Chat con IA

Streaming chat UI with stop, regenerate, copy, feedback, model picker and markdown with copyable code blocks.

AI booking assistant that shows services, stylists and open slots as tappable cards in the chat and books with a calendar invite.

Rich prompt input with slash commands, file attachments, model picker, token counter and a history of sent prompts.

Revenue insights with dollar impact, confidence, sparklines and one-click Apply, Dismiss with undo, and Why explanations.

Three-pane coding agent workspace: sessions with live status, a multi-agent thread with a self-ticking plan, expandable tool calls, thinking, approve/reject for risky commands and streaming replies, plus a syntax-colored diff viewer with file tabs, patch download and Apply.

Monitoring for AI agents: runs, success rate, cost and token KPIs, a runs-per-hour chart, an incident callout with acknowledge and resolve, filters, and a sortable runs table whose rows expand into a trace waterfall.

Folder sidebar with counts and a starred filter, searchable prompt cards with highlighted {{variables}}, tags, versions and usage, a use dialog that fills variables into a live preview with copy and send, plus new prompt and delete with undo.

Baseline-vs-candidate scoreboard (score, pass rate, latency), category and regression filters, a case table with pass/fail score badges and deltas, rows that expand into a split diff of both answers with the judge's reasoning, human votes that can override a blocked promotion, and CSV export.

Spend, tokens and requests KPIs, a stacked daily cost-by-model bar chart, a monthly budget token meter with segments, editable budget and alert threshold, and a per-feature table with cost share bars and sparklines, plus CSV export.

Model picker with speed, quality and price meters, a temperature slider with plain-English hints, max tokens, a system prompt editor with tone chips, tool permission cards with risky-tool confirmation, and a test panel that streams a sample reply.

Website, PDF, sheet and Notion sources with chunk counts, sync status, animated re-indexing progress, errors with fixes, enable toggles and remove with undo, an add-source dialog and a retrieval tester that shows matched chunks with scores.

Transcript with a latency timeline where every assistant turn opens model, tokens, cost and confidence, expandable tool calls, retrieved chunks with scores and guardrail results, plus flag, add-to-evals and JSON export.

Input/output rules for card numbers, emails, medical advice, prompt injection and more with enable toggles and block/redact/warn actions, a regex rule builder with validation, and a tester that runs the rules on a message and shows the verdict, matches and redacted text.

Job list with status filters, live training progress and loss sparklines, a detail panel with an animated loss curve, metadata, deploy-to-assistant, cancel with confirmation, retry for failures and a log console, plus a new-job dialog with a cost estimate.