> This readme is automatically generated by [create-ink-app](https://github.com/vadimdemedes/create-ink-app)
Typed client for models.dev to fetch model catalogs with friendly errors.
Tree-shakeable static models.dev catalog split by provider for TokenLens.
Up-to-date models.dev data as typed packages
Helpers for context windows, usage normalization, compaction, and cost estimation.
Pi extension that tracks AI spend per git commit across sessions
Pretty unicode tables for the command line. Based on the original cli-table.
Reuse OpenCode authentication & provider logic for clean AI calls in plugins
Get stdout window width, with two fallbacks, tty and then a default.
Typed AI model pricing and capabilities data, auto-updated from models.dev
Toggle the CLI cursor
LLM Icons - Icon collection for language models
TokenRouter provider extension for pi — dynamic model discovery with OpenRouter-compatible pricing
No description provided.
CLI to explore and query models.dev catalogue (non-interactive + interactive TUI).
Codex CLI is a coding agent from OpenAI that runs locally on your computer.
Spinners for use in the terminal
A command line utility to work with Sentry. https://docs.sentry.io/hosted/learn/cli/
The linux x64 distribution of the Sentry CLI binary.
Validate a webpack configuration.
Syntax highlighting in your terminal
CLI for the swc project
Vite as Node.js runtime
CLI tool for Angular
Call any LLM from the command line via models.dev
A fast TUI and CLI for browsing AI models, benchmarks, and coding agents
Headless coding-agent toolkit built on grain-agent-core + grain-agent-harness + grain-llm-genai. v1 ships the four read-only filesystem tools and a single-prompt CLI driver (`grain-headless`); write tools / shell / interactive loop / rig-backed semantic search land in follow-up PRs.
Rust port of @earendil-works/pi-agent-core: stateful agent framework with tool execution and event streaming.
Rust port of @earendil-works/pi-agent-core harness: session persistence, system-prompt assembly, truncation, custom message helpers.
ratatui-based terminal UI on top of grain-ai-agent-headless. Same coding-agent capabilities as `grain-headless` (read / write / bash / web / semantic-search tools, session persistence, skills, slash commands) but with a multi-pane terminal interface.
DeepSeek-specific extensions (R1 reasoning_content scavenge, subagent.done protocol) on top of grain-agent-core / grain-llm-genai.
genai-backed LlmStream implementation for grain-agent. Bridges grain-agent-core's transport-agnostic agent loop to the genai crate's multi-provider chat API. PR 3a: outbound mapping only — impl LlmStream, inbound mapping, and env/quirk config land in subsequent PRs.
Standardized LLM model registry for grain-agent. Embeds a vendored snapshot of model metadata (context window, capabilities, pricing, thinking profile); the `fetch` feature pulls live data from models.dev.
WebAssembly Component Model plugin runtime for grain. Loads `.wasm` plugins at runtime and exposes their tools through the `AgentTool` trait — no recompiling grain to add a plugin.
Boa-based JavaScript scripting layer for grain-agent. Loads user `.js` files from disk, runs them in an isolated boa_engine::Context owned by a dedicated worker thread, and exposes registered tools to the agent through the `grain_ai_agent_headless::Extension` trait.
Rhai-based scripting layer for grain-agent. Loads user `.rhai` files from disk, executes them in a shared sync-mode `rhai::Engine`, and exposes the tools they register to the agent. Sibling to `grain-script-boa` — both surfaces emit `Vec<Arc<dyn AgentTool>>` and a host crate (e.g. the TUI) can mix them however it likes.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.
No description provided.