Skip to main contentCrate battlecommand_forge
Source pub use mission::MissionRunner;
- benchmark
- codegen
- Multi-file code extraction from LLM responses.
- context
- Context compaction — prevents token limit crashes.
Ported from battleclaw-v2 context_compact.rs.
- cto
- CTO interactive chat agent with native tool calling.
- custom_commands
- db
- File-based mission database.
- editor
- Edit existing codebases with AI assistance.
- enterprise
- Enterprise features: audit logging, cost tracking, RBAC.
- github
- GitHub integration via
gh CLI.
Ported from battleclaw-v2. - hardware
- Full hardware monitoring — CPU, RAM, thermal, Ollama VRAM.
Ported from battleclaw-v2. Polls every 2 seconds.
- llm
- memory
- Self-improving memory system.
- mission
- model_config
- model_picker
- Interactive model picker overlay for TUI.
Ported from battleclaw-v2 model_picker.rs, adapted for forge’s 9-stage pipeline roles.
- models
- Model configuration, listing, and benchmarking.
- report
- Pipeline run report — structured JSON capture of every stage.
- router
- sandbox
- Sandboxed command execution with timeouts and environment isolation.
- secrets
- Atomic write helper for files holding sensitive content.
- snake
- space
- stress
- Stress test suite — 21 graded tasks (C4-C9) ported from battleclaw-v2.
- swarm
- swebench
- swebench_eval
- swebench_tools
- tui
- Terminal UI — 6 tabs: Chat, Queue, Models, Code, HW, Log.
Full snake easter egg, live streaming, hardware monitoring, code viewer.
Status bar with live cost, VRAM, task counter. Typewriter code animation.
- verifier
- Verification engine: static analysis, linting, test execution, security scanning.
- voice
- Voice TTS announcements using macOS
say command. - workspace
- Git workspace management for isolated mission execution.