and extra fields Filters support **equals**, **empty**, and **not empty** modes. For fields with known values (models, providers, keys), a dropdown shows available options with request counts. ### Pagination Logs load in pages of 30. Use **Previous** and **Next** to navigate through your request history. ## Traces View Traces group multiple requests that share the same `trace_id` into a single card. This is useful for debugging multi-step agent runs, chain-of-thought flows, or any workflow that makes several LLM calls. ### Trace Cards Each trace card shows: * **Trace ID** with copy-to-clipboard * **Start time** of the first interaction * **Call count**, total tokens, total cost, and total latency * **Success rate** across all interactions * **Cache hit rate** average Click a trace card to expand it and see every interaction in the trace. Click the **eye icon** to open the full trace detail panel. ### Trace Filters Filter traces by: * **Trace ID**: Find a specific trace * **API Key Name**: Filter by which key generated the trace * **User Email**: Filter by the user who triggered the trace ### Compare Mode Select two interactions from any trace and compare them side-by-side. Enter **Compare Mode** from the traces list, select two interactions, and view a diff of their messages, tokens, cost, and latency. ## Log Detail Panel Click any row in the Logs table to open the detail panel. Navigate between logs with the **left/right arrow** buttons or keyboard arrows. Press **Esc** to close. ### Header The panel header shows at a glance: * **Status dot** (green for success, red for failure) * **Model used** * **Request ID** (copyable) * **Cost**, **latency**, **token counts**, and **timestamp** ### Messages Tab The Messages tab is a Logfire-style viewer that reconstructs the full conversation. Trace detail panel showing a Claude Haiku interaction with tool calls The left pane shows a **timeline tree** of every message in the request: * **System** messages (amber) for system prompts * **User** messages (blue) for user inputs * **Assistant** messages (purple) for model responses * **Tool calls** (green) for function/tool invocations * **Tool results** (teal) for tool outputs Click any message in the timeline to see its full content in the right pane. Tool calls display their **arguments** with syntax-highlighted JSON and the corresponding **tool result** below. Use the **role filter buttons** (user, assistant, tool\_call, system) at the top to auto-select the last message of that role. This is useful for quickly jumping to the assistant's final response or the last tool call. The timeline pane is resizable. Drag the divider to allocate more space to the tree or the content viewer. ### Details Tab The Details tab shows structured metadata organized into sections: * **Request**: Model, provider, API key, user email, timestamp, finish reason, status code * **Performance**: Input/output/cached/reasoning tokens, latency, TTFT (time to first token), cost, and cost without cache * **IDs & Tracking**: Request ID, trace ID, session ID, user ID, context ID, response ID * **Tags**: Custom tags attached via the API * **Feedback**: User feedback data (thumbs up/down, ratings, custom fields) * **Extra Fields**: Custom metadata sent with the request ### Guardrails Tab If you have [Guardrails](/features/guardrails) enabled, this tab shows any violations detected on the request, including PII detection, content policy violations, and other safety checks. ### Model Arena Click the **Arena** button in the detail panel header to replay the same request against different models. Compare cost, latency, and response quality side-by-side to find cheaper or faster alternatives. ### Raw Request Download Download the full raw HTTP request payload for any log entry using the **download** button in the header. Useful for debugging or reproducing issues locally. ## Trace Detail Panel Click the eye icon on a trace card or select a trace to open the full trace detail panel. ### Layout The trace detail panel uses a split-pane layout: * **Left pane**: List of all interactions in the trace, or a waterfall visualization * **Right pane**: Full detail view of the selected interaction (Messages, Details, Guardrails) ### Interaction List Each interaction in the left pane shows: * **Index number** (#1, #2, #3...) * **Status dot** (success/failure) * **Model name** * **Token counts** (input and output) * **Cost** and **latency** with color coding (green for fast, yellow for moderate, red for slow) Click any interaction to view its full messages and metadata in the right pane. ### Waterfall View Toggle to **Waterfall** mode using the chart icon in the left pane header. This shows a Gantt-chart-style visualization of when each interaction started and how long it took. Bars are color-coded by provider: | Provider | Color | | --------------- | ------ | | OpenAI | Green | | Anthropic | Purple | | Vertex | Blue | | Azure | Cyan | | Failed requests | Red | The waterfall includes a time axis so you can see overlapping or sequential requests and identify bottlenecks. ### Navigation * Use **left/right arrows** (or keyboard shortcuts) to navigate between traces * Press **Esc** to close the panel * The header shows total calls, tokens, cost, and latency for the entire trace ## Keyboard Shortcuts | Key | Action | | ------- | ---------------------- | | **←** | Previous log or trace | | **→** | Next log or trace | | **Esc** | Close the detail panel | ## Integration with Other Features * [Usage Analytics](/features/usage-analytics). Aggregate cost and usage metrics across all requests * [Cost Tracking](/features/cost-tracking). Per-request and per-trace cost breakdowns * [Request Metadata](/features/request-metadata). Add custom fields that appear in log details * [Request Feedback](/features/request-feedback). Attach user feedback to individual requests * [Session Reconstruction](/features/session-reconstruction). Automatic grouping of related conversations * [Guardrails](/features/guardrails). Safety checks visible in the Guardrails tab * [Analytics Headers](/features/analytics-headers). Custom tracking headers that appear in logs