feat: add interactive configuration wizard and clap CLI

CHANGELOG.md

@@ -7,6 +7,10 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 ## [Unreleased]
 
 ### Added
+- Interactive configuration wizard via `zeph init` subcommand with 5-step setup (LLM provider, memory, channels, secrets backend, config generation)
+- clap-based CLI argument parsing with `--help`, `--version` support
+- `Serialize` derive on `Config` and all nested types for TOML generation
+- `dialoguer` dependency for interactive terminal prompts
 - Structured LLM output via `chat_typed<T>()` on `LlmProvider` trait with JSON schema enforcement (#456)
 - OpenAI/Compatible native `response_format: json_schema` structured output (#457)
 - Claude structured output via forced tool use pattern (#458)

Cargo.toml

@@ -12,6 +12,8 @@ repository = "https://github.com/bug-ops/zeph"
 
 [workspace.dependencies]
 age = { version = "0.11.2", default-features = false }
+clap = { version = "4.5", features = ["derive"] }
+dialoguer = "0.11"
 anyhow = "1.0"
 candle-core = { version = "0.9", default-features = false }
 candle-nn = { version = "0.9", default-features = false }
@@ -126,6 +128,9 @@ stt = ["zeph-llm/stt"]
 
 [dependencies]
 anyhow.workspace = true
+clap.workspace = true
+dialoguer.workspace = true
+toml.workspace = true
 tokio = { workspace = true, features = ["macros", "rt-multi-thread", "signal", "sync", "time"] }
 tokio-util.workspace = true
 tracing.workspace = true

README.md

@@ -23,30 +23,59 @@ Most AI agent frameworks are **token furnaces**. They dump every tool descriptio
 
 Zeph takes the opposite approach: **automated context engineering**. Only relevant data enters the context. Everything else is filtered, compressed, or retrieved on demand. The result — dramatically lower costs, faster responses, and an agent that runs on hardware you already have.
 
+## Installation
+
+```bash
+# From source
+cargo install --git https://github.com/bug-ops/zeph
+```
+
+Pre-built binaries for Linux, macOS, and Windows: [GitHub Releases](https://github.com/bug-ops/zeph/releases/latest) · [Docker](https://bug-ops.github.io/zeph/guide/docker.html)
+
 ## Quick Start
 
 ```bash
-git clone https://github.com/bug-ops/zeph && cd zeph
-cargo build --release
+# Interactive setup wizard — generates config.toml with provider, memory, and channel settings
+zeph init
+
+# Run the agent
+zeph
+
+# Or with TUI dashboard (requires `tui` feature)
+zeph --tui
+```
 
+Manual configuration is also supported:
+
+```bash
 # Local models — no API costs
 ollama pull mistral:7b && ollama pull qwen3-embedding
-./target/release/zeph
+zeph
 
 # Cloud providers
-ZEPH_LLM_PROVIDER=claude ZEPH_CLAUDE_API_KEY=sk-ant-... ./target/release/zeph
-ZEPH_LLM_PROVIDER=openai ZEPH_OPENAI_API_KEY=sk-... ./target/release/zeph
+ZEPH_LLM_PROVIDER=claude ZEPH_CLAUDE_API_KEY=sk-ant-... zeph
+ZEPH_LLM_PROVIDER=openai ZEPH_OPENAI_API_KEY=sk-... zeph
 
 # Any OpenAI-compatible API (Together AI, Groq, Fireworks, etc.)
 ZEPH_LLM_PROVIDER=compatible ZEPH_COMPATIBLE_BASE_URL=https://api.together.xyz/v1 \
-  ZEPH_COMPATIBLE_API_KEY=... ./target/release/zeph
+  ZEPH_COMPATIBLE_API_KEY=... zeph
 ```
 
-Pre-built binaries for Linux, macOS, and Windows: [GitHub Releases](https://github.com/bug-ops/zeph/releases/latest) · [Docker](https://bug-ops.github.io/zeph/guide/docker.html)
-
 > [!TIP]
 > Full setup walkthrough: [Installation](https://bug-ops.github.io/zeph/getting-started/installation.html) · [Configuration](https://bug-ops.github.io/zeph/getting-started/configuration.html) · [Secrets management](https://bug-ops.github.io/zeph/guide/vault.html)
 
+## CLI Usage
+
+```
+zeph                     Run the agent (default)
+zeph init                Interactive configuration wizard
+zeph init -o path.toml   Write generated config to a specific path
+zeph --tui               Run with TUI dashboard
+zeph --config <path>     Use a custom config file
+zeph --version           Print version
+zeph --help              Show help
+```
+
 ## Automated Context Engineering
 
 This is the core idea behind Zeph. Every byte that enters the LLM context window is there because it's **useful for the model** — not because the framework was too lazy to filter it.