feat: cua-driver agent integration + headless CLIs (Node + Rust)#7
Merged
feat: cua-driver agent integration + headless CLIs (Node + Rust)#7
Conversation
Three deliverables landing together because they share the EventSink decoupling:
## 1. cua-driver agent integration (Tauri + frontend)
Hawkeye's chat panel now has an "Agent" toggle: when on, Gemini receives a
curated 8-tool catalog (screenshot/list_windows/get_window_state/click/
type_text/press_key/scroll/launch_app) and can drive native macOS apps in
the background via trycua/cua's cua-driver — no focus theft, no cursor
movement, even on non-AX surfaces (Chromium, Blender, Figma).
- src-tauri/src/agent/{protocol,cua_driver,tools,runner,mod}.rs — Unix-socket
client speaking cua-driver's line-delimited JSON protocol; curated tools;
tool-use loop with MAX_TOOL_ROUNDS=8 and screenshot follow-up image attach
- src-tauri/src/commands/agent_cmd.rs — get_agent_status, start_agent,
chat_with_agent, invoke_cua_tool Tauri commands
- src-tauri/src/ai/{types,provider,gemini}.rs — FunctionDeclaration /
FunctionCall / ToolMessage / ToolTurn types; AiProvider::chat_with_tools
default-impl + full Gemini implementation with function_call /
function_response part translation
- src/hooks/useAgent.ts + components/ChatPanel.tsx — agent-mode toggle,
live tool-call streaming via agent:tool-call-* events, audit trail UI
- 5/5 unit tests pass for the cua-driver protocol layer
## 2. Phase 1 — Node CLI (@hawkeye/cli)
New workspace package wrapping @hawkeye/core. Zero changes to core.
- packages/cli/ — TypeScript ESM, tsup-bundled, 17 KB shebang'd dist/main.js
- 7 subcommands: init / perceive / plan / execute / run / chat / daemon
- Global --json flag for NDJSON output
- 4-layer config merge: defaults → ~/.config/hawkeye/cli.json → env → CLI args
- daemon falls back to polling pending an `observation` event in core
## 3. Phase 2 — Rust CLI (hawkeye-cli) + EventSink decoupling
The Tauri Rust backend now exposes a UI-agnostic EventSink trait so the same
observe + agent code paths run from a standalone CLI binary.
- src/event_sink.rs — EventSink trait + TauriSink / StdoutSink / NoopSink
- src/bin/cli.rs — clap-based CLI: config / observe / chat / agent / agent-status
- Cargo.toml — explicit [[bin]] hawkeye-{desktop,cli} + clap dep
- Refactor: lib.rs all `mod` → `pub mod`; AppState + event_sink field;
ObserveLoop::start and run_user_turn take Arc<dyn EventSink> instead of
AppHandle; perception::init dropped its unused AppHandle param
Both bins build clean. 5/5 lib tests pass. Tauri desktop app unchanged.
## Documentation
- HEADLESS_PLAN.md — full execution plan (Phases 1-3)
- HEADLESS.md — operations guide for both CLIs + cua-driver agent mode
- packages/desktop-tauri/AGENT_INTEGRATION.md — architecture deep-dive
- packages/cli/README.md — Node CLI usage
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Three deliverables landing together because they share the EventSink decoupling:
1. cua-driver agent integration (Tauri + frontend)
src-tauri/src/agent/{protocol,cua_driver,tools,runner}.rs,commands/agent_cmd.rs,src/hooks/useAgent.tsai/{types,provider,gemini}.rs—chat_with_tools+ Gemini function-calling2. Phase 1 — Node CLI (
@hawkeye/cli)packages/cli/workspace package wrapping@hawkeye/core, zero changes to core--jsonflag, 4-layer config merge, 17 KB ESM bundle3. Phase 2 — Rust CLI (
hawkeye-cli) + EventSink decouplingsrc/event_sink.rs(EventSinktrait +TauriSink/StdoutSink/NoopSink)src/bin/cli.rs— clap-based CLI: config / observe / chat / agent / agent-statusrun_user_turnandObserveLoop::startto takeArc<dyn EventSink>instead ofAppHandlecargo build --bin hawkeye-{desktop,cli}Verification
cargo test --lib agent::→ 5 passedDocumentation
HEADLESS.md,HEADLESS_PLAN.md— execution plan + ops guidepackages/desktop-tauri/AGENT_INTEGRATION.md— architecture deep-divepackages/cli/README.md— Node CLI usage