runa documentation
Local-first AI CLI — fit-check a GGUF before you download it, run it locally via ggml / llama.cpp, or hit OpenAI and Anthropic.
Local-first AI CLI — fit-check a GGUF before you download it, run it locally via ggml / llama.cpp, or hit OpenAI and Anthropic.
6 pages from pyrlyn/runa/docs at 947399f, in development
Paths into the docs
Pick where to go next.
-
Start3 pages
- Getting started runa is one CLI binary that runs AI models locally (GGUF via ggml / llama.cpp) or through the OpenAI and Anthropic APIs. Thinking controls, tools, and runa serve work the same way on both sides.
- User guide Copy-pasteable examples for the flags and behaviors. Unless noted, outputs below come from macOS CPU runs on the fixtures in tests/fixtures/. Reference docs: config.md (every key), fit.md (formulas), structured.md…
- Configuration Search order (later files win): ~/.config/runa/config.toml, then ./runa.toml. CLI flags beat RUNA_ env vars, which beat the files. Inline API keys in TOML are rejected (P3.8).
-
Concepts3 pages
- Thinking Thinking is a first-class stream, not mixed into answer text. Local and cloud backends share runa_core::ThinkConfig.
- Structured output & tools runa can force a model's answer into a JSON Schema or a GBNF grammar. Local models use llama.cpp grammar sampling, so the output always matches. OpenAI models get the schema as response_format (strict: true).
- Audio, video and vision Three routes, never mixed silently (D12).