Setup & CLI Flags

Install and launch ACM AI Coder to experience the full ACM runtime with budgeting, policies, and resumable execution.

Install

pnpm install
pnpm --filter @ddse/acm-aicoder build

Optionally link the CLI for convenience:

pnpm --filter @ddse/acm-aicoder exec pnpm link --global

Launch the TUI

acm-aicoder \
  --provider vllm \
  --model Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 \
  --base-url http://localhost:8001/v1 \
  --workspace /path/to/project

Required flags

Flag	Description
--provider	LLM provider: vllm, ollama, openai, anthropic, etc.
--model	Provider-specific model identifier
--workspace	Root directory of the repository you want to work on

Optional flags

Flag	Description
--base-url	Override provider endpoint
--temperature	Sampling temperature (default: 0.7)
--seed	Random seed for reproducibility
--plans	Number of alternative plans to generate (1 or 2)
--resume	Continue a previous run by ID
--checkpoint-dir	Custom location for checkpoints and replay bundles

Commands inside the TUI

Command	Behaviour
/help	List available commands
/exit	Quit the session
/budget	Show budget allocations and current spend
/context	Display the active context packet
/reset	Clear the current goal and start over
#path/to/file	Reference files in the indexed workspace using mentions

Replay bundles & checkpoints

• Replay bundles are saved under .aicoder/replays/<runId>/.
• Checkpoints live in .aicoder/checkpoints/<runId>/ by default.
• Use --checkpoint-dir to store artifacts elsewhere (shared volumes, object storage).

Troubleshooting

Symptom	Fix
LLM connection errors	Verify provider flags and credentials; run curl against the base URL.
Missing workspace files	Ensure --workspace points to the repository root and is readable.
Budget blocks execution	Adjust budget configuration in config/budget.ts or supply higher limits.
Replay not generated	Check for write permissions in .aicoder/ or custom checkpoint directory.

Proceed to Task surfaces to understand the capability map behind AI Coder.