Repository map
Top level
| Path |
What it holds |
src/ |
The CLI: serve, run, repl, api-bridge, acp, and their dispatch |
crates/ |
The workspace: base, app-core, session-store, harness-api, agent-sdk, provider and platform crates |
docs/ |
Behaviour documentation, compiled into the binary and served by horon_agent_docs |
packages/sdk/typescript |
@horon-ai/agent-sdk |
packages/sdk/python |
horon-agent-sdk |
packages/npm/horon |
@horon-ai/horon launcher |
packages/docs |
This documentation site |
scripts/ |
CI gates, release and install scripts, benchmarks |
tests/ |
Workspace-level integration tests |
Documents in the repository
| Topic |
Document |
| Server, sessions, sockets |
docs/architecture/SERVER.md |
| Crate ownership, platform adapters, TLS trust |
docs/architecture/CRATES.md |
| ACP v1 and v2 |
docs/protocol/ACP.md |
| Harness API credential setup and surface |
docs/protocol/HARNESS_API.md |
| Session event storage |
docs/protocol/SESSION_STORE.md |
| Tool loading, effect tiers, edit safety |
docs/runtime/TOOLS.md |
| Tool scheduling and loop prevention |
docs/runtime/TOOL_SCHEDULING.md |
Turn recovery, mid-run insert, --resume |
docs/runtime/TURNS.md |
| Inter-agent IRC mailbox |
docs/runtime/IRC.md |
Terminal client (horon-tui) |
docs/runtime/TUI.md |
| System prompt composition |
docs/runtime/PROMPT.md |
| Provider profiles and model roles |
docs/provider/MODEL.md |
| Safety, approvals, unattended runs |
docs/safety/SAFETY.md |
| Workspace trust, crash bundles |
docs/safety/SECURITY.md |
| Benchmarks and the measurement standard |
docs/performance/BENCHMARK.md |
| Measured results |
docs/performance/RESULTS.md |
| Release assets and installation |
docs/release/INSTALL.md |
| npm packages |
docs/release/NPM.md |
| Windows build and signing |
docs/release/WINDOWS.md |
Each document ships as English X.md with a Korean X.ko.md beside it. The
English file is canonical; .ko.md files are not bundled into the binary.
Checks
| Area |
Command |
| Build |
cargo build -p horon-agent |
| Format |
cargo fmt --all -- --check |
| Lints |
cargo clippy --locked --all-targets --all-features -- -D warnings |
| Tests |
cargo test --workspace |
| Headless graph |
python3 scripts/ci/check_headless_graph.py |
| ACP adapter |
cargo test -p horon-agent --lib acp |
| Docs pairs |
python3 scripts/ci/check_docs_pairs.py |
Performance measurement runs only in GitHub Actions; the comparison suite builds
the baseline and the candidate on one runner, so a before/after difference is not
a compiler difference.