aimee

aimee is the memory and agent substrate that Nehanda CLI forks. It provides the foundational capabilities that make the nehanda stack work: persistent memory, code indexing, delegate orchestration, and guardrails.

What it provides

Memory that compounds. Every session starts already knowing what the last one learned. A curator pipeline distills raw sessions into a typed knowledge base, dedupes facts, flags contradictions, and lets stale detail decay. The memory and code index form a hybrid vector-graph — vector recall fused with a typed knowledge graph and your code’s call graph, ranked together.

Your codebase as a map. aimee indexes your code into a symbol and call graph. The AI finds callers, traces an edit’s blast radius before it writes a line, and works from the graph instead of reading your files over again each session. The graph spans repositories, so one dependency map covers every repo you work across.

Delegates that cut the bill. Send summarization, review, and boilerplate to the cheapest model that can handle it. Local models on your GPU cost nothing. The primary agent gets a compact result back instead of raw content. aimee tracks cost and success per delegate and routes better over time.

Self-hosted inference. aimee ships a self-hosted inference stack: embeddings, reranking, and synthesis in one CPU or GPU container. The knowledge base curates entirely on your hardware with no outside API calls.

Autonomous development. Hand aimee a written proposal and it runs the job unattended: design, plan, implement, review, open the PR. The primary agent manages, a panel of models reviews the work, and delegates do the building.

Guardrails and isolation. Sensitive files like .env, private keys, and production configs are blocked before the AI can touch them. Known anti-patterns raise warnings. Planning mode freezes every write until you are ready. Each session runs in its own git worktree.

Performance

Core services are written in C, with hot paths running in single-digit milliseconds. Nothing phones home.

How Nehanda CLI uses it

Nehanda CLI is a fork of aimee that wires in the Nehanda model as the primary reasoning model. The aimee infrastructure — memory, code graph, delegates, guardrails — is preserved. The Nehanda-specific additions are the model configuration, system prompts, and the nehanda-server binary that manages the inference pipeline.

---

aichat

aichat is the interactive terminal UI (TUI) that Nehanda CLI uses as its user-facing interface. It is installed automatically by install.sh and configured to talk to nehanda-server over the loopback HTTP API.

What it provides

Multi-provider support. aichat works with OpenAI, Anthropic, Gemini, Mistral, Ollama, and any OpenAI-compatible endpoint. In the Nehanda CLI setup, it is pointed at nehanda-server on 127.0.0.1:8740.

Rich terminal UI. The nehanda-ui (built on aichat’s Ink-based TUI) shows the current agent, model, endpoint, bearer token, and live token usage. Slash commands provide quick access to configuration and status.

Sessions and roles. aichat supports persistent sessions, role-based system prompts, and context management — all leveraged by Nehanda CLI to maintain conversation state across sessions.

How Nehanda CLI uses it

The nehanda command launches nehanda-ui, which is aichat configured against nehanda-server. The install script sets up:

• The aichat binary at /opt/homebrew/bin/aichat (or equivalent)
• Configuration pointing to nehanda-server on 127.0.0.1:8740
• A workspace registration wrapper at ~/.local/bin/nehanda that captures the current working directory

The TUI passes cwd: process.cwd() in every chat request, ensuring file operations happen in the correct directory — the one where you invoked nehanda, not the installation directory.