A prompt input with attachments, a model picker, and a stop state.
A prompt input with attachments, a model picker, and a stop state.
Chat composer with an auto-growing textarea, Enter to send, attachments, a model picker, and a Stop button while streaming. Ships composable parts.
A chat composer: a growing message field, file attachments with previews, a tools menu (web search, deep research, image), and a send button that turns into stop while the answer streams.
Example usage of Prompt Input.
Example usage of Prompt Input.
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.
Saved prompt cards in a filterable grid showing title, description, category badge, and use count with a create prompt affordance.
Six proven prompts as cards with categories, usage counts, and mono opening lines.
Check a prompt before it ships, with pass rate by version against the bar it has to clear, a unified or side-by-side diff of what changed, and test cases that re-run live from queued to a rolling score.
Example usage of Questionnaire.
Quick reply buttons for common chat responses.
Streaming grounded RAG over its OWN text/PDF corpus — token-by-token answers through the engine-owned generator (granite-4.0-350M) with [n] markers upgraded to hover-card inline citations resolved to retrieved sources (PDF page attribution + similarity), wall-clock duration, and cancel — carrying the core-pipeline ⇄ LangChain engine toggle. Split from ui/blocks/knowledge-base (Ask tab RAG lane; absorbs langchain-rag).
Example usage of Reaction Picker.
Component which provides a realtime chat interface.
Component which provides a realtime chat interface.
A reply that shows how long it thought, folds its reasoning, runs two tool calls from running to done and cites its sources.
Example usage of Reasoning.
Example usage of Reasoning.
Supervise a coding agent and its subagents: set its autonomy, balance effort between subagents, hold the dead man's switch, and lift the cover to ship. A whole screen for a Next.js app, at /agent-console.
A Corvale research agent: a question starts a live timeline of thinking, searching and reading in the thread while the memo writes itself section by section in the panel beside it, a sheet on phones, then the answer cites numbered sources that light up and a request to share waits for Allow or Deny. Plug your agent into onSend, onApprove and onDeny.
Full agent app built from the 8 agent primitives: the agent thinks, searches and reads your workspace on a live activity timeline, writes a memo into a side panel, answers with citations, then asks before it shares the memo. Scripted end to end, so it runs with no backend until you plug in your model.
A tool-using ReAct research agent over its OWN on-device WebGPU-only WebLLM model (default Qwen3-1.7B-q4f16_1-MLC): three block-local tools (search over a bundled 8-article knowledge base, note, calculate), a step-by-step reasoning timeline, and a human-in-the-loop tool-approval gate (default on) with pause/approve/deny-feedback and immutable decision receipts. Ships a justified block-local WebLLM catalog slice (model-catalog.ts) and toolset + fixture corpus (tools.ts). Models download only behind an explicit Load action. Split from ui/blocks/agent-structured-data (Research Agent mode).
A research plan, live search activity, sources, and a cited synthesis.
A shared prompt panel above two side by side response cards, each with a model badge, answer body, latency and token stats, and a prefer button.
Showing 385–408 of 468