An interactive CLI AI agent built with LangChain, LangGraph, and Ink. It streams model responses, invokes structured tools to read/modify files in a target project, and persists conversational history for iterative workflows.
- Requirements: Go 1.25+; macOS/Linux/WSL2
- Env: export
OPENAI_API_KEYwith your key
Build:
make local-build # installs 'cli-agent'
# or
go build -o cli-agent .Run:
go run .
# or, after build
./cli-agentDev loop:
make dev # sets GOENV=dev, builds, runsgo test ./...
go vet ./...
gofmt -s -w .
go buildIf you want to use the FastAPI server with LangChain tools:
cd server
pip install -r requirements.txt
# Run with reload in dev:
ENV=dev python main.py
# Or via uvicorn:
ENV=local uvicorn agent:app --host 127.0.0.1 --port 8080The server reads OPENAI_API_KEY from env or a .env in server/.
- Smooth hand over between non-reasoning and reasoning agents.
- Proper planning step to create todo list that is not too bloated.
MIT